diff options
| author | Jonathan Gray <jsg@cvs.openbsd.org> | 2018-10-23 06:36:00 +0000 |
|---|---|---|
| committer | Jonathan Gray <jsg@cvs.openbsd.org> | 2018-10-23 06:36:00 +0000 |
| commit | b65fcab046d3a1b6b6ac315720df220925c5322e (patch) | |
| tree | ff73dcc383ac0799c655ff6194cda9dacb75dde9 /lib/mesa/src/gallium/drivers/freedreno/a5xx | |
| parent | 18d6381c51e253e4c41c62619f80d9ce745b95c8 (diff) | |
Merge Mesa 17.3.9
Mesa 18.x needs an ld with build-id for at least the intel code
Mesa 18.2 assumes linux only memfd syscalls in intel code
Tested by matthieu@, kettenis@ and myself on a variety of hardware and
architectures. ok kettenis@
Diffstat (limited to 'lib/mesa/src/gallium/drivers/freedreno/a5xx')
27 files changed, 10569 insertions, 0 deletions
diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/a5xx.xml.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/a5xx.xml.h new file mode 100644 index 000000000..ae946d81b --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/a5xx.xml.h @@ -0,0 +1,4896 @@ +#ifndef A5XX_XML +#define A5XX_XML + +/* Autogenerated file, DO NOT EDIT manually! + +This file was generated by the rules-ng-ng headergen tool in this git repository: +http://github.com/freedreno/envytools/ +git clone https://github.com/freedreno/envytools.git + +The rules-ng-ng source files this header was generated from are: +- /home/ilia/src/freedreno/envytools/rnndb/adreno/a5xx.xml ( 141938 bytes, from 2017-07-08 01:02:47) +- /home/ilia/src/freedreno/envytools/rnndb/freedreno_copyright.xml ( 1572 bytes, from 2016-02-11 01:04:14) +- /home/ilia/src/freedreno/envytools/rnndb/adreno/adreno_common.xml ( 13324 bytes, from 2017-07-04 02:59:47) +- /home/ilia/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml ( 31866 bytes, from 2017-07-04 02:59:47) + +Copyright (C) 2013-2017 by the following authors: +- Rob Clark <robdclark@gmail.com> (robclark) +- Ilia Mirkin <imirkin@alum.mit.edu> (imirkin) + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ + + +enum a5xx_color_fmt { + RB5_A8_UNORM = 2, + RB5_R8_UNORM = 3, + RB5_R8_SNORM = 4, + RB5_R8_UINT = 5, + RB5_R8_SINT = 6, + RB5_R4G4B4A4_UNORM = 8, + RB5_R5G5B5A1_UNORM = 10, + RB5_R5G6B5_UNORM = 14, + RB5_R8G8_UNORM = 15, + RB5_R8G8_SNORM = 16, + RB5_R8G8_UINT = 17, + RB5_R8G8_SINT = 18, + RB5_R16_UNORM = 21, + RB5_R16_SNORM = 22, + RB5_R16_FLOAT = 23, + RB5_R16_UINT = 24, + RB5_R16_SINT = 25, + RB5_R8G8B8A8_UNORM = 48, + RB5_R8G8B8_UNORM = 49, + RB5_R8G8B8A8_SNORM = 50, + RB5_R8G8B8A8_UINT = 51, + RB5_R8G8B8A8_SINT = 52, + RB5_R10G10B10A2_UNORM = 55, + RB5_R10G10B10A2_UINT = 58, + RB5_R11G11B10_FLOAT = 66, + RB5_R16G16_UNORM = 67, + RB5_R16G16_SNORM = 68, + RB5_R16G16_FLOAT = 69, + RB5_R16G16_UINT = 70, + RB5_R16G16_SINT = 71, + RB5_R32_FLOAT = 74, + RB5_R32_UINT = 75, + RB5_R32_SINT = 76, + RB5_R16G16B16A16_UNORM = 96, + RB5_R16G16B16A16_SNORM = 97, + RB5_R16G16B16A16_FLOAT = 98, + RB5_R16G16B16A16_UINT = 99, + RB5_R16G16B16A16_SINT = 100, + RB5_R32G32_FLOAT = 103, + RB5_R32G32_UINT = 104, + RB5_R32G32_SINT = 105, + RB5_R32G32B32A32_FLOAT = 130, + RB5_R32G32B32A32_UINT = 131, + RB5_R32G32B32A32_SINT = 132, +}; + +enum a5xx_tile_mode { + TILE5_LINEAR = 0, + TILE5_2 = 2, + TILE5_3 = 3, +}; + +enum a5xx_vtx_fmt { + VFMT5_8_UNORM = 3, + VFMT5_8_SNORM = 4, + VFMT5_8_UINT = 5, + VFMT5_8_SINT = 6, + VFMT5_8_8_UNORM = 15, + VFMT5_8_8_SNORM = 16, + VFMT5_8_8_UINT = 17, + VFMT5_8_8_SINT = 18, + VFMT5_16_UNORM = 21, + VFMT5_16_SNORM = 22, + VFMT5_16_FLOAT = 23, + VFMT5_16_UINT = 24, + VFMT5_16_SINT = 25, + VFMT5_8_8_8_UNORM = 33, + VFMT5_8_8_8_SNORM = 34, + VFMT5_8_8_8_UINT = 35, + VFMT5_8_8_8_SINT = 36, + VFMT5_8_8_8_8_UNORM = 48, + VFMT5_8_8_8_8_SNORM = 50, + VFMT5_8_8_8_8_UINT = 51, + VFMT5_8_8_8_8_SINT = 52, + VFMT5_10_10_10_2_UNORM = 54, + VFMT5_10_10_10_2_SNORM = 57, + VFMT5_10_10_10_2_UINT = 58, + VFMT5_10_10_10_2_SINT = 59, + VFMT5_11_11_10_FLOAT = 66, + VFMT5_16_16_UNORM = 67, + VFMT5_16_16_SNORM = 68, + VFMT5_16_16_FLOAT = 69, + VFMT5_16_16_UINT = 70, + VFMT5_16_16_SINT = 71, + VFMT5_32_UNORM = 72, + VFMT5_32_SNORM = 73, + VFMT5_32_FLOAT = 74, + VFMT5_32_UINT = 75, + VFMT5_32_SINT = 76, + VFMT5_32_FIXED = 77, + VFMT5_16_16_16_UNORM = 88, + VFMT5_16_16_16_SNORM = 89, + VFMT5_16_16_16_FLOAT = 90, + VFMT5_16_16_16_UINT = 91, + VFMT5_16_16_16_SINT = 92, + VFMT5_16_16_16_16_UNORM = 96, + VFMT5_16_16_16_16_SNORM = 97, + VFMT5_16_16_16_16_FLOAT = 98, + VFMT5_16_16_16_16_UINT = 99, + VFMT5_16_16_16_16_SINT = 100, + VFMT5_32_32_UNORM = 101, + VFMT5_32_32_SNORM = 102, + VFMT5_32_32_FLOAT = 103, + VFMT5_32_32_UINT = 104, + VFMT5_32_32_SINT = 105, + VFMT5_32_32_FIXED = 106, + VFMT5_32_32_32_UNORM = 112, + VFMT5_32_32_32_SNORM = 113, + VFMT5_32_32_32_UINT = 114, + VFMT5_32_32_32_SINT = 115, + VFMT5_32_32_32_FLOAT = 116, + VFMT5_32_32_32_FIXED = 117, + VFMT5_32_32_32_32_UNORM = 128, + VFMT5_32_32_32_32_SNORM = 129, + VFMT5_32_32_32_32_FLOAT = 130, + VFMT5_32_32_32_32_UINT = 131, + VFMT5_32_32_32_32_SINT = 132, + VFMT5_32_32_32_32_FIXED = 133, +}; + +enum a5xx_tex_fmt { + TFMT5_A8_UNORM = 2, + TFMT5_8_UNORM = 3, + TFMT5_8_SNORM = 4, + TFMT5_8_UINT = 5, + TFMT5_8_SINT = 6, + TFMT5_4_4_4_4_UNORM = 8, + TFMT5_5_5_5_1_UNORM = 10, + TFMT5_5_6_5_UNORM = 14, + TFMT5_8_8_UNORM = 15, + TFMT5_8_8_SNORM = 16, + TFMT5_8_8_UINT = 17, + TFMT5_8_8_SINT = 18, + TFMT5_L8_A8_UNORM = 19, + TFMT5_16_UNORM = 21, + TFMT5_16_SNORM = 22, + TFMT5_16_FLOAT = 23, + TFMT5_16_UINT = 24, + TFMT5_16_SINT = 25, + TFMT5_8_8_8_8_UNORM = 48, + TFMT5_8_8_8_UNORM = 49, + TFMT5_8_8_8_8_SNORM = 50, + TFMT5_8_8_8_8_UINT = 51, + TFMT5_8_8_8_8_SINT = 52, + TFMT5_9_9_9_E5_FLOAT = 53, + TFMT5_10_10_10_2_UNORM = 54, + TFMT5_10_10_10_2_UINT = 58, + TFMT5_11_11_10_FLOAT = 66, + TFMT5_16_16_UNORM = 67, + TFMT5_16_16_SNORM = 68, + TFMT5_16_16_FLOAT = 69, + TFMT5_16_16_UINT = 70, + TFMT5_16_16_SINT = 71, + TFMT5_32_FLOAT = 74, + TFMT5_32_UINT = 75, + TFMT5_32_SINT = 76, + TFMT5_16_16_16_16_UNORM = 96, + TFMT5_16_16_16_16_SNORM = 97, + TFMT5_16_16_16_16_FLOAT = 98, + TFMT5_16_16_16_16_UINT = 99, + TFMT5_16_16_16_16_SINT = 100, + TFMT5_32_32_FLOAT = 103, + TFMT5_32_32_UINT = 104, + TFMT5_32_32_SINT = 105, + TFMT5_32_32_32_UINT = 114, + TFMT5_32_32_32_SINT = 115, + TFMT5_32_32_32_FLOAT = 116, + TFMT5_32_32_32_32_FLOAT = 130, + TFMT5_32_32_32_32_UINT = 131, + TFMT5_32_32_32_32_SINT = 132, + TFMT5_X8Z24_UNORM = 160, + TFMT5_ETC2_RG11_UNORM = 171, + TFMT5_ETC2_RG11_SNORM = 172, + TFMT5_ETC2_R11_UNORM = 173, + TFMT5_ETC2_R11_SNORM = 174, + TFMT5_ETC1 = 175, + TFMT5_ETC2_RGB8 = 176, + TFMT5_ETC2_RGBA8 = 177, + TFMT5_ETC2_RGB8A1 = 178, + TFMT5_DXT1 = 179, + TFMT5_DXT3 = 180, + TFMT5_DXT5 = 181, + TFMT5_RGTC1_UNORM = 183, + TFMT5_RGTC1_SNORM = 184, + TFMT5_RGTC2_UNORM = 187, + TFMT5_RGTC2_SNORM = 188, + TFMT5_BPTC_UFLOAT = 190, + TFMT5_BPTC_FLOAT = 191, + TFMT5_BPTC = 192, + TFMT5_ASTC_4x4 = 193, + TFMT5_ASTC_5x4 = 194, + TFMT5_ASTC_5x5 = 195, + TFMT5_ASTC_6x5 = 196, + TFMT5_ASTC_6x6 = 197, + TFMT5_ASTC_8x5 = 198, + TFMT5_ASTC_8x6 = 199, + TFMT5_ASTC_8x8 = 200, + TFMT5_ASTC_10x5 = 201, + TFMT5_ASTC_10x6 = 202, + TFMT5_ASTC_10x8 = 203, + TFMT5_ASTC_10x10 = 204, + TFMT5_ASTC_12x10 = 205, + TFMT5_ASTC_12x12 = 206, +}; + +enum a5xx_tex_fetchsize { + TFETCH5_1_BYTE = 0, + TFETCH5_2_BYTE = 1, + TFETCH5_4_BYTE = 2, + TFETCH5_8_BYTE = 3, + TFETCH5_16_BYTE = 4, +}; + +enum a5xx_depth_format { + DEPTH5_NONE = 0, + DEPTH5_16 = 1, + DEPTH5_24_8 = 2, + DEPTH5_32 = 4, +}; + +enum a5xx_blit_buf { + BLIT_MRT0 = 0, + BLIT_MRT1 = 1, + BLIT_MRT2 = 2, + BLIT_MRT3 = 3, + BLIT_MRT4 = 4, + BLIT_MRT5 = 5, + BLIT_MRT6 = 6, + BLIT_MRT7 = 7, + BLIT_ZS = 8, + BLIT_Z32 = 9, +}; + +enum a5xx_cp_perfcounter_select { + PERF_CP_ALWAYS_COUNT = 0, + PERF_CP_BUSY_GFX_CORE_IDLE = 1, + PERF_CP_BUSY_CYCLES = 2, + PERF_CP_PFP_IDLE = 3, + PERF_CP_PFP_BUSY_WORKING = 4, + PERF_CP_PFP_STALL_CYCLES_ANY = 5, + PERF_CP_PFP_STARVE_CYCLES_ANY = 6, + PERF_CP_PFP_ICACHE_MISS = 7, + PERF_CP_PFP_ICACHE_HIT = 8, + PERF_CP_PFP_MATCH_PM4_PKT_PROFILE = 9, + PERF_CP_ME_BUSY_WORKING = 10, + PERF_CP_ME_IDLE = 11, + PERF_CP_ME_STARVE_CYCLES_ANY = 12, + PERF_CP_ME_FIFO_EMPTY_PFP_IDLE = 13, + PERF_CP_ME_FIFO_EMPTY_PFP_BUSY = 14, + PERF_CP_ME_FIFO_FULL_ME_BUSY = 15, + PERF_CP_ME_FIFO_FULL_ME_NON_WORKING = 16, + PERF_CP_ME_STALL_CYCLES_ANY = 17, + PERF_CP_ME_ICACHE_MISS = 18, + PERF_CP_ME_ICACHE_HIT = 19, + PERF_CP_NUM_PREEMPTIONS = 20, + PERF_CP_PREEMPTION_REACTION_DELAY = 21, + PERF_CP_PREEMPTION_SWITCH_OUT_TIME = 22, + PERF_CP_PREEMPTION_SWITCH_IN_TIME = 23, + PERF_CP_DEAD_DRAWS_IN_BIN_RENDER = 24, + PERF_CP_PREDICATED_DRAWS_KILLED = 25, + PERF_CP_MODE_SWITCH = 26, + PERF_CP_ZPASS_DONE = 27, + PERF_CP_CONTEXT_DONE = 28, + PERF_CP_CACHE_FLUSH = 29, + PERF_CP_LONG_PREEMPTIONS = 30, +}; + +enum a5xx_rbbm_perfcounter_select { + PERF_RBBM_ALWAYS_COUNT = 0, + PERF_RBBM_ALWAYS_ON = 1, + PERF_RBBM_TSE_BUSY = 2, + PERF_RBBM_RAS_BUSY = 3, + PERF_RBBM_PC_DCALL_BUSY = 4, + PERF_RBBM_PC_VSD_BUSY = 5, + PERF_RBBM_STATUS_MASKED = 6, + PERF_RBBM_COM_BUSY = 7, + PERF_RBBM_DCOM_BUSY = 8, + PERF_RBBM_VBIF_BUSY = 9, + PERF_RBBM_VSC_BUSY = 10, + PERF_RBBM_TESS_BUSY = 11, + PERF_RBBM_UCHE_BUSY = 12, + PERF_RBBM_HLSQ_BUSY = 13, +}; + +enum a5xx_pc_perfcounter_select { + PERF_PC_BUSY_CYCLES = 0, + PERF_PC_WORKING_CYCLES = 1, + PERF_PC_STALL_CYCLES_VFD = 2, + PERF_PC_STALL_CYCLES_TSE = 3, + PERF_PC_STALL_CYCLES_VPC = 4, + PERF_PC_STALL_CYCLES_UCHE = 5, + PERF_PC_STALL_CYCLES_TESS = 6, + PERF_PC_STALL_CYCLES_TSE_ONLY = 7, + PERF_PC_STALL_CYCLES_VPC_ONLY = 8, + PERF_PC_PASS1_TF_STALL_CYCLES = 9, + PERF_PC_STARVE_CYCLES_FOR_INDEX = 10, + PERF_PC_STARVE_CYCLES_FOR_TESS_FACTOR = 11, + PERF_PC_STARVE_CYCLES_FOR_VIZ_STREAM = 12, + PERF_PC_STARVE_CYCLES_FOR_POSITION = 13, + PERF_PC_STARVE_CYCLES_DI = 14, + PERF_PC_VIS_STREAMS_LOADED = 15, + PERF_PC_INSTANCES = 16, + PERF_PC_VPC_PRIMITIVES = 17, + PERF_PC_DEAD_PRIM = 18, + PERF_PC_LIVE_PRIM = 19, + PERF_PC_VERTEX_HITS = 20, + PERF_PC_IA_VERTICES = 21, + PERF_PC_IA_PRIMITIVES = 22, + PERF_PC_GS_PRIMITIVES = 23, + PERF_PC_HS_INVOCATIONS = 24, + PERF_PC_DS_INVOCATIONS = 25, + PERF_PC_VS_INVOCATIONS = 26, + PERF_PC_GS_INVOCATIONS = 27, + PERF_PC_DS_PRIMITIVES = 28, + PERF_PC_VPC_POS_DATA_TRANSACTION = 29, + PERF_PC_3D_DRAWCALLS = 30, + PERF_PC_2D_DRAWCALLS = 31, + PERF_PC_NON_DRAWCALL_GLOBAL_EVENTS = 32, + PERF_TESS_BUSY_CYCLES = 33, + PERF_TESS_WORKING_CYCLES = 34, + PERF_TESS_STALL_CYCLES_PC = 35, + PERF_TESS_STARVE_CYCLES_PC = 36, +}; + +enum a5xx_vfd_perfcounter_select { + PERF_VFD_BUSY_CYCLES = 0, + PERF_VFD_STALL_CYCLES_UCHE = 1, + PERF_VFD_STALL_CYCLES_VPC_ALLOC = 2, + PERF_VFD_STALL_CYCLES_MISS_VB = 3, + PERF_VFD_STALL_CYCLES_MISS_Q = 4, + PERF_VFD_STALL_CYCLES_SP_INFO = 5, + PERF_VFD_STALL_CYCLES_SP_ATTR = 6, + PERF_VFD_STALL_CYCLES_VFDP_VB = 7, + PERF_VFD_STALL_CYCLES_VFDP_Q = 8, + PERF_VFD_DECODER_PACKER_STALL = 9, + PERF_VFD_STARVE_CYCLES_UCHE = 10, + PERF_VFD_RBUFFER_FULL = 11, + PERF_VFD_ATTR_INFO_FIFO_FULL = 12, + PERF_VFD_DECODED_ATTRIBUTE_BYTES = 13, + PERF_VFD_NUM_ATTRIBUTES = 14, + PERF_VFD_INSTRUCTIONS = 15, + PERF_VFD_UPPER_SHADER_FIBERS = 16, + PERF_VFD_LOWER_SHADER_FIBERS = 17, + PERF_VFD_MODE_0_FIBERS = 18, + PERF_VFD_MODE_1_FIBERS = 19, + PERF_VFD_MODE_2_FIBERS = 20, + PERF_VFD_MODE_3_FIBERS = 21, + PERF_VFD_MODE_4_FIBERS = 22, + PERF_VFD_TOTAL_VERTICES = 23, + PERF_VFD_NUM_ATTR_MISS = 24, + PERF_VFD_1_BURST_REQ = 25, + PERF_VFDP_STALL_CYCLES_VFD = 26, + PERF_VFDP_STALL_CYCLES_VFD_INDEX = 27, + PERF_VFDP_STALL_CYCLES_VFD_PROG = 28, + PERF_VFDP_STARVE_CYCLES_PC = 29, + PERF_VFDP_VS_STAGE_32_WAVES = 30, +}; + +enum a5xx_hlsq_perfcounter_select { + PERF_HLSQ_BUSY_CYCLES = 0, + PERF_HLSQ_STALL_CYCLES_UCHE = 1, + PERF_HLSQ_STALL_CYCLES_SP_STATE = 2, + PERF_HLSQ_STALL_CYCLES_SP_FS_STAGE = 3, + PERF_HLSQ_UCHE_LATENCY_CYCLES = 4, + PERF_HLSQ_UCHE_LATENCY_COUNT = 5, + PERF_HLSQ_FS_STAGE_32_WAVES = 6, + PERF_HLSQ_FS_STAGE_64_WAVES = 7, + PERF_HLSQ_QUADS = 8, + PERF_HLSQ_SP_STATE_COPY_TRANS_FS_STAGE = 9, + PERF_HLSQ_SP_STATE_COPY_TRANS_VS_STAGE = 10, + PERF_HLSQ_TP_STATE_COPY_TRANS_FS_STAGE = 11, + PERF_HLSQ_TP_STATE_COPY_TRANS_VS_STAGE = 12, + PERF_HLSQ_CS_INVOCATIONS = 13, + PERF_HLSQ_COMPUTE_DRAWCALLS = 14, +}; + +enum a5xx_vpc_perfcounter_select { + PERF_VPC_BUSY_CYCLES = 0, + PERF_VPC_WORKING_CYCLES = 1, + PERF_VPC_STALL_CYCLES_UCHE = 2, + PERF_VPC_STALL_CYCLES_VFD_WACK = 3, + PERF_VPC_STALL_CYCLES_HLSQ_PRIM_ALLOC = 4, + PERF_VPC_STALL_CYCLES_PC = 5, + PERF_VPC_STALL_CYCLES_SP_LM = 6, + PERF_VPC_POS_EXPORT_STALL_CYCLES = 7, + PERF_VPC_STARVE_CYCLES_SP = 8, + PERF_VPC_STARVE_CYCLES_LRZ = 9, + PERF_VPC_PC_PRIMITIVES = 10, + PERF_VPC_SP_COMPONENTS = 11, + PERF_VPC_SP_LM_PRIMITIVES = 12, + PERF_VPC_SP_LM_COMPONENTS = 13, + PERF_VPC_SP_LM_DWORDS = 14, + PERF_VPC_STREAMOUT_COMPONENTS = 15, + PERF_VPC_GRANT_PHASES = 16, +}; + +enum a5xx_tse_perfcounter_select { + PERF_TSE_BUSY_CYCLES = 0, + PERF_TSE_CLIPPING_CYCLES = 1, + PERF_TSE_STALL_CYCLES_RAS = 2, + PERF_TSE_STALL_CYCLES_LRZ_BARYPLANE = 3, + PERF_TSE_STALL_CYCLES_LRZ_ZPLANE = 4, + PERF_TSE_STARVE_CYCLES_PC = 5, + PERF_TSE_INPUT_PRIM = 6, + PERF_TSE_INPUT_NULL_PRIM = 7, + PERF_TSE_TRIVAL_REJ_PRIM = 8, + PERF_TSE_CLIPPED_PRIM = 9, + PERF_TSE_ZERO_AREA_PRIM = 10, + PERF_TSE_FACENESS_CULLED_PRIM = 11, + PERF_TSE_ZERO_PIXEL_PRIM = 12, + PERF_TSE_OUTPUT_NULL_PRIM = 13, + PERF_TSE_OUTPUT_VISIBLE_PRIM = 14, + PERF_TSE_CINVOCATION = 15, + PERF_TSE_CPRIMITIVES = 16, + PERF_TSE_2D_INPUT_PRIM = 17, + PERF_TSE_2D_ALIVE_CLCLES = 18, +}; + +enum a5xx_ras_perfcounter_select { + PERF_RAS_BUSY_CYCLES = 0, + PERF_RAS_SUPERTILE_ACTIVE_CYCLES = 1, + PERF_RAS_STALL_CYCLES_LRZ = 2, + PERF_RAS_STARVE_CYCLES_TSE = 3, + PERF_RAS_SUPER_TILES = 4, + PERF_RAS_8X4_TILES = 5, + PERF_RAS_MASKGEN_ACTIVE = 6, + PERF_RAS_FULLY_COVERED_SUPER_TILES = 7, + PERF_RAS_FULLY_COVERED_8X4_TILES = 8, + PERF_RAS_PRIM_KILLED_INVISILBE = 9, +}; + +enum a5xx_lrz_perfcounter_select { + PERF_LRZ_BUSY_CYCLES = 0, + PERF_LRZ_STARVE_CYCLES_RAS = 1, + PERF_LRZ_STALL_CYCLES_RB = 2, + PERF_LRZ_STALL_CYCLES_VSC = 3, + PERF_LRZ_STALL_CYCLES_VPC = 4, + PERF_LRZ_STALL_CYCLES_FLAG_PREFETCH = 5, + PERF_LRZ_STALL_CYCLES_UCHE = 6, + PERF_LRZ_LRZ_READ = 7, + PERF_LRZ_LRZ_WRITE = 8, + PERF_LRZ_READ_LATENCY = 9, + PERF_LRZ_MERGE_CACHE_UPDATING = 10, + PERF_LRZ_PRIM_KILLED_BY_MASKGEN = 11, + PERF_LRZ_PRIM_KILLED_BY_LRZ = 12, + PERF_LRZ_VISIBLE_PRIM_AFTER_LRZ = 13, + PERF_LRZ_FULL_8X8_TILES = 14, + PERF_LRZ_PARTIAL_8X8_TILES = 15, + PERF_LRZ_TILE_KILLED = 16, + PERF_LRZ_TOTAL_PIXEL = 17, + PERF_LRZ_VISIBLE_PIXEL_AFTER_LRZ = 18, +}; + +enum a5xx_uche_perfcounter_select { + PERF_UCHE_BUSY_CYCLES = 0, + PERF_UCHE_STALL_CYCLES_VBIF = 1, + PERF_UCHE_VBIF_LATENCY_CYCLES = 2, + PERF_UCHE_VBIF_LATENCY_SAMPLES = 3, + PERF_UCHE_VBIF_READ_BEATS_TP = 4, + PERF_UCHE_VBIF_READ_BEATS_VFD = 5, + PERF_UCHE_VBIF_READ_BEATS_HLSQ = 6, + PERF_UCHE_VBIF_READ_BEATS_LRZ = 7, + PERF_UCHE_VBIF_READ_BEATS_SP = 8, + PERF_UCHE_READ_REQUESTS_TP = 9, + PERF_UCHE_READ_REQUESTS_VFD = 10, + PERF_UCHE_READ_REQUESTS_HLSQ = 11, + PERF_UCHE_READ_REQUESTS_LRZ = 12, + PERF_UCHE_READ_REQUESTS_SP = 13, + PERF_UCHE_WRITE_REQUESTS_LRZ = 14, + PERF_UCHE_WRITE_REQUESTS_SP = 15, + PERF_UCHE_WRITE_REQUESTS_VPC = 16, + PERF_UCHE_WRITE_REQUESTS_VSC = 17, + PERF_UCHE_EVICTS = 18, + PERF_UCHE_BANK_REQ0 = 19, + PERF_UCHE_BANK_REQ1 = 20, + PERF_UCHE_BANK_REQ2 = 21, + PERF_UCHE_BANK_REQ3 = 22, + PERF_UCHE_BANK_REQ4 = 23, + PERF_UCHE_BANK_REQ5 = 24, + PERF_UCHE_BANK_REQ6 = 25, + PERF_UCHE_BANK_REQ7 = 26, + PERF_UCHE_VBIF_READ_BEATS_CH0 = 27, + PERF_UCHE_VBIF_READ_BEATS_CH1 = 28, + PERF_UCHE_GMEM_READ_BEATS = 29, + PERF_UCHE_FLAG_COUNT = 30, +}; + +enum a5xx_tp_perfcounter_select { + PERF_TP_BUSY_CYCLES = 0, + PERF_TP_STALL_CYCLES_UCHE = 1, + PERF_TP_LATENCY_CYCLES = 2, + PERF_TP_LATENCY_TRANS = 3, + PERF_TP_FLAG_CACHE_REQUEST_SAMPLES = 4, + PERF_TP_FLAG_CACHE_REQUEST_LATENCY = 5, + PERF_TP_L1_CACHELINE_REQUESTS = 6, + PERF_TP_L1_CACHELINE_MISSES = 7, + PERF_TP_SP_TP_TRANS = 8, + PERF_TP_TP_SP_TRANS = 9, + PERF_TP_OUTPUT_PIXELS = 10, + PERF_TP_FILTER_WORKLOAD_16BIT = 11, + PERF_TP_FILTER_WORKLOAD_32BIT = 12, + PERF_TP_QUADS_RECEIVED = 13, + PERF_TP_QUADS_OFFSET = 14, + PERF_TP_QUADS_SHADOW = 15, + PERF_TP_QUADS_ARRAY = 16, + PERF_TP_QUADS_GRADIENT = 17, + PERF_TP_QUADS_1D = 18, + PERF_TP_QUADS_2D = 19, + PERF_TP_QUADS_BUFFER = 20, + PERF_TP_QUADS_3D = 21, + PERF_TP_QUADS_CUBE = 22, + PERF_TP_STATE_CACHE_REQUESTS = 23, + PERF_TP_STATE_CACHE_MISSES = 24, + PERF_TP_DIVERGENT_QUADS_RECEIVED = 25, + PERF_TP_BINDLESS_STATE_CACHE_REQUESTS = 26, + PERF_TP_BINDLESS_STATE_CACHE_MISSES = 27, + PERF_TP_PRT_NON_RESIDENT_EVENTS = 28, + PERF_TP_OUTPUT_PIXELS_POINT = 29, + PERF_TP_OUTPUT_PIXELS_BILINEAR = 30, + PERF_TP_OUTPUT_PIXELS_MIP = 31, + PERF_TP_OUTPUT_PIXELS_ANISO = 32, + PERF_TP_OUTPUT_PIXELS_ZERO_LOD = 33, + PERF_TP_FLAG_CACHE_REQUESTS = 34, + PERF_TP_FLAG_CACHE_MISSES = 35, + PERF_TP_L1_5_L2_REQUESTS = 36, + PERF_TP_2D_OUTPUT_PIXELS = 37, + PERF_TP_2D_OUTPUT_PIXELS_POINT = 38, + PERF_TP_2D_OUTPUT_PIXELS_BILINEAR = 39, + PERF_TP_2D_FILTER_WORKLOAD_16BIT = 40, + PERF_TP_2D_FILTER_WORKLOAD_32BIT = 41, +}; + +enum a5xx_sp_perfcounter_select { + PERF_SP_BUSY_CYCLES = 0, + PERF_SP_ALU_WORKING_CYCLES = 1, + PERF_SP_EFU_WORKING_CYCLES = 2, + PERF_SP_STALL_CYCLES_VPC = 3, + PERF_SP_STALL_CYCLES_TP = 4, + PERF_SP_STALL_CYCLES_UCHE = 5, + PERF_SP_STALL_CYCLES_RB = 6, + PERF_SP_SCHEDULER_NON_WORKING = 7, + PERF_SP_WAVE_CONTEXTS = 8, + PERF_SP_WAVE_CONTEXT_CYCLES = 9, + PERF_SP_FS_STAGE_WAVE_CYCLES = 10, + PERF_SP_FS_STAGE_WAVE_SAMPLES = 11, + PERF_SP_VS_STAGE_WAVE_CYCLES = 12, + PERF_SP_VS_STAGE_WAVE_SAMPLES = 13, + PERF_SP_FS_STAGE_DURATION_CYCLES = 14, + PERF_SP_VS_STAGE_DURATION_CYCLES = 15, + PERF_SP_WAVE_CTRL_CYCLES = 16, + PERF_SP_WAVE_LOAD_CYCLES = 17, + PERF_SP_WAVE_EMIT_CYCLES = 18, + PERF_SP_WAVE_NOP_CYCLES = 19, + PERF_SP_WAVE_WAIT_CYCLES = 20, + PERF_SP_WAVE_FETCH_CYCLES = 21, + PERF_SP_WAVE_IDLE_CYCLES = 22, + PERF_SP_WAVE_END_CYCLES = 23, + PERF_SP_WAVE_LONG_SYNC_CYCLES = 24, + PERF_SP_WAVE_SHORT_SYNC_CYCLES = 25, + PERF_SP_WAVE_JOIN_CYCLES = 26, + PERF_SP_LM_LOAD_INSTRUCTIONS = 27, + PERF_SP_LM_STORE_INSTRUCTIONS = 28, + PERF_SP_LM_ATOMICS = 29, + PERF_SP_GM_LOAD_INSTRUCTIONS = 30, + PERF_SP_GM_STORE_INSTRUCTIONS = 31, + PERF_SP_GM_ATOMICS = 32, + PERF_SP_VS_STAGE_TEX_INSTRUCTIONS = 33, + PERF_SP_VS_STAGE_CFLOW_INSTRUCTIONS = 34, + PERF_SP_VS_STAGE_EFU_INSTRUCTIONS = 35, + PERF_SP_VS_STAGE_FULL_ALU_INSTRUCTIONS = 36, + PERF_SP_VS_STAGE_HALF_ALU_INSTRUCTIONS = 37, + PERF_SP_FS_STAGE_TEX_INSTRUCTIONS = 38, + PERF_SP_FS_STAGE_CFLOW_INSTRUCTIONS = 39, + PERF_SP_FS_STAGE_EFU_INSTRUCTIONS = 40, + PERF_SP_FS_STAGE_FULL_ALU_INSTRUCTIONS = 41, + PERF_SP_FS_STAGE_HALF_ALU_INSTRUCTIONS = 42, + PERF_SP_FS_STAGE_BARY_INSTRUCTIONS = 43, + PERF_SP_VS_INSTRUCTIONS = 44, + PERF_SP_FS_INSTRUCTIONS = 45, + PERF_SP_ADDR_LOCK_COUNT = 46, + PERF_SP_UCHE_READ_TRANS = 47, + PERF_SP_UCHE_WRITE_TRANS = 48, + PERF_SP_EXPORT_VPC_TRANS = 49, + PERF_SP_EXPORT_RB_TRANS = 50, + PERF_SP_PIXELS_KILLED = 51, + PERF_SP_ICL1_REQUESTS = 52, + PERF_SP_ICL1_MISSES = 53, + PERF_SP_ICL0_REQUESTS = 54, + PERF_SP_ICL0_MISSES = 55, + PERF_SP_HS_INSTRUCTIONS = 56, + PERF_SP_DS_INSTRUCTIONS = 57, + PERF_SP_GS_INSTRUCTIONS = 58, + PERF_SP_CS_INSTRUCTIONS = 59, + PERF_SP_GPR_READ = 60, + PERF_SP_GPR_WRITE = 61, + PERF_SP_LM_CH0_REQUESTS = 62, + PERF_SP_LM_CH1_REQUESTS = 63, + PERF_SP_LM_BANK_CONFLICTS = 64, +}; + +enum a5xx_rb_perfcounter_select { + PERF_RB_BUSY_CYCLES = 0, + PERF_RB_STALL_CYCLES_CCU = 1, + PERF_RB_STALL_CYCLES_HLSQ = 2, + PERF_RB_STALL_CYCLES_FIFO0_FULL = 3, + PERF_RB_STALL_CYCLES_FIFO1_FULL = 4, + PERF_RB_STALL_CYCLES_FIFO2_FULL = 5, + PERF_RB_STARVE_CYCLES_SP = 6, + PERF_RB_STARVE_CYCLES_LRZ_TILE = 7, + PERF_RB_STARVE_CYCLES_CCU = 8, + PERF_RB_STARVE_CYCLES_Z_PLANE = 9, + PERF_RB_STARVE_CYCLES_BARY_PLANE = 10, + PERF_RB_Z_WORKLOAD = 11, + PERF_RB_HLSQ_ACTIVE = 12, + PERF_RB_Z_READ = 13, + PERF_RB_Z_WRITE = 14, + PERF_RB_C_READ = 15, + PERF_RB_C_WRITE = 16, + PERF_RB_TOTAL_PASS = 17, + PERF_RB_Z_PASS = 18, + PERF_RB_Z_FAIL = 19, + PERF_RB_S_FAIL = 20, + PERF_RB_BLENDED_FXP_COMPONENTS = 21, + PERF_RB_BLENDED_FP16_COMPONENTS = 22, + RB_RESERVED = 23, + PERF_RB_2D_ALIVE_CYCLES = 24, + PERF_RB_2D_STALL_CYCLES_A2D = 25, + PERF_RB_2D_STARVE_CYCLES_SRC = 26, + PERF_RB_2D_STARVE_CYCLES_SP = 27, + PERF_RB_2D_STARVE_CYCLES_DST = 28, + PERF_RB_2D_VALID_PIXELS = 29, +}; + +enum a5xx_rb_samples_perfcounter_select { + TOTAL_SAMPLES = 0, + ZPASS_SAMPLES = 1, + ZFAIL_SAMPLES = 2, + SFAIL_SAMPLES = 3, +}; + +enum a5xx_vsc_perfcounter_select { + PERF_VSC_BUSY_CYCLES = 0, + PERF_VSC_WORKING_CYCLES = 1, + PERF_VSC_STALL_CYCLES_UCHE = 2, + PERF_VSC_EOT_NUM = 3, +}; + +enum a5xx_ccu_perfcounter_select { + PERF_CCU_BUSY_CYCLES = 0, + PERF_CCU_STALL_CYCLES_RB_DEPTH_RETURN = 1, + PERF_CCU_STALL_CYCLES_RB_COLOR_RETURN = 2, + PERF_CCU_STARVE_CYCLES_FLAG_RETURN = 3, + PERF_CCU_DEPTH_BLOCKS = 4, + PERF_CCU_COLOR_BLOCKS = 5, + PERF_CCU_DEPTH_BLOCK_HIT = 6, + PERF_CCU_COLOR_BLOCK_HIT = 7, + PERF_CCU_PARTIAL_BLOCK_READ = 8, + PERF_CCU_GMEM_READ = 9, + PERF_CCU_GMEM_WRITE = 10, + PERF_CCU_DEPTH_READ_FLAG0_COUNT = 11, + PERF_CCU_DEPTH_READ_FLAG1_COUNT = 12, + PERF_CCU_DEPTH_READ_FLAG2_COUNT = 13, + PERF_CCU_DEPTH_READ_FLAG3_COUNT = 14, + PERF_CCU_DEPTH_READ_FLAG4_COUNT = 15, + PERF_CCU_COLOR_READ_FLAG0_COUNT = 16, + PERF_CCU_COLOR_READ_FLAG1_COUNT = 17, + PERF_CCU_COLOR_READ_FLAG2_COUNT = 18, + PERF_CCU_COLOR_READ_FLAG3_COUNT = 19, + PERF_CCU_COLOR_READ_FLAG4_COUNT = 20, + PERF_CCU_2D_BUSY_CYCLES = 21, + PERF_CCU_2D_RD_REQ = 22, + PERF_CCU_2D_WR_REQ = 23, + PERF_CCU_2D_REORDER_STARVE_CYCLES = 24, + PERF_CCU_2D_PIXELS = 25, +}; + +enum a5xx_cmp_perfcounter_select { + PERF_CMPDECMP_STALL_CYCLES_VBIF = 0, + PERF_CMPDECMP_VBIF_LATENCY_CYCLES = 1, + PERF_CMPDECMP_VBIF_LATENCY_SAMPLES = 2, + PERF_CMPDECMP_VBIF_READ_DATA_CCU = 3, + PERF_CMPDECMP_VBIF_WRITE_DATA_CCU = 4, + PERF_CMPDECMP_VBIF_READ_REQUEST = 5, + PERF_CMPDECMP_VBIF_WRITE_REQUEST = 6, + PERF_CMPDECMP_VBIF_READ_DATA = 7, + PERF_CMPDECMP_VBIF_WRITE_DATA = 8, + PERF_CMPDECMP_FLAG_FETCH_CYCLES = 9, + PERF_CMPDECMP_FLAG_FETCH_SAMPLES = 10, + PERF_CMPDECMP_DEPTH_WRITE_FLAG1_COUNT = 11, + PERF_CMPDECMP_DEPTH_WRITE_FLAG2_COUNT = 12, + PERF_CMPDECMP_DEPTH_WRITE_FLAG3_COUNT = 13, + PERF_CMPDECMP_DEPTH_WRITE_FLAG4_COUNT = 14, + PERF_CMPDECMP_COLOR_WRITE_FLAG1_COUNT = 15, + PERF_CMPDECMP_COLOR_WRITE_FLAG2_COUNT = 16, + PERF_CMPDECMP_COLOR_WRITE_FLAG3_COUNT = 17, + PERF_CMPDECMP_COLOR_WRITE_FLAG4_COUNT = 18, + PERF_CMPDECMP_2D_STALL_CYCLES_VBIF_REQ = 19, + PERF_CMPDECMP_2D_STALL_CYCLES_VBIF_WR = 20, + PERF_CMPDECMP_2D_STALL_CYCLES_VBIF_RETURN = 21, + PERF_CMPDECMP_2D_RD_DATA = 22, + PERF_CMPDECMP_2D_WR_DATA = 23, +}; + +enum a5xx_vbif_perfcounter_select { + AXI_READ_REQUESTS_ID_0 = 0, + AXI_READ_REQUESTS_ID_1 = 1, + AXI_READ_REQUESTS_ID_2 = 2, + AXI_READ_REQUESTS_ID_3 = 3, + AXI_READ_REQUESTS_ID_4 = 4, + AXI_READ_REQUESTS_ID_5 = 5, + AXI_READ_REQUESTS_ID_6 = 6, + AXI_READ_REQUESTS_ID_7 = 7, + AXI_READ_REQUESTS_ID_8 = 8, + AXI_READ_REQUESTS_ID_9 = 9, + AXI_READ_REQUESTS_ID_10 = 10, + AXI_READ_REQUESTS_ID_11 = 11, + AXI_READ_REQUESTS_ID_12 = 12, + AXI_READ_REQUESTS_ID_13 = 13, + AXI_READ_REQUESTS_ID_14 = 14, + AXI_READ_REQUESTS_ID_15 = 15, + AXI0_READ_REQUESTS_TOTAL = 16, + AXI1_READ_REQUESTS_TOTAL = 17, + AXI2_READ_REQUESTS_TOTAL = 18, + AXI3_READ_REQUESTS_TOTAL = 19, + AXI_READ_REQUESTS_TOTAL = 20, + AXI_WRITE_REQUESTS_ID_0 = 21, + AXI_WRITE_REQUESTS_ID_1 = 22, + AXI_WRITE_REQUESTS_ID_2 = 23, + AXI_WRITE_REQUESTS_ID_3 = 24, + AXI_WRITE_REQUESTS_ID_4 = 25, + AXI_WRITE_REQUESTS_ID_5 = 26, + AXI_WRITE_REQUESTS_ID_6 = 27, + AXI_WRITE_REQUESTS_ID_7 = 28, + AXI_WRITE_REQUESTS_ID_8 = 29, + AXI_WRITE_REQUESTS_ID_9 = 30, + AXI_WRITE_REQUESTS_ID_10 = 31, + AXI_WRITE_REQUESTS_ID_11 = 32, + AXI_WRITE_REQUESTS_ID_12 = 33, + AXI_WRITE_REQUESTS_ID_13 = 34, + AXI_WRITE_REQUESTS_ID_14 = 35, + AXI_WRITE_REQUESTS_ID_15 = 36, + AXI0_WRITE_REQUESTS_TOTAL = 37, + AXI1_WRITE_REQUESTS_TOTAL = 38, + AXI2_WRITE_REQUESTS_TOTAL = 39, + AXI3_WRITE_REQUESTS_TOTAL = 40, + AXI_WRITE_REQUESTS_TOTAL = 41, + AXI_TOTAL_REQUESTS = 42, + AXI_READ_DATA_BEATS_ID_0 = 43, + AXI_READ_DATA_BEATS_ID_1 = 44, + AXI_READ_DATA_BEATS_ID_2 = 45, + AXI_READ_DATA_BEATS_ID_3 = 46, + AXI_READ_DATA_BEATS_ID_4 = 47, + AXI_READ_DATA_BEATS_ID_5 = 48, + AXI_READ_DATA_BEATS_ID_6 = 49, + AXI_READ_DATA_BEATS_ID_7 = 50, + AXI_READ_DATA_BEATS_ID_8 = 51, + AXI_READ_DATA_BEATS_ID_9 = 52, + AXI_READ_DATA_BEATS_ID_10 = 53, + AXI_READ_DATA_BEATS_ID_11 = 54, + AXI_READ_DATA_BEATS_ID_12 = 55, + AXI_READ_DATA_BEATS_ID_13 = 56, + AXI_READ_DATA_BEATS_ID_14 = 57, + AXI_READ_DATA_BEATS_ID_15 = 58, + AXI0_READ_DATA_BEATS_TOTAL = 59, + AXI1_READ_DATA_BEATS_TOTAL = 60, + AXI2_READ_DATA_BEATS_TOTAL = 61, + AXI3_READ_DATA_BEATS_TOTAL = 62, + AXI_READ_DATA_BEATS_TOTAL = 63, + AXI_WRITE_DATA_BEATS_ID_0 = 64, + AXI_WRITE_DATA_BEATS_ID_1 = 65, + AXI_WRITE_DATA_BEATS_ID_2 = 66, + AXI_WRITE_DATA_BEATS_ID_3 = 67, + AXI_WRITE_DATA_BEATS_ID_4 = 68, + AXI_WRITE_DATA_BEATS_ID_5 = 69, + AXI_WRITE_DATA_BEATS_ID_6 = 70, + AXI_WRITE_DATA_BEATS_ID_7 = 71, + AXI_WRITE_DATA_BEATS_ID_8 = 72, + AXI_WRITE_DATA_BEATS_ID_9 = 73, + AXI_WRITE_DATA_BEATS_ID_10 = 74, + AXI_WRITE_DATA_BEATS_ID_11 = 75, + AXI_WRITE_DATA_BEATS_ID_12 = 76, + AXI_WRITE_DATA_BEATS_ID_13 = 77, + AXI_WRITE_DATA_BEATS_ID_14 = 78, + AXI_WRITE_DATA_BEATS_ID_15 = 79, + AXI0_WRITE_DATA_BEATS_TOTAL = 80, + AXI1_WRITE_DATA_BEATS_TOTAL = 81, + AXI2_WRITE_DATA_BEATS_TOTAL = 82, + AXI3_WRITE_DATA_BEATS_TOTAL = 83, + AXI_WRITE_DATA_BEATS_TOTAL = 84, + AXI_DATA_BEATS_TOTAL = 85, +}; + +enum a5xx_tex_filter { + A5XX_TEX_NEAREST = 0, + A5XX_TEX_LINEAR = 1, + A5XX_TEX_ANISO = 2, +}; + +enum a5xx_tex_clamp { + A5XX_TEX_REPEAT = 0, + A5XX_TEX_CLAMP_TO_EDGE = 1, + A5XX_TEX_MIRROR_REPEAT = 2, + A5XX_TEX_CLAMP_TO_BORDER = 3, + A5XX_TEX_MIRROR_CLAMP = 4, +}; + +enum a5xx_tex_aniso { + A5XX_TEX_ANISO_1 = 0, + A5XX_TEX_ANISO_2 = 1, + A5XX_TEX_ANISO_4 = 2, + A5XX_TEX_ANISO_8 = 3, + A5XX_TEX_ANISO_16 = 4, +}; + +enum a5xx_tex_swiz { + A5XX_TEX_X = 0, + A5XX_TEX_Y = 1, + A5XX_TEX_Z = 2, + A5XX_TEX_W = 3, + A5XX_TEX_ZERO = 4, + A5XX_TEX_ONE = 5, +}; + +enum a5xx_tex_type { + A5XX_TEX_1D = 0, + A5XX_TEX_2D = 1, + A5XX_TEX_CUBE = 2, + A5XX_TEX_3D = 3, +}; + +#define A5XX_INT0_RBBM_GPU_IDLE 0x00000001 +#define A5XX_INT0_RBBM_AHB_ERROR 0x00000002 +#define A5XX_INT0_RBBM_TRANSFER_TIMEOUT 0x00000004 +#define A5XX_INT0_RBBM_ME_MS_TIMEOUT 0x00000008 +#define A5XX_INT0_RBBM_PFP_MS_TIMEOUT 0x00000010 +#define A5XX_INT0_RBBM_ETS_MS_TIMEOUT 0x00000020 +#define A5XX_INT0_RBBM_ATB_ASYNC_OVERFLOW 0x00000040 +#define A5XX_INT0_RBBM_GPC_ERROR 0x00000080 +#define A5XX_INT0_CP_SW 0x00000100 +#define A5XX_INT0_CP_HW_ERROR 0x00000200 +#define A5XX_INT0_CP_CCU_FLUSH_DEPTH_TS 0x00000400 +#define A5XX_INT0_CP_CCU_FLUSH_COLOR_TS 0x00000800 +#define A5XX_INT0_CP_CCU_RESOLVE_TS 0x00001000 +#define A5XX_INT0_CP_IB2 0x00002000 +#define A5XX_INT0_CP_IB1 0x00004000 +#define A5XX_INT0_CP_RB 0x00008000 +#define A5XX_INT0_CP_UNUSED_1 0x00010000 +#define A5XX_INT0_CP_RB_DONE_TS 0x00020000 +#define A5XX_INT0_CP_WT_DONE_TS 0x00040000 +#define A5XX_INT0_UNKNOWN_1 0x00080000 +#define A5XX_INT0_CP_CACHE_FLUSH_TS 0x00100000 +#define A5XX_INT0_UNUSED_2 0x00200000 +#define A5XX_INT0_RBBM_ATB_BUS_OVERFLOW 0x00400000 +#define A5XX_INT0_MISC_HANG_DETECT 0x00800000 +#define A5XX_INT0_UCHE_OOB_ACCESS 0x01000000 +#define A5XX_INT0_UCHE_TRAP_INTR 0x02000000 +#define A5XX_INT0_DEBBUS_INTR_0 0x04000000 +#define A5XX_INT0_DEBBUS_INTR_1 0x08000000 +#define A5XX_INT0_GPMU_VOLTAGE_DROOP 0x10000000 +#define A5XX_INT0_GPMU_FIRMWARE 0x20000000 +#define A5XX_INT0_ISDB_CPU_IRQ 0x40000000 +#define A5XX_INT0_ISDB_UNDER_DEBUG 0x80000000 +#define A5XX_CP_INT_CP_OPCODE_ERROR 0x00000001 +#define A5XX_CP_INT_CP_RESERVED_BIT_ERROR 0x00000002 +#define A5XX_CP_INT_CP_HW_FAULT_ERROR 0x00000004 +#define A5XX_CP_INT_CP_DMA_ERROR 0x00000008 +#define A5XX_CP_INT_CP_REGISTER_PROTECTION_ERROR 0x00000010 +#define A5XX_CP_INT_CP_AHB_ERROR 0x00000020 +#define REG_A5XX_CP_RB_BASE 0x00000800 + +#define REG_A5XX_CP_RB_BASE_HI 0x00000801 + +#define REG_A5XX_CP_RB_CNTL 0x00000802 + +#define REG_A5XX_CP_RB_RPTR_ADDR 0x00000804 + +#define REG_A5XX_CP_RB_RPTR_ADDR_HI 0x00000805 + +#define REG_A5XX_CP_RB_RPTR 0x00000806 + +#define REG_A5XX_CP_RB_WPTR 0x00000807 + +#define REG_A5XX_CP_PFP_STAT_ADDR 0x00000808 + +#define REG_A5XX_CP_PFP_STAT_DATA 0x00000809 + +#define REG_A5XX_CP_DRAW_STATE_ADDR 0x0000080b + +#define REG_A5XX_CP_DRAW_STATE_DATA 0x0000080c + +#define REG_A5XX_CP_CRASH_SCRIPT_BASE_LO 0x00000817 + +#define REG_A5XX_CP_CRASH_SCRIPT_BASE_HI 0x00000818 + +#define REG_A5XX_CP_CRASH_DUMP_CNTL 0x00000819 + +#define REG_A5XX_CP_ME_STAT_ADDR 0x0000081a + +#define REG_A5XX_CP_ROQ_THRESHOLDS_1 0x0000081f + +#define REG_A5XX_CP_ROQ_THRESHOLDS_2 0x00000820 + +#define REG_A5XX_CP_ROQ_DBG_ADDR 0x00000821 + +#define REG_A5XX_CP_ROQ_DBG_DATA 0x00000822 + +#define REG_A5XX_CP_MEQ_DBG_ADDR 0x00000823 + +#define REG_A5XX_CP_MEQ_DBG_DATA 0x00000824 + +#define REG_A5XX_CP_MEQ_THRESHOLDS 0x00000825 + +#define REG_A5XX_CP_MERCIU_SIZE 0x00000826 + +#define REG_A5XX_CP_MERCIU_DBG_ADDR 0x00000827 + +#define REG_A5XX_CP_MERCIU_DBG_DATA_1 0x00000828 + +#define REG_A5XX_CP_MERCIU_DBG_DATA_2 0x00000829 + +#define REG_A5XX_CP_PFP_UCODE_DBG_ADDR 0x0000082a + +#define REG_A5XX_CP_PFP_UCODE_DBG_DATA 0x0000082b + +#define REG_A5XX_CP_ME_UCODE_DBG_ADDR 0x0000082f + +#define REG_A5XX_CP_ME_UCODE_DBG_DATA 0x00000830 + +#define REG_A5XX_CP_CNTL 0x00000831 + +#define REG_A5XX_CP_PFP_ME_CNTL 0x00000832 + +#define REG_A5XX_CP_CHICKEN_DBG 0x00000833 + +#define REG_A5XX_CP_PFP_INSTR_BASE_LO 0x00000835 + +#define REG_A5XX_CP_PFP_INSTR_BASE_HI 0x00000836 + +#define REG_A5XX_CP_ME_INSTR_BASE_LO 0x00000838 + +#define REG_A5XX_CP_ME_INSTR_BASE_HI 0x00000839 + +#define REG_A5XX_CP_CONTEXT_SWITCH_CNTL 0x0000083b + +#define REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_LO 0x0000083c + +#define REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_HI 0x0000083d + +#define REG_A5XX_CP_CONTEXT_SWITCH_SAVE_ADDR_LO 0x0000083e + +#define REG_A5XX_CP_CONTEXT_SWITCH_SAVE_ADDR_HI 0x0000083f + +#define REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_LO 0x00000840 + +#define REG_A5XX_CP_CONTEXT_SWITCH_SMMU_INFO_HI 0x00000841 + +#define REG_A5XX_CP_ADDR_MODE_CNTL 0x00000860 + +#define REG_A5XX_CP_ME_STAT_DATA 0x00000b14 + +#define REG_A5XX_CP_WFI_PEND_CTR 0x00000b15 + +#define REG_A5XX_CP_INTERRUPT_STATUS 0x00000b18 + +#define REG_A5XX_CP_HW_FAULT 0x00000b1a + +#define REG_A5XX_CP_PROTECT_STATUS 0x00000b1c + +#define REG_A5XX_CP_IB1_BASE 0x00000b1f + +#define REG_A5XX_CP_IB1_BASE_HI 0x00000b20 + +#define REG_A5XX_CP_IB1_BUFSZ 0x00000b21 + +#define REG_A5XX_CP_IB2_BASE 0x00000b22 + +#define REG_A5XX_CP_IB2_BASE_HI 0x00000b23 + +#define REG_A5XX_CP_IB2_BUFSZ 0x00000b24 + +static inline uint32_t REG_A5XX_CP_SCRATCH(uint32_t i0) { return 0x00000b78 + 0x1*i0; } + +static inline uint32_t REG_A5XX_CP_SCRATCH_REG(uint32_t i0) { return 0x00000b78 + 0x1*i0; } + +static inline uint32_t REG_A5XX_CP_PROTECT(uint32_t i0) { return 0x00000880 + 0x1*i0; } + +static inline uint32_t REG_A5XX_CP_PROTECT_REG(uint32_t i0) { return 0x00000880 + 0x1*i0; } +#define A5XX_CP_PROTECT_REG_BASE_ADDR__MASK 0x0001ffff +#define A5XX_CP_PROTECT_REG_BASE_ADDR__SHIFT 0 +static inline uint32_t A5XX_CP_PROTECT_REG_BASE_ADDR(uint32_t val) +{ + return ((val) << A5XX_CP_PROTECT_REG_BASE_ADDR__SHIFT) & A5XX_CP_PROTECT_REG_BASE_ADDR__MASK; +} +#define A5XX_CP_PROTECT_REG_MASK_LEN__MASK 0x1f000000 +#define A5XX_CP_PROTECT_REG_MASK_LEN__SHIFT 24 +static inline uint32_t A5XX_CP_PROTECT_REG_MASK_LEN(uint32_t val) +{ + return ((val) << A5XX_CP_PROTECT_REG_MASK_LEN__SHIFT) & A5XX_CP_PROTECT_REG_MASK_LEN__MASK; +} +#define A5XX_CP_PROTECT_REG_TRAP_WRITE 0x20000000 +#define A5XX_CP_PROTECT_REG_TRAP_READ 0x40000000 + +#define REG_A5XX_CP_PROTECT_CNTL 0x000008a0 + +#define REG_A5XX_CP_AHB_FAULT 0x00000b1b + +#define REG_A5XX_CP_PERFCTR_CP_SEL_0 0x00000bb0 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_1 0x00000bb1 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_2 0x00000bb2 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_3 0x00000bb3 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_4 0x00000bb4 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_5 0x00000bb5 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_6 0x00000bb6 + +#define REG_A5XX_CP_PERFCTR_CP_SEL_7 0x00000bb7 + +#define REG_A5XX_VSC_ADDR_MODE_CNTL 0x00000bc1 + +#define REG_A5XX_CP_POWERCTR_CP_SEL_0 0x00000bba + +#define REG_A5XX_CP_POWERCTR_CP_SEL_1 0x00000bbb + +#define REG_A5XX_CP_POWERCTR_CP_SEL_2 0x00000bbc + +#define REG_A5XX_CP_POWERCTR_CP_SEL_3 0x00000bbd + +#define REG_A5XX_RBBM_CFG_DBGBUS_SEL_A 0x00000004 + +#define REG_A5XX_RBBM_CFG_DBGBUS_SEL_B 0x00000005 + +#define REG_A5XX_RBBM_CFG_DBGBUS_SEL_C 0x00000006 + +#define REG_A5XX_RBBM_CFG_DBGBUS_SEL_D 0x00000007 + +#define REG_A5XX_RBBM_CFG_DBGBUS_CNTLT 0x00000008 + +#define REG_A5XX_RBBM_CFG_DBGBUS_CNTLM 0x00000009 + +#define REG_A5XX_RBBM_CFG_DEBBUS_CTLTM_ENABLE_SHIFT 0x00000018 + +#define REG_A5XX_RBBM_CFG_DBGBUS_OPL 0x0000000a + +#define REG_A5XX_RBBM_CFG_DBGBUS_OPE 0x0000000b + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTL_0 0x0000000c + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTL_1 0x0000000d + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTL_2 0x0000000e + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTL_3 0x0000000f + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKL_0 0x00000010 + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKL_1 0x00000011 + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKL_2 0x00000012 + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKL_3 0x00000013 + +#define REG_A5XX_RBBM_CFG_DBGBUS_BYTEL_0 0x00000014 + +#define REG_A5XX_RBBM_CFG_DBGBUS_BYTEL_1 0x00000015 + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTE_0 0x00000016 + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTE_1 0x00000017 + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTE_2 0x00000018 + +#define REG_A5XX_RBBM_CFG_DBGBUS_IVTE_3 0x00000019 + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKE_0 0x0000001a + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKE_1 0x0000001b + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKE_2 0x0000001c + +#define REG_A5XX_RBBM_CFG_DBGBUS_MASKE_3 0x0000001d + +#define REG_A5XX_RBBM_CFG_DBGBUS_NIBBLEE 0x0000001e + +#define REG_A5XX_RBBM_CFG_DBGBUS_PTRC0 0x0000001f + +#define REG_A5XX_RBBM_CFG_DBGBUS_PTRC1 0x00000020 + +#define REG_A5XX_RBBM_CFG_DBGBUS_LOADREG 0x00000021 + +#define REG_A5XX_RBBM_CFG_DBGBUS_IDX 0x00000022 + +#define REG_A5XX_RBBM_CFG_DBGBUS_CLRC 0x00000023 + +#define REG_A5XX_RBBM_CFG_DBGBUS_LOADIVT 0x00000024 + +#define REG_A5XX_RBBM_INTERFACE_HANG_INT_CNTL 0x0000002f + +#define REG_A5XX_RBBM_INT_CLEAR_CMD 0x00000037 + +#define REG_A5XX_RBBM_INT_0_MASK 0x00000038 +#define A5XX_RBBM_INT_0_MASK_RBBM_GPU_IDLE 0x00000001 +#define A5XX_RBBM_INT_0_MASK_RBBM_AHB_ERROR 0x00000002 +#define A5XX_RBBM_INT_0_MASK_RBBM_TRANSFER_TIMEOUT 0x00000004 +#define A5XX_RBBM_INT_0_MASK_RBBM_ME_MS_TIMEOUT 0x00000008 +#define A5XX_RBBM_INT_0_MASK_RBBM_PFP_MS_TIMEOUT 0x00000010 +#define A5XX_RBBM_INT_0_MASK_RBBM_ETS_MS_TIMEOUT 0x00000020 +#define A5XX_RBBM_INT_0_MASK_RBBM_ATB_ASYNC_OVERFLOW 0x00000040 +#define A5XX_RBBM_INT_0_MASK_RBBM_GPC_ERROR 0x00000080 +#define A5XX_RBBM_INT_0_MASK_CP_SW 0x00000100 +#define A5XX_RBBM_INT_0_MASK_CP_HW_ERROR 0x00000200 +#define A5XX_RBBM_INT_0_MASK_CP_CCU_FLUSH_DEPTH_TS 0x00000400 +#define A5XX_RBBM_INT_0_MASK_CP_CCU_FLUSH_COLOR_TS 0x00000800 +#define A5XX_RBBM_INT_0_MASK_CP_CCU_RESOLVE_TS 0x00001000 +#define A5XX_RBBM_INT_0_MASK_CP_IB2 0x00002000 +#define A5XX_RBBM_INT_0_MASK_CP_IB1 0x00004000 +#define A5XX_RBBM_INT_0_MASK_CP_RB 0x00008000 +#define A5XX_RBBM_INT_0_MASK_CP_RB_DONE_TS 0x00020000 +#define A5XX_RBBM_INT_0_MASK_CP_WT_DONE_TS 0x00040000 +#define A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS 0x00100000 +#define A5XX_RBBM_INT_0_MASK_RBBM_ATB_BUS_OVERFLOW 0x00400000 +#define A5XX_RBBM_INT_0_MASK_MISC_HANG_DETECT 0x00800000 +#define A5XX_RBBM_INT_0_MASK_UCHE_OOB_ACCESS 0x01000000 +#define A5XX_RBBM_INT_0_MASK_UCHE_TRAP_INTR 0x02000000 +#define A5XX_RBBM_INT_0_MASK_DEBBUS_INTR_0 0x04000000 +#define A5XX_RBBM_INT_0_MASK_DEBBUS_INTR_1 0x08000000 +#define A5XX_RBBM_INT_0_MASK_GPMU_VOLTAGE_DROOP 0x10000000 +#define A5XX_RBBM_INT_0_MASK_GPMU_FIRMWARE 0x20000000 +#define A5XX_RBBM_INT_0_MASK_ISDB_CPU_IRQ 0x40000000 +#define A5XX_RBBM_INT_0_MASK_ISDB_UNDER_DEBUG 0x80000000 + +#define REG_A5XX_RBBM_AHB_DBG_CNTL 0x0000003f + +#define REG_A5XX_RBBM_EXT_VBIF_DBG_CNTL 0x00000041 + +#define REG_A5XX_RBBM_SW_RESET_CMD 0x00000043 + +#define REG_A5XX_RBBM_BLOCK_SW_RESET_CMD 0x00000045 + +#define REG_A5XX_RBBM_BLOCK_SW_RESET_CMD2 0x00000046 + +#define REG_A5XX_RBBM_DBG_LO_HI_GPIO 0x00000048 + +#define REG_A5XX_RBBM_EXT_TRACE_BUS_CNTL 0x00000049 + +#define REG_A5XX_RBBM_CLOCK_CNTL_TP0 0x0000004a + +#define REG_A5XX_RBBM_CLOCK_CNTL_TP1 0x0000004b + +#define REG_A5XX_RBBM_CLOCK_CNTL_TP2 0x0000004c + +#define REG_A5XX_RBBM_CLOCK_CNTL_TP3 0x0000004d + +#define REG_A5XX_RBBM_CLOCK_CNTL2_TP0 0x0000004e + +#define REG_A5XX_RBBM_CLOCK_CNTL2_TP1 0x0000004f + +#define REG_A5XX_RBBM_CLOCK_CNTL2_TP2 0x00000050 + +#define REG_A5XX_RBBM_CLOCK_CNTL2_TP3 0x00000051 + +#define REG_A5XX_RBBM_CLOCK_CNTL3_TP0 0x00000052 + +#define REG_A5XX_RBBM_CLOCK_CNTL3_TP1 0x00000053 + +#define REG_A5XX_RBBM_CLOCK_CNTL3_TP2 0x00000054 + +#define REG_A5XX_RBBM_CLOCK_CNTL3_TP3 0x00000055 + +#define REG_A5XX_RBBM_READ_AHB_THROUGH_DBG 0x00000059 + +#define REG_A5XX_RBBM_CLOCK_CNTL_UCHE 0x0000005a + +#define REG_A5XX_RBBM_CLOCK_CNTL2_UCHE 0x0000005b + +#define REG_A5XX_RBBM_CLOCK_CNTL3_UCHE 0x0000005c + +#define REG_A5XX_RBBM_CLOCK_CNTL4_UCHE 0x0000005d + +#define REG_A5XX_RBBM_CLOCK_HYST_UCHE 0x0000005e + +#define REG_A5XX_RBBM_CLOCK_DELAY_UCHE 0x0000005f + +#define REG_A5XX_RBBM_CLOCK_MODE_GPC 0x00000060 + +#define REG_A5XX_RBBM_CLOCK_DELAY_GPC 0x00000061 + +#define REG_A5XX_RBBM_CLOCK_HYST_GPC 0x00000062 + +#define REG_A5XX_RBBM_CLOCK_CNTL_TSE_RAS_RBBM 0x00000063 + +#define REG_A5XX_RBBM_CLOCK_HYST_TSE_RAS_RBBM 0x00000064 + +#define REG_A5XX_RBBM_CLOCK_DELAY_TSE_RAS_RBBM 0x00000065 + +#define REG_A5XX_RBBM_CLOCK_DELAY_HLSQ 0x00000066 + +#define REG_A5XX_RBBM_CLOCK_CNTL 0x00000067 + +#define REG_A5XX_RBBM_CLOCK_CNTL_SP0 0x00000068 + +#define REG_A5XX_RBBM_CLOCK_CNTL_SP1 0x00000069 + +#define REG_A5XX_RBBM_CLOCK_CNTL_SP2 0x0000006a + +#define REG_A5XX_RBBM_CLOCK_CNTL_SP3 0x0000006b + +#define REG_A5XX_RBBM_CLOCK_CNTL2_SP0 0x0000006c + +#define REG_A5XX_RBBM_CLOCK_CNTL2_SP1 0x0000006d + +#define REG_A5XX_RBBM_CLOCK_CNTL2_SP2 0x0000006e + +#define REG_A5XX_RBBM_CLOCK_CNTL2_SP3 0x0000006f + +#define REG_A5XX_RBBM_CLOCK_HYST_SP0 0x00000070 + +#define REG_A5XX_RBBM_CLOCK_HYST_SP1 0x00000071 + +#define REG_A5XX_RBBM_CLOCK_HYST_SP2 0x00000072 + +#define REG_A5XX_RBBM_CLOCK_HYST_SP3 0x00000073 + +#define REG_A5XX_RBBM_CLOCK_DELAY_SP0 0x00000074 + +#define REG_A5XX_RBBM_CLOCK_DELAY_SP1 0x00000075 + +#define REG_A5XX_RBBM_CLOCK_DELAY_SP2 0x00000076 + +#define REG_A5XX_RBBM_CLOCK_DELAY_SP3 0x00000077 + +#define REG_A5XX_RBBM_CLOCK_CNTL_RB0 0x00000078 + +#define REG_A5XX_RBBM_CLOCK_CNTL_RB1 0x00000079 + +#define REG_A5XX_RBBM_CLOCK_CNTL_RB2 0x0000007a + +#define REG_A5XX_RBBM_CLOCK_CNTL_RB3 0x0000007b + +#define REG_A5XX_RBBM_CLOCK_CNTL2_RB0 0x0000007c + +#define REG_A5XX_RBBM_CLOCK_CNTL2_RB1 0x0000007d + +#define REG_A5XX_RBBM_CLOCK_CNTL2_RB2 0x0000007e + +#define REG_A5XX_RBBM_CLOCK_CNTL2_RB3 0x0000007f + +#define REG_A5XX_RBBM_CLOCK_HYST_RAC 0x00000080 + +#define REG_A5XX_RBBM_CLOCK_DELAY_RAC 0x00000081 + +#define REG_A5XX_RBBM_CLOCK_CNTL_CCU0 0x00000082 + +#define REG_A5XX_RBBM_CLOCK_CNTL_CCU1 0x00000083 + +#define REG_A5XX_RBBM_CLOCK_CNTL_CCU2 0x00000084 + +#define REG_A5XX_RBBM_CLOCK_CNTL_CCU3 0x00000085 + +#define REG_A5XX_RBBM_CLOCK_HYST_RB_CCU0 0x00000086 + +#define REG_A5XX_RBBM_CLOCK_HYST_RB_CCU1 0x00000087 + +#define REG_A5XX_RBBM_CLOCK_HYST_RB_CCU2 0x00000088 + +#define REG_A5XX_RBBM_CLOCK_HYST_RB_CCU3 0x00000089 + +#define REG_A5XX_RBBM_CLOCK_CNTL_RAC 0x0000008a + +#define REG_A5XX_RBBM_CLOCK_CNTL2_RAC 0x0000008b + +#define REG_A5XX_RBBM_CLOCK_DELAY_RB_CCU_L1_0 0x0000008c + +#define REG_A5XX_RBBM_CLOCK_DELAY_RB_CCU_L1_1 0x0000008d + +#define REG_A5XX_RBBM_CLOCK_DELAY_RB_CCU_L1_2 0x0000008e + +#define REG_A5XX_RBBM_CLOCK_DELAY_RB_CCU_L1_3 0x0000008f + +#define REG_A5XX_RBBM_CLOCK_HYST_VFD 0x00000090 + +#define REG_A5XX_RBBM_CLOCK_MODE_VFD 0x00000091 + +#define REG_A5XX_RBBM_CLOCK_DELAY_VFD 0x00000092 + +#define REG_A5XX_RBBM_AHB_CNTL0 0x00000093 + +#define REG_A5XX_RBBM_AHB_CNTL1 0x00000094 + +#define REG_A5XX_RBBM_AHB_CNTL2 0x00000095 + +#define REG_A5XX_RBBM_AHB_CMD 0x00000096 + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL11 0x0000009c + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL12 0x0000009d + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL13 0x0000009e + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL14 0x0000009f + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL15 0x000000a0 + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL16 0x000000a1 + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL17 0x000000a2 + +#define REG_A5XX_RBBM_INTERFACE_HANG_MASK_CNTL18 0x000000a3 + +#define REG_A5XX_RBBM_CLOCK_DELAY_TP0 0x000000a4 + +#define REG_A5XX_RBBM_CLOCK_DELAY_TP1 0x000000a5 + +#define REG_A5XX_RBBM_CLOCK_DELAY_TP2 0x000000a6 + +#define REG_A5XX_RBBM_CLOCK_DELAY_TP3 0x000000a7 + +#define REG_A5XX_RBBM_CLOCK_DELAY2_TP0 0x000000a8 + +#define REG_A5XX_RBBM_CLOCK_DELAY2_TP1 0x000000a9 + +#define REG_A5XX_RBBM_CLOCK_DELAY2_TP2 0x000000aa + +#define REG_A5XX_RBBM_CLOCK_DELAY2_TP3 0x000000ab + +#define REG_A5XX_RBBM_CLOCK_DELAY3_TP0 0x000000ac + +#define REG_A5XX_RBBM_CLOCK_DELAY3_TP1 0x000000ad + +#define REG_A5XX_RBBM_CLOCK_DELAY3_TP2 0x000000ae + +#define REG_A5XX_RBBM_CLOCK_DELAY3_TP3 0x000000af + +#define REG_A5XX_RBBM_CLOCK_HYST_TP0 0x000000b0 + +#define REG_A5XX_RBBM_CLOCK_HYST_TP1 0x000000b1 + +#define REG_A5XX_RBBM_CLOCK_HYST_TP2 0x000000b2 + +#define REG_A5XX_RBBM_CLOCK_HYST_TP3 0x000000b3 + +#define REG_A5XX_RBBM_CLOCK_HYST2_TP0 0x000000b4 + +#define REG_A5XX_RBBM_CLOCK_HYST2_TP1 0x000000b5 + +#define REG_A5XX_RBBM_CLOCK_HYST2_TP2 0x000000b6 + +#define REG_A5XX_RBBM_CLOCK_HYST2_TP3 0x000000b7 + +#define REG_A5XX_RBBM_CLOCK_HYST3_TP0 0x000000b8 + +#define REG_A5XX_RBBM_CLOCK_HYST3_TP1 0x000000b9 + +#define REG_A5XX_RBBM_CLOCK_HYST3_TP2 0x000000ba + +#define REG_A5XX_RBBM_CLOCK_HYST3_TP3 0x000000bb + +#define REG_A5XX_RBBM_CLOCK_CNTL_GPMU 0x000000c8 + +#define REG_A5XX_RBBM_CLOCK_DELAY_GPMU 0x000000c9 + +#define REG_A5XX_RBBM_CLOCK_HYST_GPMU 0x000000ca + +#define REG_A5XX_RBBM_PERFCTR_CP_0_LO 0x000003a0 + +#define REG_A5XX_RBBM_PERFCTR_CP_0_HI 0x000003a1 + +#define REG_A5XX_RBBM_PERFCTR_CP_1_LO 0x000003a2 + +#define REG_A5XX_RBBM_PERFCTR_CP_1_HI 0x000003a3 + +#define REG_A5XX_RBBM_PERFCTR_CP_2_LO 0x000003a4 + +#define REG_A5XX_RBBM_PERFCTR_CP_2_HI 0x000003a5 + +#define REG_A5XX_RBBM_PERFCTR_CP_3_LO 0x000003a6 + +#define REG_A5XX_RBBM_PERFCTR_CP_3_HI 0x000003a7 + +#define REG_A5XX_RBBM_PERFCTR_CP_4_LO 0x000003a8 + +#define REG_A5XX_RBBM_PERFCTR_CP_4_HI 0x000003a9 + +#define REG_A5XX_RBBM_PERFCTR_CP_5_LO 0x000003aa + +#define REG_A5XX_RBBM_PERFCTR_CP_5_HI 0x000003ab + +#define REG_A5XX_RBBM_PERFCTR_CP_6_LO 0x000003ac + +#define REG_A5XX_RBBM_PERFCTR_CP_6_HI 0x000003ad + +#define REG_A5XX_RBBM_PERFCTR_CP_7_LO 0x000003ae + +#define REG_A5XX_RBBM_PERFCTR_CP_7_HI 0x000003af + +#define REG_A5XX_RBBM_PERFCTR_RBBM_0_LO 0x000003b0 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_0_HI 0x000003b1 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_1_LO 0x000003b2 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_1_HI 0x000003b3 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_2_LO 0x000003b4 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_2_HI 0x000003b5 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_3_LO 0x000003b6 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_3_HI 0x000003b7 + +#define REG_A5XX_RBBM_PERFCTR_PC_0_LO 0x000003b8 + +#define REG_A5XX_RBBM_PERFCTR_PC_0_HI 0x000003b9 + +#define REG_A5XX_RBBM_PERFCTR_PC_1_LO 0x000003ba + +#define REG_A5XX_RBBM_PERFCTR_PC_1_HI 0x000003bb + +#define REG_A5XX_RBBM_PERFCTR_PC_2_LO 0x000003bc + +#define REG_A5XX_RBBM_PERFCTR_PC_2_HI 0x000003bd + +#define REG_A5XX_RBBM_PERFCTR_PC_3_LO 0x000003be + +#define REG_A5XX_RBBM_PERFCTR_PC_3_HI 0x000003bf + +#define REG_A5XX_RBBM_PERFCTR_PC_4_LO 0x000003c0 + +#define REG_A5XX_RBBM_PERFCTR_PC_4_HI 0x000003c1 + +#define REG_A5XX_RBBM_PERFCTR_PC_5_LO 0x000003c2 + +#define REG_A5XX_RBBM_PERFCTR_PC_5_HI 0x000003c3 + +#define REG_A5XX_RBBM_PERFCTR_PC_6_LO 0x000003c4 + +#define REG_A5XX_RBBM_PERFCTR_PC_6_HI 0x000003c5 + +#define REG_A5XX_RBBM_PERFCTR_PC_7_LO 0x000003c6 + +#define REG_A5XX_RBBM_PERFCTR_PC_7_HI 0x000003c7 + +#define REG_A5XX_RBBM_PERFCTR_VFD_0_LO 0x000003c8 + +#define REG_A5XX_RBBM_PERFCTR_VFD_0_HI 0x000003c9 + +#define REG_A5XX_RBBM_PERFCTR_VFD_1_LO 0x000003ca + +#define REG_A5XX_RBBM_PERFCTR_VFD_1_HI 0x000003cb + +#define REG_A5XX_RBBM_PERFCTR_VFD_2_LO 0x000003cc + +#define REG_A5XX_RBBM_PERFCTR_VFD_2_HI 0x000003cd + +#define REG_A5XX_RBBM_PERFCTR_VFD_3_LO 0x000003ce + +#define REG_A5XX_RBBM_PERFCTR_VFD_3_HI 0x000003cf + +#define REG_A5XX_RBBM_PERFCTR_VFD_4_LO 0x000003d0 + +#define REG_A5XX_RBBM_PERFCTR_VFD_4_HI 0x000003d1 + +#define REG_A5XX_RBBM_PERFCTR_VFD_5_LO 0x000003d2 + +#define REG_A5XX_RBBM_PERFCTR_VFD_5_HI 0x000003d3 + +#define REG_A5XX_RBBM_PERFCTR_VFD_6_LO 0x000003d4 + +#define REG_A5XX_RBBM_PERFCTR_VFD_6_HI 0x000003d5 + +#define REG_A5XX_RBBM_PERFCTR_VFD_7_LO 0x000003d6 + +#define REG_A5XX_RBBM_PERFCTR_VFD_7_HI 0x000003d7 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_0_LO 0x000003d8 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_0_HI 0x000003d9 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_1_LO 0x000003da + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_1_HI 0x000003db + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_2_LO 0x000003dc + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_2_HI 0x000003dd + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_3_LO 0x000003de + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_3_HI 0x000003df + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_4_LO 0x000003e0 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_4_HI 0x000003e1 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_5_LO 0x000003e2 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_5_HI 0x000003e3 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_6_LO 0x000003e4 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_6_HI 0x000003e5 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_7_LO 0x000003e6 + +#define REG_A5XX_RBBM_PERFCTR_HLSQ_7_HI 0x000003e7 + +#define REG_A5XX_RBBM_PERFCTR_VPC_0_LO 0x000003e8 + +#define REG_A5XX_RBBM_PERFCTR_VPC_0_HI 0x000003e9 + +#define REG_A5XX_RBBM_PERFCTR_VPC_1_LO 0x000003ea + +#define REG_A5XX_RBBM_PERFCTR_VPC_1_HI 0x000003eb + +#define REG_A5XX_RBBM_PERFCTR_VPC_2_LO 0x000003ec + +#define REG_A5XX_RBBM_PERFCTR_VPC_2_HI 0x000003ed + +#define REG_A5XX_RBBM_PERFCTR_VPC_3_LO 0x000003ee + +#define REG_A5XX_RBBM_PERFCTR_VPC_3_HI 0x000003ef + +#define REG_A5XX_RBBM_PERFCTR_CCU_0_LO 0x000003f0 + +#define REG_A5XX_RBBM_PERFCTR_CCU_0_HI 0x000003f1 + +#define REG_A5XX_RBBM_PERFCTR_CCU_1_LO 0x000003f2 + +#define REG_A5XX_RBBM_PERFCTR_CCU_1_HI 0x000003f3 + +#define REG_A5XX_RBBM_PERFCTR_CCU_2_LO 0x000003f4 + +#define REG_A5XX_RBBM_PERFCTR_CCU_2_HI 0x000003f5 + +#define REG_A5XX_RBBM_PERFCTR_CCU_3_LO 0x000003f6 + +#define REG_A5XX_RBBM_PERFCTR_CCU_3_HI 0x000003f7 + +#define REG_A5XX_RBBM_PERFCTR_TSE_0_LO 0x000003f8 + +#define REG_A5XX_RBBM_PERFCTR_TSE_0_HI 0x000003f9 + +#define REG_A5XX_RBBM_PERFCTR_TSE_1_LO 0x000003fa + +#define REG_A5XX_RBBM_PERFCTR_TSE_1_HI 0x000003fb + +#define REG_A5XX_RBBM_PERFCTR_TSE_2_LO 0x000003fc + +#define REG_A5XX_RBBM_PERFCTR_TSE_2_HI 0x000003fd + +#define REG_A5XX_RBBM_PERFCTR_TSE_3_LO 0x000003fe + +#define REG_A5XX_RBBM_PERFCTR_TSE_3_HI 0x000003ff + +#define REG_A5XX_RBBM_PERFCTR_RAS_0_LO 0x00000400 + +#define REG_A5XX_RBBM_PERFCTR_RAS_0_HI 0x00000401 + +#define REG_A5XX_RBBM_PERFCTR_RAS_1_LO 0x00000402 + +#define REG_A5XX_RBBM_PERFCTR_RAS_1_HI 0x00000403 + +#define REG_A5XX_RBBM_PERFCTR_RAS_2_LO 0x00000404 + +#define REG_A5XX_RBBM_PERFCTR_RAS_2_HI 0x00000405 + +#define REG_A5XX_RBBM_PERFCTR_RAS_3_LO 0x00000406 + +#define REG_A5XX_RBBM_PERFCTR_RAS_3_HI 0x00000407 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_0_LO 0x00000408 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_0_HI 0x00000409 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_1_LO 0x0000040a + +#define REG_A5XX_RBBM_PERFCTR_UCHE_1_HI 0x0000040b + +#define REG_A5XX_RBBM_PERFCTR_UCHE_2_LO 0x0000040c + +#define REG_A5XX_RBBM_PERFCTR_UCHE_2_HI 0x0000040d + +#define REG_A5XX_RBBM_PERFCTR_UCHE_3_LO 0x0000040e + +#define REG_A5XX_RBBM_PERFCTR_UCHE_3_HI 0x0000040f + +#define REG_A5XX_RBBM_PERFCTR_UCHE_4_LO 0x00000410 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_4_HI 0x00000411 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_5_LO 0x00000412 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_5_HI 0x00000413 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_6_LO 0x00000414 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_6_HI 0x00000415 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_7_LO 0x00000416 + +#define REG_A5XX_RBBM_PERFCTR_UCHE_7_HI 0x00000417 + +#define REG_A5XX_RBBM_PERFCTR_TP_0_LO 0x00000418 + +#define REG_A5XX_RBBM_PERFCTR_TP_0_HI 0x00000419 + +#define REG_A5XX_RBBM_PERFCTR_TP_1_LO 0x0000041a + +#define REG_A5XX_RBBM_PERFCTR_TP_1_HI 0x0000041b + +#define REG_A5XX_RBBM_PERFCTR_TP_2_LO 0x0000041c + +#define REG_A5XX_RBBM_PERFCTR_TP_2_HI 0x0000041d + +#define REG_A5XX_RBBM_PERFCTR_TP_3_LO 0x0000041e + +#define REG_A5XX_RBBM_PERFCTR_TP_3_HI 0x0000041f + +#define REG_A5XX_RBBM_PERFCTR_TP_4_LO 0x00000420 + +#define REG_A5XX_RBBM_PERFCTR_TP_4_HI 0x00000421 + +#define REG_A5XX_RBBM_PERFCTR_TP_5_LO 0x00000422 + +#define REG_A5XX_RBBM_PERFCTR_TP_5_HI 0x00000423 + +#define REG_A5XX_RBBM_PERFCTR_TP_6_LO 0x00000424 + +#define REG_A5XX_RBBM_PERFCTR_TP_6_HI 0x00000425 + +#define REG_A5XX_RBBM_PERFCTR_TP_7_LO 0x00000426 + +#define REG_A5XX_RBBM_PERFCTR_TP_7_HI 0x00000427 + +#define REG_A5XX_RBBM_PERFCTR_SP_0_LO 0x00000428 + +#define REG_A5XX_RBBM_PERFCTR_SP_0_HI 0x00000429 + +#define REG_A5XX_RBBM_PERFCTR_SP_1_LO 0x0000042a + +#define REG_A5XX_RBBM_PERFCTR_SP_1_HI 0x0000042b + +#define REG_A5XX_RBBM_PERFCTR_SP_2_LO 0x0000042c + +#define REG_A5XX_RBBM_PERFCTR_SP_2_HI 0x0000042d + +#define REG_A5XX_RBBM_PERFCTR_SP_3_LO 0x0000042e + +#define REG_A5XX_RBBM_PERFCTR_SP_3_HI 0x0000042f + +#define REG_A5XX_RBBM_PERFCTR_SP_4_LO 0x00000430 + +#define REG_A5XX_RBBM_PERFCTR_SP_4_HI 0x00000431 + +#define REG_A5XX_RBBM_PERFCTR_SP_5_LO 0x00000432 + +#define REG_A5XX_RBBM_PERFCTR_SP_5_HI 0x00000433 + +#define REG_A5XX_RBBM_PERFCTR_SP_6_LO 0x00000434 + +#define REG_A5XX_RBBM_PERFCTR_SP_6_HI 0x00000435 + +#define REG_A5XX_RBBM_PERFCTR_SP_7_LO 0x00000436 + +#define REG_A5XX_RBBM_PERFCTR_SP_7_HI 0x00000437 + +#define REG_A5XX_RBBM_PERFCTR_SP_8_LO 0x00000438 + +#define REG_A5XX_RBBM_PERFCTR_SP_8_HI 0x00000439 + +#define REG_A5XX_RBBM_PERFCTR_SP_9_LO 0x0000043a + +#define REG_A5XX_RBBM_PERFCTR_SP_9_HI 0x0000043b + +#define REG_A5XX_RBBM_PERFCTR_SP_10_LO 0x0000043c + +#define REG_A5XX_RBBM_PERFCTR_SP_10_HI 0x0000043d + +#define REG_A5XX_RBBM_PERFCTR_SP_11_LO 0x0000043e + +#define REG_A5XX_RBBM_PERFCTR_SP_11_HI 0x0000043f + +#define REG_A5XX_RBBM_PERFCTR_RB_0_LO 0x00000440 + +#define REG_A5XX_RBBM_PERFCTR_RB_0_HI 0x00000441 + +#define REG_A5XX_RBBM_PERFCTR_RB_1_LO 0x00000442 + +#define REG_A5XX_RBBM_PERFCTR_RB_1_HI 0x00000443 + +#define REG_A5XX_RBBM_PERFCTR_RB_2_LO 0x00000444 + +#define REG_A5XX_RBBM_PERFCTR_RB_2_HI 0x00000445 + +#define REG_A5XX_RBBM_PERFCTR_RB_3_LO 0x00000446 + +#define REG_A5XX_RBBM_PERFCTR_RB_3_HI 0x00000447 + +#define REG_A5XX_RBBM_PERFCTR_RB_4_LO 0x00000448 + +#define REG_A5XX_RBBM_PERFCTR_RB_4_HI 0x00000449 + +#define REG_A5XX_RBBM_PERFCTR_RB_5_LO 0x0000044a + +#define REG_A5XX_RBBM_PERFCTR_RB_5_HI 0x0000044b + +#define REG_A5XX_RBBM_PERFCTR_RB_6_LO 0x0000044c + +#define REG_A5XX_RBBM_PERFCTR_RB_6_HI 0x0000044d + +#define REG_A5XX_RBBM_PERFCTR_RB_7_LO 0x0000044e + +#define REG_A5XX_RBBM_PERFCTR_RB_7_HI 0x0000044f + +#define REG_A5XX_RBBM_PERFCTR_VSC_0_LO 0x00000450 + +#define REG_A5XX_RBBM_PERFCTR_VSC_0_HI 0x00000451 + +#define REG_A5XX_RBBM_PERFCTR_VSC_1_LO 0x00000452 + +#define REG_A5XX_RBBM_PERFCTR_VSC_1_HI 0x00000453 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_0_LO 0x00000454 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_0_HI 0x00000455 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_1_LO 0x00000456 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_1_HI 0x00000457 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_2_LO 0x00000458 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_2_HI 0x00000459 + +#define REG_A5XX_RBBM_PERFCTR_LRZ_3_LO 0x0000045a + +#define REG_A5XX_RBBM_PERFCTR_LRZ_3_HI 0x0000045b + +#define REG_A5XX_RBBM_PERFCTR_CMP_0_LO 0x0000045c + +#define REG_A5XX_RBBM_PERFCTR_CMP_0_HI 0x0000045d + +#define REG_A5XX_RBBM_PERFCTR_CMP_1_LO 0x0000045e + +#define REG_A5XX_RBBM_PERFCTR_CMP_1_HI 0x0000045f + +#define REG_A5XX_RBBM_PERFCTR_CMP_2_LO 0x00000460 + +#define REG_A5XX_RBBM_PERFCTR_CMP_2_HI 0x00000461 + +#define REG_A5XX_RBBM_PERFCTR_CMP_3_LO 0x00000462 + +#define REG_A5XX_RBBM_PERFCTR_CMP_3_HI 0x00000463 + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_0 0x0000046b + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_1 0x0000046c + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_2 0x0000046d + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_3 0x0000046e + +#define REG_A5XX_RBBM_ALWAYSON_COUNTER_LO 0x000004d2 + +#define REG_A5XX_RBBM_ALWAYSON_COUNTER_HI 0x000004d3 + +#define REG_A5XX_RBBM_STATUS 0x000004f5 +#define A5XX_RBBM_STATUS_GPU_BUSY_IGN_AHB 0x80000000 +#define A5XX_RBBM_STATUS_GPU_BUSY_IGN_AHB_CP 0x40000000 +#define A5XX_RBBM_STATUS_HLSQ_BUSY 0x20000000 +#define A5XX_RBBM_STATUS_VSC_BUSY 0x10000000 +#define A5XX_RBBM_STATUS_TPL1_BUSY 0x08000000 +#define A5XX_RBBM_STATUS_SP_BUSY 0x04000000 +#define A5XX_RBBM_STATUS_UCHE_BUSY 0x02000000 +#define A5XX_RBBM_STATUS_VPC_BUSY 0x01000000 +#define A5XX_RBBM_STATUS_VFDP_BUSY 0x00800000 +#define A5XX_RBBM_STATUS_VFD_BUSY 0x00400000 +#define A5XX_RBBM_STATUS_TESS_BUSY 0x00200000 +#define A5XX_RBBM_STATUS_PC_VSD_BUSY 0x00100000 +#define A5XX_RBBM_STATUS_PC_DCALL_BUSY 0x00080000 +#define A5XX_RBBM_STATUS_GPMU_SLAVE_BUSY 0x00040000 +#define A5XX_RBBM_STATUS_DCOM_BUSY 0x00020000 +#define A5XX_RBBM_STATUS_COM_BUSY 0x00010000 +#define A5XX_RBBM_STATUS_LRZ_BUZY 0x00008000 +#define A5XX_RBBM_STATUS_A2D_DSP_BUSY 0x00004000 +#define A5XX_RBBM_STATUS_CCUFCHE_BUSY 0x00002000 +#define A5XX_RBBM_STATUS_RB_BUSY 0x00001000 +#define A5XX_RBBM_STATUS_RAS_BUSY 0x00000800 +#define A5XX_RBBM_STATUS_TSE_BUSY 0x00000400 +#define A5XX_RBBM_STATUS_VBIF_BUSY 0x00000200 +#define A5XX_RBBM_STATUS_GPU_BUSY_IGN_AHB_HYST 0x00000100 +#define A5XX_RBBM_STATUS_CP_BUSY_IGN_HYST 0x00000080 +#define A5XX_RBBM_STATUS_CP_BUSY 0x00000040 +#define A5XX_RBBM_STATUS_GPMU_MASTER_BUSY 0x00000020 +#define A5XX_RBBM_STATUS_CP_CRASH_BUSY 0x00000010 +#define A5XX_RBBM_STATUS_CP_ETS_BUSY 0x00000008 +#define A5XX_RBBM_STATUS_CP_PFP_BUSY 0x00000004 +#define A5XX_RBBM_STATUS_CP_ME_BUSY 0x00000002 +#define A5XX_RBBM_STATUS_HI_BUSY 0x00000001 + +#define REG_A5XX_RBBM_STATUS3 0x00000530 + +#define REG_A5XX_RBBM_INT_0_STATUS 0x000004e1 + +#define REG_A5XX_RBBM_AHB_ME_SPLIT_STATUS 0x000004f0 + +#define REG_A5XX_RBBM_AHB_PFP_SPLIT_STATUS 0x000004f1 + +#define REG_A5XX_RBBM_AHB_ETS_SPLIT_STATUS 0x000004f3 + +#define REG_A5XX_RBBM_AHB_ERROR_STATUS 0x000004f4 + +#define REG_A5XX_RBBM_PERFCTR_CNTL 0x00000464 + +#define REG_A5XX_RBBM_PERFCTR_LOAD_CMD0 0x00000465 + +#define REG_A5XX_RBBM_PERFCTR_LOAD_CMD1 0x00000466 + +#define REG_A5XX_RBBM_PERFCTR_LOAD_CMD2 0x00000467 + +#define REG_A5XX_RBBM_PERFCTR_LOAD_CMD3 0x00000468 + +#define REG_A5XX_RBBM_PERFCTR_LOAD_VALUE_LO 0x00000469 + +#define REG_A5XX_RBBM_PERFCTR_LOAD_VALUE_HI 0x0000046a + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_0 0x0000046b + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_1 0x0000046c + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_2 0x0000046d + +#define REG_A5XX_RBBM_PERFCTR_RBBM_SEL_3 0x0000046e + +#define REG_A5XX_RBBM_PERFCTR_GPU_BUSY_MASKED 0x0000046f + +#define REG_A5XX_RBBM_AHB_ERROR 0x000004ed + +#define REG_A5XX_RBBM_CFG_DBGBUS_EVENT_LOGIC 0x00000504 + +#define REG_A5XX_RBBM_CFG_DBGBUS_OVER 0x00000505 + +#define REG_A5XX_RBBM_CFG_DBGBUS_COUNT0 0x00000506 + +#define REG_A5XX_RBBM_CFG_DBGBUS_COUNT1 0x00000507 + +#define REG_A5XX_RBBM_CFG_DBGBUS_COUNT2 0x00000508 + +#define REG_A5XX_RBBM_CFG_DBGBUS_COUNT3 0x00000509 + +#define REG_A5XX_RBBM_CFG_DBGBUS_COUNT4 0x0000050a + +#define REG_A5XX_RBBM_CFG_DBGBUS_COUNT5 0x0000050b + +#define REG_A5XX_RBBM_CFG_DBGBUS_TRACE_ADDR 0x0000050c + +#define REG_A5XX_RBBM_CFG_DBGBUS_TRACE_BUF0 0x0000050d + +#define REG_A5XX_RBBM_CFG_DBGBUS_TRACE_BUF1 0x0000050e + +#define REG_A5XX_RBBM_CFG_DBGBUS_TRACE_BUF2 0x0000050f + +#define REG_A5XX_RBBM_CFG_DBGBUS_TRACE_BUF3 0x00000510 + +#define REG_A5XX_RBBM_CFG_DBGBUS_TRACE_BUF4 0x00000511 + +#define REG_A5XX_RBBM_CFG_DBGBUS_MISR0 0x00000512 + +#define REG_A5XX_RBBM_CFG_DBGBUS_MISR1 0x00000513 + +#define REG_A5XX_RBBM_ISDB_CNT 0x00000533 + +#define REG_A5XX_RBBM_SECVID_TRUST_CONFIG 0x0000f000 + +#define REG_A5XX_RBBM_SECVID_TRUST_CNTL 0x0000f400 + +#define REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO 0x0000f800 + +#define REG_A5XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI 0x0000f801 + +#define REG_A5XX_RBBM_SECVID_TSB_TRUSTED_SIZE 0x0000f802 + +#define REG_A5XX_RBBM_SECVID_TSB_CNTL 0x0000f803 + +#define REG_A5XX_RBBM_SECVID_TSB_COMP_STATUS_LO 0x0000f804 + +#define REG_A5XX_RBBM_SECVID_TSB_COMP_STATUS_HI 0x0000f805 + +#define REG_A5XX_RBBM_SECVID_TSB_UCHE_STATUS_LO 0x0000f806 + +#define REG_A5XX_RBBM_SECVID_TSB_UCHE_STATUS_HI 0x0000f807 + +#define REG_A5XX_RBBM_SECVID_TSB_ADDR_MODE_CNTL 0x0000f810 + +#define REG_A5XX_VSC_BIN_SIZE 0x00000bc2 +#define A5XX_VSC_BIN_SIZE_WIDTH__MASK 0x000000ff +#define A5XX_VSC_BIN_SIZE_WIDTH__SHIFT 0 +static inline uint32_t A5XX_VSC_BIN_SIZE_WIDTH(uint32_t val) +{ + assert(!(val & 0x1f)); + return ((val >> 5) << A5XX_VSC_BIN_SIZE_WIDTH__SHIFT) & A5XX_VSC_BIN_SIZE_WIDTH__MASK; +} +#define A5XX_VSC_BIN_SIZE_HEIGHT__MASK 0x0001fe00 +#define A5XX_VSC_BIN_SIZE_HEIGHT__SHIFT 9 +static inline uint32_t A5XX_VSC_BIN_SIZE_HEIGHT(uint32_t val) +{ + assert(!(val & 0x1f)); + return ((val >> 5) << A5XX_VSC_BIN_SIZE_HEIGHT__SHIFT) & A5XX_VSC_BIN_SIZE_HEIGHT__MASK; +} + +#define REG_A5XX_VSC_SIZE_ADDRESS_LO 0x00000bc3 + +#define REG_A5XX_VSC_SIZE_ADDRESS_HI 0x00000bc4 + +#define REG_A5XX_UNKNOWN_0BC5 0x00000bc5 + +#define REG_A5XX_UNKNOWN_0BC6 0x00000bc6 + +static inline uint32_t REG_A5XX_VSC_PIPE_CONFIG(uint32_t i0) { return 0x00000bd0 + 0x1*i0; } + +static inline uint32_t REG_A5XX_VSC_PIPE_CONFIG_REG(uint32_t i0) { return 0x00000bd0 + 0x1*i0; } +#define A5XX_VSC_PIPE_CONFIG_REG_X__MASK 0x000003ff +#define A5XX_VSC_PIPE_CONFIG_REG_X__SHIFT 0 +static inline uint32_t A5XX_VSC_PIPE_CONFIG_REG_X(uint32_t val) +{ + return ((val) << A5XX_VSC_PIPE_CONFIG_REG_X__SHIFT) & A5XX_VSC_PIPE_CONFIG_REG_X__MASK; +} +#define A5XX_VSC_PIPE_CONFIG_REG_Y__MASK 0x000ffc00 +#define A5XX_VSC_PIPE_CONFIG_REG_Y__SHIFT 10 +static inline uint32_t A5XX_VSC_PIPE_CONFIG_REG_Y(uint32_t val) +{ + return ((val) << A5XX_VSC_PIPE_CONFIG_REG_Y__SHIFT) & A5XX_VSC_PIPE_CONFIG_REG_Y__MASK; +} +#define A5XX_VSC_PIPE_CONFIG_REG_W__MASK 0x00f00000 +#define A5XX_VSC_PIPE_CONFIG_REG_W__SHIFT 20 +static inline uint32_t A5XX_VSC_PIPE_CONFIG_REG_W(uint32_t val) +{ + return ((val) << A5XX_VSC_PIPE_CONFIG_REG_W__SHIFT) & A5XX_VSC_PIPE_CONFIG_REG_W__MASK; +} +#define A5XX_VSC_PIPE_CONFIG_REG_H__MASK 0x0f000000 +#define A5XX_VSC_PIPE_CONFIG_REG_H__SHIFT 24 +static inline uint32_t A5XX_VSC_PIPE_CONFIG_REG_H(uint32_t val) +{ + return ((val) << A5XX_VSC_PIPE_CONFIG_REG_H__SHIFT) & A5XX_VSC_PIPE_CONFIG_REG_H__MASK; +} + +static inline uint32_t REG_A5XX_VSC_PIPE_DATA_ADDRESS(uint32_t i0) { return 0x00000be0 + 0x2*i0; } + +static inline uint32_t REG_A5XX_VSC_PIPE_DATA_ADDRESS_LO(uint32_t i0) { return 0x00000be0 + 0x2*i0; } + +static inline uint32_t REG_A5XX_VSC_PIPE_DATA_ADDRESS_HI(uint32_t i0) { return 0x00000be1 + 0x2*i0; } + +static inline uint32_t REG_A5XX_VSC_PIPE_DATA_LENGTH(uint32_t i0) { return 0x00000c00 + 0x1*i0; } + +static inline uint32_t REG_A5XX_VSC_PIPE_DATA_LENGTH_REG(uint32_t i0) { return 0x00000c00 + 0x1*i0; } + +#define REG_A5XX_VSC_PERFCTR_VSC_SEL_0 0x00000c60 + +#define REG_A5XX_VSC_PERFCTR_VSC_SEL_1 0x00000c61 + +#define REG_A5XX_VSC_RESOLVE_CNTL 0x00000cdd +#define A5XX_VSC_RESOLVE_CNTL_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_VSC_RESOLVE_CNTL_X__MASK 0x00007fff +#define A5XX_VSC_RESOLVE_CNTL_X__SHIFT 0 +static inline uint32_t A5XX_VSC_RESOLVE_CNTL_X(uint32_t val) +{ + return ((val) << A5XX_VSC_RESOLVE_CNTL_X__SHIFT) & A5XX_VSC_RESOLVE_CNTL_X__MASK; +} +#define A5XX_VSC_RESOLVE_CNTL_Y__MASK 0x7fff0000 +#define A5XX_VSC_RESOLVE_CNTL_Y__SHIFT 16 +static inline uint32_t A5XX_VSC_RESOLVE_CNTL_Y(uint32_t val) +{ + return ((val) << A5XX_VSC_RESOLVE_CNTL_Y__SHIFT) & A5XX_VSC_RESOLVE_CNTL_Y__MASK; +} + +#define REG_A5XX_GRAS_ADDR_MODE_CNTL 0x00000c81 + +#define REG_A5XX_GRAS_PERFCTR_TSE_SEL_0 0x00000c90 + +#define REG_A5XX_GRAS_PERFCTR_TSE_SEL_1 0x00000c91 + +#define REG_A5XX_GRAS_PERFCTR_TSE_SEL_2 0x00000c92 + +#define REG_A5XX_GRAS_PERFCTR_TSE_SEL_3 0x00000c93 + +#define REG_A5XX_GRAS_PERFCTR_RAS_SEL_0 0x00000c94 + +#define REG_A5XX_GRAS_PERFCTR_RAS_SEL_1 0x00000c95 + +#define REG_A5XX_GRAS_PERFCTR_RAS_SEL_2 0x00000c96 + +#define REG_A5XX_GRAS_PERFCTR_RAS_SEL_3 0x00000c97 + +#define REG_A5XX_GRAS_PERFCTR_LRZ_SEL_0 0x00000c98 + +#define REG_A5XX_GRAS_PERFCTR_LRZ_SEL_1 0x00000c99 + +#define REG_A5XX_GRAS_PERFCTR_LRZ_SEL_2 0x00000c9a + +#define REG_A5XX_GRAS_PERFCTR_LRZ_SEL_3 0x00000c9b + +#define REG_A5XX_RB_DBG_ECO_CNTL 0x00000cc4 + +#define REG_A5XX_RB_ADDR_MODE_CNTL 0x00000cc5 + +#define REG_A5XX_RB_MODE_CNTL 0x00000cc6 + +#define REG_A5XX_RB_CCU_CNTL 0x00000cc7 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_0 0x00000cd0 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_1 0x00000cd1 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_2 0x00000cd2 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_3 0x00000cd3 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_4 0x00000cd4 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_5 0x00000cd5 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_6 0x00000cd6 + +#define REG_A5XX_RB_PERFCTR_RB_SEL_7 0x00000cd7 + +#define REG_A5XX_RB_PERFCTR_CCU_SEL_0 0x00000cd8 + +#define REG_A5XX_RB_PERFCTR_CCU_SEL_1 0x00000cd9 + +#define REG_A5XX_RB_PERFCTR_CCU_SEL_2 0x00000cda + +#define REG_A5XX_RB_PERFCTR_CCU_SEL_3 0x00000cdb + +#define REG_A5XX_RB_POWERCTR_RB_SEL_0 0x00000ce0 + +#define REG_A5XX_RB_POWERCTR_RB_SEL_1 0x00000ce1 + +#define REG_A5XX_RB_POWERCTR_RB_SEL_2 0x00000ce2 + +#define REG_A5XX_RB_POWERCTR_RB_SEL_3 0x00000ce3 + +#define REG_A5XX_RB_POWERCTR_CCU_SEL_0 0x00000ce4 + +#define REG_A5XX_RB_POWERCTR_CCU_SEL_1 0x00000ce5 + +#define REG_A5XX_RB_PERFCTR_CMP_SEL_0 0x00000cec + +#define REG_A5XX_RB_PERFCTR_CMP_SEL_1 0x00000ced + +#define REG_A5XX_RB_PERFCTR_CMP_SEL_2 0x00000cee + +#define REG_A5XX_RB_PERFCTR_CMP_SEL_3 0x00000cef + +#define REG_A5XX_PC_DBG_ECO_CNTL 0x00000d00 +#define A5XX_PC_DBG_ECO_CNTL_TWOPASSUSEWFI 0x00000100 + +#define REG_A5XX_PC_ADDR_MODE_CNTL 0x00000d01 + +#define REG_A5XX_PC_MODE_CNTL 0x00000d02 + +#define REG_A5XX_UNKNOWN_0D08 0x00000d08 + +#define REG_A5XX_UNKNOWN_0D09 0x00000d09 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_0 0x00000d10 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_1 0x00000d11 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_2 0x00000d12 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_3 0x00000d13 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_4 0x00000d14 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_5 0x00000d15 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_6 0x00000d16 + +#define REG_A5XX_PC_PERFCTR_PC_SEL_7 0x00000d17 + +#define REG_A5XX_HLSQ_TIMEOUT_THRESHOLD_0 0x00000e00 + +#define REG_A5XX_HLSQ_TIMEOUT_THRESHOLD_1 0x00000e01 + +#define REG_A5XX_HLSQ_ADDR_MODE_CNTL 0x00000e05 + +#define REG_A5XX_HLSQ_MODE_CNTL 0x00000e06 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_0 0x00000e10 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_1 0x00000e11 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_2 0x00000e12 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_3 0x00000e13 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_4 0x00000e14 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_5 0x00000e15 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_6 0x00000e16 + +#define REG_A5XX_HLSQ_PERFCTR_HLSQ_SEL_7 0x00000e17 + +#define REG_A5XX_HLSQ_SPTP_RDSEL 0x00000f08 + +#define REG_A5XX_HLSQ_DBG_READ_SEL 0x0000bc00 + +#define REG_A5XX_HLSQ_DBG_AHB_READ_APERTURE 0x0000a000 + +#define REG_A5XX_VFD_ADDR_MODE_CNTL 0x00000e41 + +#define REG_A5XX_VFD_MODE_CNTL 0x00000e42 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_0 0x00000e50 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_1 0x00000e51 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_2 0x00000e52 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_3 0x00000e53 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_4 0x00000e54 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_5 0x00000e55 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_6 0x00000e56 + +#define REG_A5XX_VFD_PERFCTR_VFD_SEL_7 0x00000e57 + +#define REG_A5XX_VPC_DBG_ECO_CNTL 0x00000e60 + +#define REG_A5XX_VPC_ADDR_MODE_CNTL 0x00000e61 + +#define REG_A5XX_VPC_MODE_CNTL 0x00000e62 +#define A5XX_VPC_MODE_CNTL_BINNING_PASS 0x00000001 + +#define REG_A5XX_VPC_PERFCTR_VPC_SEL_0 0x00000e64 + +#define REG_A5XX_VPC_PERFCTR_VPC_SEL_1 0x00000e65 + +#define REG_A5XX_VPC_PERFCTR_VPC_SEL_2 0x00000e66 + +#define REG_A5XX_VPC_PERFCTR_VPC_SEL_3 0x00000e67 + +#define REG_A5XX_UCHE_ADDR_MODE_CNTL 0x00000e80 + +#define REG_A5XX_UCHE_SVM_CNTL 0x00000e82 + +#define REG_A5XX_UCHE_WRITE_THRU_BASE_LO 0x00000e87 + +#define REG_A5XX_UCHE_WRITE_THRU_BASE_HI 0x00000e88 + +#define REG_A5XX_UCHE_TRAP_BASE_LO 0x00000e89 + +#define REG_A5XX_UCHE_TRAP_BASE_HI 0x00000e8a + +#define REG_A5XX_UCHE_GMEM_RANGE_MIN_LO 0x00000e8b + +#define REG_A5XX_UCHE_GMEM_RANGE_MIN_HI 0x00000e8c + +#define REG_A5XX_UCHE_GMEM_RANGE_MAX_LO 0x00000e8d + +#define REG_A5XX_UCHE_GMEM_RANGE_MAX_HI 0x00000e8e + +#define REG_A5XX_UCHE_DBG_ECO_CNTL_2 0x00000e8f + +#define REG_A5XX_UCHE_DBG_ECO_CNTL 0x00000e90 + +#define REG_A5XX_UCHE_CACHE_INVALIDATE_MIN_LO 0x00000e91 + +#define REG_A5XX_UCHE_CACHE_INVALIDATE_MIN_HI 0x00000e92 + +#define REG_A5XX_UCHE_CACHE_INVALIDATE_MAX_LO 0x00000e93 + +#define REG_A5XX_UCHE_CACHE_INVALIDATE_MAX_HI 0x00000e94 + +#define REG_A5XX_UCHE_CACHE_INVALIDATE 0x00000e95 + +#define REG_A5XX_UCHE_CACHE_WAYS 0x00000e96 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_0 0x00000ea0 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_1 0x00000ea1 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_2 0x00000ea2 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_3 0x00000ea3 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_4 0x00000ea4 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_5 0x00000ea5 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_6 0x00000ea6 + +#define REG_A5XX_UCHE_PERFCTR_UCHE_SEL_7 0x00000ea7 + +#define REG_A5XX_UCHE_POWERCTR_UCHE_SEL_0 0x00000ea8 + +#define REG_A5XX_UCHE_POWERCTR_UCHE_SEL_1 0x00000ea9 + +#define REG_A5XX_UCHE_POWERCTR_UCHE_SEL_2 0x00000eaa + +#define REG_A5XX_UCHE_POWERCTR_UCHE_SEL_3 0x00000eab + +#define REG_A5XX_UCHE_TRAP_LOG_LO 0x00000eb1 + +#define REG_A5XX_UCHE_TRAP_LOG_HI 0x00000eb2 + +#define REG_A5XX_SP_DBG_ECO_CNTL 0x00000ec0 + +#define REG_A5XX_SP_ADDR_MODE_CNTL 0x00000ec1 + +#define REG_A5XX_SP_MODE_CNTL 0x00000ec2 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_0 0x00000ed0 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_1 0x00000ed1 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_2 0x00000ed2 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_3 0x00000ed3 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_4 0x00000ed4 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_5 0x00000ed5 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_6 0x00000ed6 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_7 0x00000ed7 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_8 0x00000ed8 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_9 0x00000ed9 + +#define REG_A5XX_SP_PERFCTR_SP_SEL_10 0x00000eda + +#define REG_A5XX_SP_PERFCTR_SP_SEL_11 0x00000edb + +#define REG_A5XX_SP_POWERCTR_SP_SEL_0 0x00000edc + +#define REG_A5XX_SP_POWERCTR_SP_SEL_1 0x00000edd + +#define REG_A5XX_SP_POWERCTR_SP_SEL_2 0x00000ede + +#define REG_A5XX_SP_POWERCTR_SP_SEL_3 0x00000edf + +#define REG_A5XX_TPL1_ADDR_MODE_CNTL 0x00000f01 + +#define REG_A5XX_TPL1_MODE_CNTL 0x00000f02 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_0 0x00000f10 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_1 0x00000f11 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_2 0x00000f12 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_3 0x00000f13 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_4 0x00000f14 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_5 0x00000f15 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_6 0x00000f16 + +#define REG_A5XX_TPL1_PERFCTR_TP_SEL_7 0x00000f17 + +#define REG_A5XX_TPL1_POWERCTR_TP_SEL_0 0x00000f18 + +#define REG_A5XX_TPL1_POWERCTR_TP_SEL_1 0x00000f19 + +#define REG_A5XX_TPL1_POWERCTR_TP_SEL_2 0x00000f1a + +#define REG_A5XX_TPL1_POWERCTR_TP_SEL_3 0x00000f1b + +#define REG_A5XX_VBIF_VERSION 0x00003000 + +#define REG_A5XX_VBIF_CLKON 0x00003001 + +#define REG_A5XX_VBIF_ABIT_SORT 0x00003028 + +#define REG_A5XX_VBIF_ABIT_SORT_CONF 0x00003029 + +#define REG_A5XX_VBIF_ROUND_ROBIN_QOS_ARB 0x00003049 + +#define REG_A5XX_VBIF_GATE_OFF_WRREQ_EN 0x0000302a + +#define REG_A5XX_VBIF_IN_RD_LIM_CONF0 0x0000302c + +#define REG_A5XX_VBIF_IN_RD_LIM_CONF1 0x0000302d + +#define REG_A5XX_VBIF_XIN_HALT_CTRL0 0x00003080 + +#define REG_A5XX_VBIF_XIN_HALT_CTRL1 0x00003081 + +#define REG_A5XX_VBIF_TEST_BUS_OUT_CTRL 0x00003084 + +#define REG_A5XX_VBIF_TEST_BUS1_CTRL0 0x00003085 + +#define REG_A5XX_VBIF_TEST_BUS1_CTRL1 0x00003086 + +#define REG_A5XX_VBIF_TEST_BUS2_CTRL0 0x00003087 + +#define REG_A5XX_VBIF_TEST_BUS2_CTRL1 0x00003088 + +#define REG_A5XX_VBIF_TEST_BUS_OUT 0x0000308c + +#define REG_A5XX_VBIF_PERF_CNT_EN0 0x000030c0 + +#define REG_A5XX_VBIF_PERF_CNT_EN1 0x000030c1 + +#define REG_A5XX_VBIF_PERF_CNT_EN2 0x000030c2 + +#define REG_A5XX_VBIF_PERF_CNT_EN3 0x000030c3 + +#define REG_A5XX_VBIF_PERF_CNT_SEL0 0x000030d0 + +#define REG_A5XX_VBIF_PERF_CNT_SEL1 0x000030d1 + +#define REG_A5XX_VBIF_PERF_CNT_SEL2 0x000030d2 + +#define REG_A5XX_VBIF_PERF_CNT_SEL3 0x000030d3 + +#define REG_A5XX_VBIF_PERF_CNT_LOW0 0x000030d8 + +#define REG_A5XX_VBIF_PERF_CNT_LOW1 0x000030d9 + +#define REG_A5XX_VBIF_PERF_CNT_LOW2 0x000030da + +#define REG_A5XX_VBIF_PERF_CNT_LOW3 0x000030db + +#define REG_A5XX_VBIF_PERF_CNT_HIGH0 0x000030e0 + +#define REG_A5XX_VBIF_PERF_CNT_HIGH1 0x000030e1 + +#define REG_A5XX_VBIF_PERF_CNT_HIGH2 0x000030e2 + +#define REG_A5XX_VBIF_PERF_CNT_HIGH3 0x000030e3 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_EN0 0x00003100 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_EN1 0x00003101 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_EN2 0x00003102 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_LOW0 0x00003110 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_LOW1 0x00003111 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_LOW2 0x00003112 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_HIGH0 0x00003118 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_HIGH1 0x00003119 + +#define REG_A5XX_VBIF_PERF_PWR_CNT_HIGH2 0x0000311a + +#define REG_A5XX_GPMU_INST_RAM_BASE 0x00008800 + +#define REG_A5XX_GPMU_DATA_RAM_BASE 0x00009800 + +#define REG_A5XX_GPMU_SP_POWER_CNTL 0x0000a881 + +#define REG_A5XX_GPMU_RBCCU_CLOCK_CNTL 0x0000a886 + +#define REG_A5XX_GPMU_RBCCU_POWER_CNTL 0x0000a887 + +#define REG_A5XX_GPMU_SP_PWR_CLK_STATUS 0x0000a88b +#define A5XX_GPMU_SP_PWR_CLK_STATUS_PWR_ON 0x00100000 + +#define REG_A5XX_GPMU_RBCCU_PWR_CLK_STATUS 0x0000a88d +#define A5XX_GPMU_RBCCU_PWR_CLK_STATUS_PWR_ON 0x00100000 + +#define REG_A5XX_GPMU_PWR_COL_STAGGER_DELAY 0x0000a891 + +#define REG_A5XX_GPMU_PWR_COL_INTER_FRAME_CTRL 0x0000a892 + +#define REG_A5XX_GPMU_PWR_COL_INTER_FRAME_HYST 0x0000a893 + +#define REG_A5XX_GPMU_PWR_COL_BINNING_CTRL 0x0000a894 + +#define REG_A5XX_GPMU_CLOCK_THROTTLE_CTRL 0x0000a8a3 + +#define REG_A5XX_GPMU_WFI_CONFIG 0x0000a8c1 + +#define REG_A5XX_GPMU_RBBM_INTR_INFO 0x0000a8d6 + +#define REG_A5XX_GPMU_CM3_SYSRESET 0x0000a8d8 + +#define REG_A5XX_GPMU_GENERAL_0 0x0000a8e0 + +#define REG_A5XX_GPMU_GENERAL_1 0x0000a8e1 + +#define REG_A5XX_SP_POWER_COUNTER_0_LO 0x0000a840 + +#define REG_A5XX_SP_POWER_COUNTER_0_HI 0x0000a841 + +#define REG_A5XX_SP_POWER_COUNTER_1_LO 0x0000a842 + +#define REG_A5XX_SP_POWER_COUNTER_1_HI 0x0000a843 + +#define REG_A5XX_SP_POWER_COUNTER_2_LO 0x0000a844 + +#define REG_A5XX_SP_POWER_COUNTER_2_HI 0x0000a845 + +#define REG_A5XX_SP_POWER_COUNTER_3_LO 0x0000a846 + +#define REG_A5XX_SP_POWER_COUNTER_3_HI 0x0000a847 + +#define REG_A5XX_TP_POWER_COUNTER_0_LO 0x0000a848 + +#define REG_A5XX_TP_POWER_COUNTER_0_HI 0x0000a849 + +#define REG_A5XX_TP_POWER_COUNTER_1_LO 0x0000a84a + +#define REG_A5XX_TP_POWER_COUNTER_1_HI 0x0000a84b + +#define REG_A5XX_TP_POWER_COUNTER_2_LO 0x0000a84c + +#define REG_A5XX_TP_POWER_COUNTER_2_HI 0x0000a84d + +#define REG_A5XX_TP_POWER_COUNTER_3_LO 0x0000a84e + +#define REG_A5XX_TP_POWER_COUNTER_3_HI 0x0000a84f + +#define REG_A5XX_RB_POWER_COUNTER_0_LO 0x0000a850 + +#define REG_A5XX_RB_POWER_COUNTER_0_HI 0x0000a851 + +#define REG_A5XX_RB_POWER_COUNTER_1_LO 0x0000a852 + +#define REG_A5XX_RB_POWER_COUNTER_1_HI 0x0000a853 + +#define REG_A5XX_RB_POWER_COUNTER_2_LO 0x0000a854 + +#define REG_A5XX_RB_POWER_COUNTER_2_HI 0x0000a855 + +#define REG_A5XX_RB_POWER_COUNTER_3_LO 0x0000a856 + +#define REG_A5XX_RB_POWER_COUNTER_3_HI 0x0000a857 + +#define REG_A5XX_CCU_POWER_COUNTER_0_LO 0x0000a858 + +#define REG_A5XX_CCU_POWER_COUNTER_0_HI 0x0000a859 + +#define REG_A5XX_CCU_POWER_COUNTER_1_LO 0x0000a85a + +#define REG_A5XX_CCU_POWER_COUNTER_1_HI 0x0000a85b + +#define REG_A5XX_UCHE_POWER_COUNTER_0_LO 0x0000a85c + +#define REG_A5XX_UCHE_POWER_COUNTER_0_HI 0x0000a85d + +#define REG_A5XX_UCHE_POWER_COUNTER_1_LO 0x0000a85e + +#define REG_A5XX_UCHE_POWER_COUNTER_1_HI 0x0000a85f + +#define REG_A5XX_UCHE_POWER_COUNTER_2_LO 0x0000a860 + +#define REG_A5XX_UCHE_POWER_COUNTER_2_HI 0x0000a861 + +#define REG_A5XX_UCHE_POWER_COUNTER_3_LO 0x0000a862 + +#define REG_A5XX_UCHE_POWER_COUNTER_3_HI 0x0000a863 + +#define REG_A5XX_CP_POWER_COUNTER_0_LO 0x0000a864 + +#define REG_A5XX_CP_POWER_COUNTER_0_HI 0x0000a865 + +#define REG_A5XX_CP_POWER_COUNTER_1_LO 0x0000a866 + +#define REG_A5XX_CP_POWER_COUNTER_1_HI 0x0000a867 + +#define REG_A5XX_CP_POWER_COUNTER_2_LO 0x0000a868 + +#define REG_A5XX_CP_POWER_COUNTER_2_HI 0x0000a869 + +#define REG_A5XX_CP_POWER_COUNTER_3_LO 0x0000a86a + +#define REG_A5XX_CP_POWER_COUNTER_3_HI 0x0000a86b + +#define REG_A5XX_GPMU_POWER_COUNTER_0_LO 0x0000a86c + +#define REG_A5XX_GPMU_POWER_COUNTER_0_HI 0x0000a86d + +#define REG_A5XX_GPMU_POWER_COUNTER_1_LO 0x0000a86e + +#define REG_A5XX_GPMU_POWER_COUNTER_1_HI 0x0000a86f + +#define REG_A5XX_GPMU_POWER_COUNTER_2_LO 0x0000a870 + +#define REG_A5XX_GPMU_POWER_COUNTER_2_HI 0x0000a871 + +#define REG_A5XX_GPMU_POWER_COUNTER_3_LO 0x0000a872 + +#define REG_A5XX_GPMU_POWER_COUNTER_3_HI 0x0000a873 + +#define REG_A5XX_GPMU_POWER_COUNTER_4_LO 0x0000a874 + +#define REG_A5XX_GPMU_POWER_COUNTER_4_HI 0x0000a875 + +#define REG_A5XX_GPMU_POWER_COUNTER_5_LO 0x0000a876 + +#define REG_A5XX_GPMU_POWER_COUNTER_5_HI 0x0000a877 + +#define REG_A5XX_GPMU_POWER_COUNTER_ENABLE 0x0000a878 + +#define REG_A5XX_GPMU_ALWAYS_ON_COUNTER_LO 0x0000a879 + +#define REG_A5XX_GPMU_ALWAYS_ON_COUNTER_HI 0x0000a87a + +#define REG_A5XX_GPMU_ALWAYS_ON_COUNTER_RESET 0x0000a87b + +#define REG_A5XX_GPMU_POWER_COUNTER_SELECT_0 0x0000a87c + +#define REG_A5XX_GPMU_POWER_COUNTER_SELECT_1 0x0000a87d + +#define REG_A5XX_GPMU_CLOCK_THROTTLE_CTRL 0x0000a8a3 + +#define REG_A5XX_GPMU_THROTTLE_UNMASK_FORCE_CTRL 0x0000a8a8 + +#define REG_A5XX_GPMU_TEMP_SENSOR_ID 0x0000ac00 + +#define REG_A5XX_GPMU_TEMP_SENSOR_CONFIG 0x0000ac01 + +#define REG_A5XX_GPMU_TEMP_VAL 0x0000ac02 + +#define REG_A5XX_GPMU_DELTA_TEMP_THRESHOLD 0x0000ac03 + +#define REG_A5XX_GPMU_TEMP_THRESHOLD_INTR_STATUS 0x0000ac05 + +#define REG_A5XX_GPMU_TEMP_THRESHOLD_INTR_EN_MASK 0x0000ac06 + +#define REG_A5XX_GPMU_LEAKAGE_TEMP_COEFF_0_1 0x0000ac40 + +#define REG_A5XX_GPMU_LEAKAGE_TEMP_COEFF_2_3 0x0000ac41 + +#define REG_A5XX_GPMU_LEAKAGE_VTG_COEFF_0_1 0x0000ac42 + +#define REG_A5XX_GPMU_LEAKAGE_VTG_COEFF_2_3 0x0000ac43 + +#define REG_A5XX_GPMU_BASE_LEAKAGE 0x0000ac46 + +#define REG_A5XX_GPMU_GPMU_VOLTAGE 0x0000ac60 + +#define REG_A5XX_GPMU_GPMU_VOLTAGE_INTR_STATUS 0x0000ac61 + +#define REG_A5XX_GPMU_GPMU_VOLTAGE_INTR_EN_MASK 0x0000ac62 + +#define REG_A5XX_GPMU_GPMU_PWR_THRESHOLD 0x0000ac80 + +#define REG_A5XX_GPMU_GPMU_LLM_GLM_SLEEP_CTRL 0x0000acc4 + +#define REG_A5XX_GPMU_GPMU_LLM_GLM_SLEEP_STATUS 0x0000acc5 + +#define REG_A5XX_GDPM_CONFIG1 0x0000b80c + +#define REG_A5XX_GDPM_CONFIG2 0x0000b80d + +#define REG_A5XX_GDPM_INT_EN 0x0000b80f + +#define REG_A5XX_GDPM_INT_MASK 0x0000b811 + +#define REG_A5XX_GPMU_BEC_ENABLE 0x0000b9a0 + +#define REG_A5XX_GPU_CS_SENSOR_GENERAL_STATUS 0x0000c41a + +#define REG_A5XX_GPU_CS_AMP_CALIBRATION_STATUS1_0 0x0000c41d + +#define REG_A5XX_GPU_CS_AMP_CALIBRATION_STATUS1_2 0x0000c41f + +#define REG_A5XX_GPU_CS_AMP_CALIBRATION_STATUS1_4 0x0000c421 + +#define REG_A5XX_GPU_CS_ENABLE_REG 0x0000c520 + +#define REG_A5XX_GPU_CS_AMP_CALIBRATION_CONTROL1 0x0000c557 + +#define REG_A5XX_GRAS_CL_CNTL 0x0000e000 +#define A5XX_GRAS_CL_CNTL_ZERO_GB_SCALE_Z 0x00000040 + +#define REG_A5XX_UNKNOWN_E001 0x0000e001 + +#define REG_A5XX_UNKNOWN_E004 0x0000e004 + +#define REG_A5XX_GRAS_CNTL 0x0000e005 +#define A5XX_GRAS_CNTL_VARYING 0x00000001 +#define A5XX_GRAS_CNTL_UNK3 0x00000008 +#define A5XX_GRAS_CNTL_XCOORD 0x00000040 +#define A5XX_GRAS_CNTL_YCOORD 0x00000080 +#define A5XX_GRAS_CNTL_ZCOORD 0x00000100 +#define A5XX_GRAS_CNTL_WCOORD 0x00000200 + +#define REG_A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ 0x0000e006 +#define A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_HORZ__MASK 0x000003ff +#define A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_HORZ__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_HORZ(uint32_t val) +{ + return ((val) << A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_HORZ__SHIFT) & A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_HORZ__MASK; +} +#define A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_VERT__MASK 0x000ffc00 +#define A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_VERT__SHIFT 10 +static inline uint32_t A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_VERT(uint32_t val) +{ + return ((val) << A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_VERT__SHIFT) & A5XX_GRAS_CL_GUARDBAND_CLIP_ADJ_VERT__MASK; +} + +#define REG_A5XX_GRAS_CL_VPORT_XOFFSET_0 0x0000e010 +#define A5XX_GRAS_CL_VPORT_XOFFSET_0__MASK 0xffffffff +#define A5XX_GRAS_CL_VPORT_XOFFSET_0__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_VPORT_XOFFSET_0(float val) +{ + return ((fui(val)) << A5XX_GRAS_CL_VPORT_XOFFSET_0__SHIFT) & A5XX_GRAS_CL_VPORT_XOFFSET_0__MASK; +} + +#define REG_A5XX_GRAS_CL_VPORT_XSCALE_0 0x0000e011 +#define A5XX_GRAS_CL_VPORT_XSCALE_0__MASK 0xffffffff +#define A5XX_GRAS_CL_VPORT_XSCALE_0__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_VPORT_XSCALE_0(float val) +{ + return ((fui(val)) << A5XX_GRAS_CL_VPORT_XSCALE_0__SHIFT) & A5XX_GRAS_CL_VPORT_XSCALE_0__MASK; +} + +#define REG_A5XX_GRAS_CL_VPORT_YOFFSET_0 0x0000e012 +#define A5XX_GRAS_CL_VPORT_YOFFSET_0__MASK 0xffffffff +#define A5XX_GRAS_CL_VPORT_YOFFSET_0__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_VPORT_YOFFSET_0(float val) +{ + return ((fui(val)) << A5XX_GRAS_CL_VPORT_YOFFSET_0__SHIFT) & A5XX_GRAS_CL_VPORT_YOFFSET_0__MASK; +} + +#define REG_A5XX_GRAS_CL_VPORT_YSCALE_0 0x0000e013 +#define A5XX_GRAS_CL_VPORT_YSCALE_0__MASK 0xffffffff +#define A5XX_GRAS_CL_VPORT_YSCALE_0__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_VPORT_YSCALE_0(float val) +{ + return ((fui(val)) << A5XX_GRAS_CL_VPORT_YSCALE_0__SHIFT) & A5XX_GRAS_CL_VPORT_YSCALE_0__MASK; +} + +#define REG_A5XX_GRAS_CL_VPORT_ZOFFSET_0 0x0000e014 +#define A5XX_GRAS_CL_VPORT_ZOFFSET_0__MASK 0xffffffff +#define A5XX_GRAS_CL_VPORT_ZOFFSET_0__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_VPORT_ZOFFSET_0(float val) +{ + return ((fui(val)) << A5XX_GRAS_CL_VPORT_ZOFFSET_0__SHIFT) & A5XX_GRAS_CL_VPORT_ZOFFSET_0__MASK; +} + +#define REG_A5XX_GRAS_CL_VPORT_ZSCALE_0 0x0000e015 +#define A5XX_GRAS_CL_VPORT_ZSCALE_0__MASK 0xffffffff +#define A5XX_GRAS_CL_VPORT_ZSCALE_0__SHIFT 0 +static inline uint32_t A5XX_GRAS_CL_VPORT_ZSCALE_0(float val) +{ + return ((fui(val)) << A5XX_GRAS_CL_VPORT_ZSCALE_0__SHIFT) & A5XX_GRAS_CL_VPORT_ZSCALE_0__MASK; +} + +#define REG_A5XX_GRAS_SU_CNTL 0x0000e090 +#define A5XX_GRAS_SU_CNTL_CULL_FRONT 0x00000001 +#define A5XX_GRAS_SU_CNTL_CULL_BACK 0x00000002 +#define A5XX_GRAS_SU_CNTL_FRONT_CW 0x00000004 +#define A5XX_GRAS_SU_CNTL_LINEHALFWIDTH__MASK 0x000007f8 +#define A5XX_GRAS_SU_CNTL_LINEHALFWIDTH__SHIFT 3 +static inline uint32_t A5XX_GRAS_SU_CNTL_LINEHALFWIDTH(float val) +{ + return ((((int32_t)(val * 4.0))) << A5XX_GRAS_SU_CNTL_LINEHALFWIDTH__SHIFT) & A5XX_GRAS_SU_CNTL_LINEHALFWIDTH__MASK; +} +#define A5XX_GRAS_SU_CNTL_POLY_OFFSET 0x00000800 +#define A5XX_GRAS_SU_CNTL_MSAA_ENABLE 0x00002000 + +#define REG_A5XX_GRAS_SU_POINT_MINMAX 0x0000e091 +#define A5XX_GRAS_SU_POINT_MINMAX_MIN__MASK 0x0000ffff +#define A5XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT 0 +static inline uint32_t A5XX_GRAS_SU_POINT_MINMAX_MIN(float val) +{ + return ((((uint32_t)(val * 16.0))) << A5XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A5XX_GRAS_SU_POINT_MINMAX_MIN__MASK; +} +#define A5XX_GRAS_SU_POINT_MINMAX_MAX__MASK 0xffff0000 +#define A5XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT 16 +static inline uint32_t A5XX_GRAS_SU_POINT_MINMAX_MAX(float val) +{ + return ((((uint32_t)(val * 16.0))) << A5XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A5XX_GRAS_SU_POINT_MINMAX_MAX__MASK; +} + +#define REG_A5XX_GRAS_SU_POINT_SIZE 0x0000e092 +#define A5XX_GRAS_SU_POINT_SIZE__MASK 0xffffffff +#define A5XX_GRAS_SU_POINT_SIZE__SHIFT 0 +static inline uint32_t A5XX_GRAS_SU_POINT_SIZE(float val) +{ + return ((((int32_t)(val * 16.0))) << A5XX_GRAS_SU_POINT_SIZE__SHIFT) & A5XX_GRAS_SU_POINT_SIZE__MASK; +} + +#define REG_A5XX_UNKNOWN_E093 0x0000e093 + +#define REG_A5XX_GRAS_SU_DEPTH_PLANE_CNTL 0x0000e094 +#define A5XX_GRAS_SU_DEPTH_PLANE_CNTL_FRAG_WRITES_Z 0x00000001 +#define A5XX_GRAS_SU_DEPTH_PLANE_CNTL_UNK1 0x00000002 + +#define REG_A5XX_GRAS_SU_POLY_OFFSET_SCALE 0x0000e095 +#define A5XX_GRAS_SU_POLY_OFFSET_SCALE__MASK 0xffffffff +#define A5XX_GRAS_SU_POLY_OFFSET_SCALE__SHIFT 0 +static inline uint32_t A5XX_GRAS_SU_POLY_OFFSET_SCALE(float val) +{ + return ((fui(val)) << A5XX_GRAS_SU_POLY_OFFSET_SCALE__SHIFT) & A5XX_GRAS_SU_POLY_OFFSET_SCALE__MASK; +} + +#define REG_A5XX_GRAS_SU_POLY_OFFSET_OFFSET 0x0000e096 +#define A5XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK 0xffffffff +#define A5XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT 0 +static inline uint32_t A5XX_GRAS_SU_POLY_OFFSET_OFFSET(float val) +{ + return ((fui(val)) << A5XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A5XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK; +} + +#define REG_A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP 0x0000e097 +#define A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP__MASK 0xffffffff +#define A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP__SHIFT 0 +static inline uint32_t A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP(float val) +{ + return ((fui(val)) << A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP__SHIFT) & A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP__MASK; +} + +#define REG_A5XX_GRAS_SU_DEPTH_BUFFER_INFO 0x0000e098 +#define A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT__MASK 0x00000007 +#define A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT__SHIFT 0 +static inline uint32_t A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT(enum a5xx_depth_format val) +{ + return ((val) << A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT__SHIFT) & A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT__MASK; +} + +#define REG_A5XX_GRAS_SU_CONSERVATIVE_RAS_CNTL 0x0000e099 + +#define REG_A5XX_GRAS_SC_CNTL 0x0000e0a0 +#define A5XX_GRAS_SC_CNTL_BINNING_PASS 0x00000001 +#define A5XX_GRAS_SC_CNTL_SAMPLES_PASSED 0x00008000 + +#define REG_A5XX_GRAS_SC_BIN_CNTL 0x0000e0a1 + +#define REG_A5XX_GRAS_SC_RAS_MSAA_CNTL 0x0000e0a2 +#define A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES__MASK 0x00000003 +#define A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES(enum a3xx_msaa_samples val) +{ + return ((val) << A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES__SHIFT) & A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES__MASK; +} + +#define REG_A5XX_GRAS_SC_DEST_MSAA_CNTL 0x0000e0a3 +#define A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES__MASK 0x00000003 +#define A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES(enum a3xx_msaa_samples val) +{ + return ((val) << A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES__SHIFT) & A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES__MASK; +} +#define A5XX_GRAS_SC_DEST_MSAA_CNTL_MSAA_DISABLE 0x00000004 + +#define REG_A5XX_GRAS_SC_SCREEN_SCISSOR_CNTL 0x0000e0a4 + +#define REG_A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0 0x0000e0aa +#define A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X__MASK 0x00007fff +#define A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X__SHIFT) & A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X__MASK; +} +#define A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y__MASK 0x7fff0000 +#define A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y__SHIFT 16 +static inline uint32_t A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y__SHIFT) & A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y__MASK; +} + +#define REG_A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0 0x0000e0ab +#define A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_X__MASK 0x00007fff +#define A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_X__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_X(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_X__SHIFT) & A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_X__MASK; +} +#define A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_Y__MASK 0x7fff0000 +#define A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_Y__SHIFT 16 +static inline uint32_t A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_Y(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_Y__SHIFT) & A5XX_GRAS_SC_SCREEN_SCISSOR_BR_0_Y__MASK; +} + +#define REG_A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0 0x0000e0ca +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X__MASK 0x00007fff +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X__SHIFT) & A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X__MASK; +} +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y__MASK 0x7fff0000 +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y__SHIFT 16 +static inline uint32_t A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y__SHIFT) & A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y__MASK; +} + +#define REG_A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0 0x0000e0cb +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_X__MASK 0x00007fff +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_X__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_X(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_X__SHIFT) & A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_X__MASK; +} +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_Y__MASK 0x7fff0000 +#define A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_Y__SHIFT 16 +static inline uint32_t A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_Y(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_Y__SHIFT) & A5XX_GRAS_SC_VIEWPORT_SCISSOR_BR_0_Y__MASK; +} + +#define REG_A5XX_GRAS_SC_WINDOW_SCISSOR_TL 0x0000e0ea +#define A5XX_GRAS_SC_WINDOW_SCISSOR_TL_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X__MASK 0x00007fff +#define A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X__SHIFT) & A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X__MASK; +} +#define A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__MASK 0x7fff0000 +#define A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__SHIFT 16 +static inline uint32_t A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__SHIFT) & A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__MASK; +} + +#define REG_A5XX_GRAS_SC_WINDOW_SCISSOR_BR 0x0000e0eb +#define A5XX_GRAS_SC_WINDOW_SCISSOR_BR_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X__MASK 0x00007fff +#define A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X__SHIFT 0 +static inline uint32_t A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X__SHIFT) & A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X__MASK; +} +#define A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__MASK 0x7fff0000 +#define A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__SHIFT 16 +static inline uint32_t A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(uint32_t val) +{ + return ((val) << A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__SHIFT) & A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__MASK; +} + +#define REG_A5XX_GRAS_LRZ_CNTL 0x0000e100 +#define A5XX_GRAS_LRZ_CNTL_ENABLE 0x00000001 +#define A5XX_GRAS_LRZ_CNTL_LRZ_WRITE 0x00000002 +#define A5XX_GRAS_LRZ_CNTL_GREATER 0x00000004 + +#define REG_A5XX_GRAS_LRZ_BUFFER_BASE_LO 0x0000e101 + +#define REG_A5XX_GRAS_LRZ_BUFFER_BASE_HI 0x0000e102 + +#define REG_A5XX_GRAS_LRZ_BUFFER_PITCH 0x0000e103 +#define A5XX_GRAS_LRZ_BUFFER_PITCH__MASK 0xffffffff +#define A5XX_GRAS_LRZ_BUFFER_PITCH__SHIFT 0 +static inline uint32_t A5XX_GRAS_LRZ_BUFFER_PITCH(uint32_t val) +{ + assert(!(val & 0x1f)); + return ((val >> 5) << A5XX_GRAS_LRZ_BUFFER_PITCH__SHIFT) & A5XX_GRAS_LRZ_BUFFER_PITCH__MASK; +} + +#define REG_A5XX_GRAS_LRZ_FAST_CLEAR_BUFFER_BASE_LO 0x0000e104 + +#define REG_A5XX_GRAS_LRZ_FAST_CLEAR_BUFFER_BASE_HI 0x0000e105 + +#define REG_A5XX_RB_CNTL 0x0000e140 +#define A5XX_RB_CNTL_WIDTH__MASK 0x000000ff +#define A5XX_RB_CNTL_WIDTH__SHIFT 0 +static inline uint32_t A5XX_RB_CNTL_WIDTH(uint32_t val) +{ + assert(!(val & 0x1f)); + return ((val >> 5) << A5XX_RB_CNTL_WIDTH__SHIFT) & A5XX_RB_CNTL_WIDTH__MASK; +} +#define A5XX_RB_CNTL_HEIGHT__MASK 0x0001fe00 +#define A5XX_RB_CNTL_HEIGHT__SHIFT 9 +static inline uint32_t A5XX_RB_CNTL_HEIGHT(uint32_t val) +{ + assert(!(val & 0x1f)); + return ((val >> 5) << A5XX_RB_CNTL_HEIGHT__SHIFT) & A5XX_RB_CNTL_HEIGHT__MASK; +} +#define A5XX_RB_CNTL_BYPASS 0x00020000 + +#define REG_A5XX_RB_RENDER_CNTL 0x0000e141 +#define A5XX_RB_RENDER_CNTL_BINNING_PASS 0x00000001 +#define A5XX_RB_RENDER_CNTL_SAMPLES_PASSED 0x00000040 +#define A5XX_RB_RENDER_CNTL_DISABLE_COLOR_PIPE 0x00000080 +#define A5XX_RB_RENDER_CNTL_FLAG_DEPTH 0x00004000 +#define A5XX_RB_RENDER_CNTL_FLAG_DEPTH2 0x00008000 +#define A5XX_RB_RENDER_CNTL_FLAG_MRTS__MASK 0x00ff0000 +#define A5XX_RB_RENDER_CNTL_FLAG_MRTS__SHIFT 16 +static inline uint32_t A5XX_RB_RENDER_CNTL_FLAG_MRTS(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_CNTL_FLAG_MRTS__SHIFT) & A5XX_RB_RENDER_CNTL_FLAG_MRTS__MASK; +} +#define A5XX_RB_RENDER_CNTL_FLAG_MRTS2__MASK 0xff000000 +#define A5XX_RB_RENDER_CNTL_FLAG_MRTS2__SHIFT 24 +static inline uint32_t A5XX_RB_RENDER_CNTL_FLAG_MRTS2(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_CNTL_FLAG_MRTS2__SHIFT) & A5XX_RB_RENDER_CNTL_FLAG_MRTS2__MASK; +} + +#define REG_A5XX_RB_RAS_MSAA_CNTL 0x0000e142 +#define A5XX_RB_RAS_MSAA_CNTL_SAMPLES__MASK 0x00000003 +#define A5XX_RB_RAS_MSAA_CNTL_SAMPLES__SHIFT 0 +static inline uint32_t A5XX_RB_RAS_MSAA_CNTL_SAMPLES(enum a3xx_msaa_samples val) +{ + return ((val) << A5XX_RB_RAS_MSAA_CNTL_SAMPLES__SHIFT) & A5XX_RB_RAS_MSAA_CNTL_SAMPLES__MASK; +} + +#define REG_A5XX_RB_DEST_MSAA_CNTL 0x0000e143 +#define A5XX_RB_DEST_MSAA_CNTL_SAMPLES__MASK 0x00000003 +#define A5XX_RB_DEST_MSAA_CNTL_SAMPLES__SHIFT 0 +static inline uint32_t A5XX_RB_DEST_MSAA_CNTL_SAMPLES(enum a3xx_msaa_samples val) +{ + return ((val) << A5XX_RB_DEST_MSAA_CNTL_SAMPLES__SHIFT) & A5XX_RB_DEST_MSAA_CNTL_SAMPLES__MASK; +} +#define A5XX_RB_DEST_MSAA_CNTL_MSAA_DISABLE 0x00000004 + +#define REG_A5XX_RB_RENDER_CONTROL0 0x0000e144 +#define A5XX_RB_RENDER_CONTROL0_VARYING 0x00000001 +#define A5XX_RB_RENDER_CONTROL0_UNK3 0x00000008 +#define A5XX_RB_RENDER_CONTROL0_XCOORD 0x00000040 +#define A5XX_RB_RENDER_CONTROL0_YCOORD 0x00000080 +#define A5XX_RB_RENDER_CONTROL0_ZCOORD 0x00000100 +#define A5XX_RB_RENDER_CONTROL0_WCOORD 0x00000200 + +#define REG_A5XX_RB_RENDER_CONTROL1 0x0000e145 +#define A5XX_RB_RENDER_CONTROL1_FACENESS 0x00000002 + +#define REG_A5XX_RB_FS_OUTPUT_CNTL 0x0000e146 +#define A5XX_RB_FS_OUTPUT_CNTL_MRT__MASK 0x0000000f +#define A5XX_RB_FS_OUTPUT_CNTL_MRT__SHIFT 0 +static inline uint32_t A5XX_RB_FS_OUTPUT_CNTL_MRT(uint32_t val) +{ + return ((val) << A5XX_RB_FS_OUTPUT_CNTL_MRT__SHIFT) & A5XX_RB_FS_OUTPUT_CNTL_MRT__MASK; +} +#define A5XX_RB_FS_OUTPUT_CNTL_FRAG_WRITES_Z 0x00000020 + +#define REG_A5XX_RB_RENDER_COMPONENTS 0x0000e147 +#define A5XX_RB_RENDER_COMPONENTS_RT0__MASK 0x0000000f +#define A5XX_RB_RENDER_COMPONENTS_RT0__SHIFT 0 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT0(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT0__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT0__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT1__MASK 0x000000f0 +#define A5XX_RB_RENDER_COMPONENTS_RT1__SHIFT 4 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT1(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT1__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT1__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT2__MASK 0x00000f00 +#define A5XX_RB_RENDER_COMPONENTS_RT2__SHIFT 8 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT2(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT2__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT2__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT3__MASK 0x0000f000 +#define A5XX_RB_RENDER_COMPONENTS_RT3__SHIFT 12 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT3(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT3__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT3__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT4__MASK 0x000f0000 +#define A5XX_RB_RENDER_COMPONENTS_RT4__SHIFT 16 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT4(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT4__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT4__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT5__MASK 0x00f00000 +#define A5XX_RB_RENDER_COMPONENTS_RT5__SHIFT 20 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT5(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT5__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT5__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT6__MASK 0x0f000000 +#define A5XX_RB_RENDER_COMPONENTS_RT6__SHIFT 24 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT6(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT6__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT6__MASK; +} +#define A5XX_RB_RENDER_COMPONENTS_RT7__MASK 0xf0000000 +#define A5XX_RB_RENDER_COMPONENTS_RT7__SHIFT 28 +static inline uint32_t A5XX_RB_RENDER_COMPONENTS_RT7(uint32_t val) +{ + return ((val) << A5XX_RB_RENDER_COMPONENTS_RT7__SHIFT) & A5XX_RB_RENDER_COMPONENTS_RT7__MASK; +} + +static inline uint32_t REG_A5XX_RB_MRT(uint32_t i0) { return 0x0000e150 + 0x7*i0; } + +static inline uint32_t REG_A5XX_RB_MRT_CONTROL(uint32_t i0) { return 0x0000e150 + 0x7*i0; } +#define A5XX_RB_MRT_CONTROL_BLEND 0x00000001 +#define A5XX_RB_MRT_CONTROL_BLEND2 0x00000002 +#define A5XX_RB_MRT_CONTROL_ROP_ENABLE 0x00000004 +#define A5XX_RB_MRT_CONTROL_ROP_CODE__MASK 0x00000078 +#define A5XX_RB_MRT_CONTROL_ROP_CODE__SHIFT 3 +static inline uint32_t A5XX_RB_MRT_CONTROL_ROP_CODE(enum a3xx_rop_code val) +{ + return ((val) << A5XX_RB_MRT_CONTROL_ROP_CODE__SHIFT) & A5XX_RB_MRT_CONTROL_ROP_CODE__MASK; +} +#define A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK 0x00000780 +#define A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE__SHIFT 7 +static inline uint32_t A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE(uint32_t val) +{ + return ((val) << A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE__SHIFT) & A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK; +} + +static inline uint32_t REG_A5XX_RB_MRT_BLEND_CONTROL(uint32_t i0) { return 0x0000e151 + 0x7*i0; } +#define A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__MASK 0x0000001f +#define A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__SHIFT 0 +static inline uint32_t A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__SHIFT) & A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__MASK; +} +#define A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__MASK 0x000000e0 +#define A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__SHIFT 5 +static inline uint32_t A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE(enum a3xx_rb_blend_opcode val) +{ + return ((val) << A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__SHIFT) & A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__MASK; +} +#define A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__MASK 0x00001f00 +#define A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__SHIFT 8 +static inline uint32_t A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__SHIFT) & A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__MASK; +} +#define A5XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__MASK 0x001f0000 +#define A5XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__SHIFT 16 +static inline uint32_t A5XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A5XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__SHIFT) & A5XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__MASK; +} +#define A5XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__MASK 0x00e00000 +#define A5XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__SHIFT 21 +static inline uint32_t A5XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE(enum a3xx_rb_blend_opcode val) +{ + return ((val) << A5XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__SHIFT) & A5XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__MASK; +} +#define A5XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__MASK 0x1f000000 +#define A5XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__SHIFT 24 +static inline uint32_t A5XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR(enum adreno_rb_blend_factor val) +{ + return ((val) << A5XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__SHIFT) & A5XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__MASK; +} + +static inline uint32_t REG_A5XX_RB_MRT_BUF_INFO(uint32_t i0) { return 0x0000e152 + 0x7*i0; } +#define A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT__MASK 0x000000ff +#define A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT(enum a5xx_color_fmt val) +{ + return ((val) << A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT__SHIFT) & A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT__MASK; +} +#define A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE__MASK 0x00000300 +#define A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE__SHIFT 8 +static inline uint32_t A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE(enum a5xx_tile_mode val) +{ + return ((val) << A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE__SHIFT) & A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE__MASK; +} +#define A5XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK 0x00006000 +#define A5XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT 13 +static inline uint32_t A5XX_RB_MRT_BUF_INFO_COLOR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT) & A5XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK; +} +#define A5XX_RB_MRT_BUF_INFO_COLOR_SRGB 0x00008000 + +static inline uint32_t REG_A5XX_RB_MRT_PITCH(uint32_t i0) { return 0x0000e153 + 0x7*i0; } +#define A5XX_RB_MRT_PITCH__MASK 0xffffffff +#define A5XX_RB_MRT_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_MRT_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_MRT_PITCH__SHIFT) & A5XX_RB_MRT_PITCH__MASK; +} + +static inline uint32_t REG_A5XX_RB_MRT_ARRAY_PITCH(uint32_t i0) { return 0x0000e154 + 0x7*i0; } +#define A5XX_RB_MRT_ARRAY_PITCH__MASK 0xffffffff +#define A5XX_RB_MRT_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_MRT_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_MRT_ARRAY_PITCH__SHIFT) & A5XX_RB_MRT_ARRAY_PITCH__MASK; +} + +static inline uint32_t REG_A5XX_RB_MRT_BASE_LO(uint32_t i0) { return 0x0000e155 + 0x7*i0; } + +static inline uint32_t REG_A5XX_RB_MRT_BASE_HI(uint32_t i0) { return 0x0000e156 + 0x7*i0; } + +#define REG_A5XX_RB_BLEND_RED 0x0000e1a0 +#define A5XX_RB_BLEND_RED_UINT__MASK 0x000000ff +#define A5XX_RB_BLEND_RED_UINT__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_RED_UINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_RED_UINT__SHIFT) & A5XX_RB_BLEND_RED_UINT__MASK; +} +#define A5XX_RB_BLEND_RED_SINT__MASK 0x0000ff00 +#define A5XX_RB_BLEND_RED_SINT__SHIFT 8 +static inline uint32_t A5XX_RB_BLEND_RED_SINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_RED_SINT__SHIFT) & A5XX_RB_BLEND_RED_SINT__MASK; +} +#define A5XX_RB_BLEND_RED_FLOAT__MASK 0xffff0000 +#define A5XX_RB_BLEND_RED_FLOAT__SHIFT 16 +static inline uint32_t A5XX_RB_BLEND_RED_FLOAT(float val) +{ + return ((util_float_to_half(val)) << A5XX_RB_BLEND_RED_FLOAT__SHIFT) & A5XX_RB_BLEND_RED_FLOAT__MASK; +} + +#define REG_A5XX_RB_BLEND_RED_F32 0x0000e1a1 +#define A5XX_RB_BLEND_RED_F32__MASK 0xffffffff +#define A5XX_RB_BLEND_RED_F32__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_RED_F32(float val) +{ + return ((fui(val)) << A5XX_RB_BLEND_RED_F32__SHIFT) & A5XX_RB_BLEND_RED_F32__MASK; +} + +#define REG_A5XX_RB_BLEND_GREEN 0x0000e1a2 +#define A5XX_RB_BLEND_GREEN_UINT__MASK 0x000000ff +#define A5XX_RB_BLEND_GREEN_UINT__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_GREEN_UINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_GREEN_UINT__SHIFT) & A5XX_RB_BLEND_GREEN_UINT__MASK; +} +#define A5XX_RB_BLEND_GREEN_SINT__MASK 0x0000ff00 +#define A5XX_RB_BLEND_GREEN_SINT__SHIFT 8 +static inline uint32_t A5XX_RB_BLEND_GREEN_SINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_GREEN_SINT__SHIFT) & A5XX_RB_BLEND_GREEN_SINT__MASK; +} +#define A5XX_RB_BLEND_GREEN_FLOAT__MASK 0xffff0000 +#define A5XX_RB_BLEND_GREEN_FLOAT__SHIFT 16 +static inline uint32_t A5XX_RB_BLEND_GREEN_FLOAT(float val) +{ + return ((util_float_to_half(val)) << A5XX_RB_BLEND_GREEN_FLOAT__SHIFT) & A5XX_RB_BLEND_GREEN_FLOAT__MASK; +} + +#define REG_A5XX_RB_BLEND_GREEN_F32 0x0000e1a3 +#define A5XX_RB_BLEND_GREEN_F32__MASK 0xffffffff +#define A5XX_RB_BLEND_GREEN_F32__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_GREEN_F32(float val) +{ + return ((fui(val)) << A5XX_RB_BLEND_GREEN_F32__SHIFT) & A5XX_RB_BLEND_GREEN_F32__MASK; +} + +#define REG_A5XX_RB_BLEND_BLUE 0x0000e1a4 +#define A5XX_RB_BLEND_BLUE_UINT__MASK 0x000000ff +#define A5XX_RB_BLEND_BLUE_UINT__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_BLUE_UINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_BLUE_UINT__SHIFT) & A5XX_RB_BLEND_BLUE_UINT__MASK; +} +#define A5XX_RB_BLEND_BLUE_SINT__MASK 0x0000ff00 +#define A5XX_RB_BLEND_BLUE_SINT__SHIFT 8 +static inline uint32_t A5XX_RB_BLEND_BLUE_SINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_BLUE_SINT__SHIFT) & A5XX_RB_BLEND_BLUE_SINT__MASK; +} +#define A5XX_RB_BLEND_BLUE_FLOAT__MASK 0xffff0000 +#define A5XX_RB_BLEND_BLUE_FLOAT__SHIFT 16 +static inline uint32_t A5XX_RB_BLEND_BLUE_FLOAT(float val) +{ + return ((util_float_to_half(val)) << A5XX_RB_BLEND_BLUE_FLOAT__SHIFT) & A5XX_RB_BLEND_BLUE_FLOAT__MASK; +} + +#define REG_A5XX_RB_BLEND_BLUE_F32 0x0000e1a5 +#define A5XX_RB_BLEND_BLUE_F32__MASK 0xffffffff +#define A5XX_RB_BLEND_BLUE_F32__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_BLUE_F32(float val) +{ + return ((fui(val)) << A5XX_RB_BLEND_BLUE_F32__SHIFT) & A5XX_RB_BLEND_BLUE_F32__MASK; +} + +#define REG_A5XX_RB_BLEND_ALPHA 0x0000e1a6 +#define A5XX_RB_BLEND_ALPHA_UINT__MASK 0x000000ff +#define A5XX_RB_BLEND_ALPHA_UINT__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_ALPHA_UINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_ALPHA_UINT__SHIFT) & A5XX_RB_BLEND_ALPHA_UINT__MASK; +} +#define A5XX_RB_BLEND_ALPHA_SINT__MASK 0x0000ff00 +#define A5XX_RB_BLEND_ALPHA_SINT__SHIFT 8 +static inline uint32_t A5XX_RB_BLEND_ALPHA_SINT(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_ALPHA_SINT__SHIFT) & A5XX_RB_BLEND_ALPHA_SINT__MASK; +} +#define A5XX_RB_BLEND_ALPHA_FLOAT__MASK 0xffff0000 +#define A5XX_RB_BLEND_ALPHA_FLOAT__SHIFT 16 +static inline uint32_t A5XX_RB_BLEND_ALPHA_FLOAT(float val) +{ + return ((util_float_to_half(val)) << A5XX_RB_BLEND_ALPHA_FLOAT__SHIFT) & A5XX_RB_BLEND_ALPHA_FLOAT__MASK; +} + +#define REG_A5XX_RB_BLEND_ALPHA_F32 0x0000e1a7 +#define A5XX_RB_BLEND_ALPHA_F32__MASK 0xffffffff +#define A5XX_RB_BLEND_ALPHA_F32__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_ALPHA_F32(float val) +{ + return ((fui(val)) << A5XX_RB_BLEND_ALPHA_F32__SHIFT) & A5XX_RB_BLEND_ALPHA_F32__MASK; +} + +#define REG_A5XX_RB_ALPHA_CONTROL 0x0000e1a8 +#define A5XX_RB_ALPHA_CONTROL_ALPHA_REF__MASK 0x000000ff +#define A5XX_RB_ALPHA_CONTROL_ALPHA_REF__SHIFT 0 +static inline uint32_t A5XX_RB_ALPHA_CONTROL_ALPHA_REF(uint32_t val) +{ + return ((val) << A5XX_RB_ALPHA_CONTROL_ALPHA_REF__SHIFT) & A5XX_RB_ALPHA_CONTROL_ALPHA_REF__MASK; +} +#define A5XX_RB_ALPHA_CONTROL_ALPHA_TEST 0x00000100 +#define A5XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__MASK 0x00000e00 +#define A5XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__SHIFT 9 +static inline uint32_t A5XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC(enum adreno_compare_func val) +{ + return ((val) << A5XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__SHIFT) & A5XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__MASK; +} + +#define REG_A5XX_RB_BLEND_CNTL 0x0000e1a9 +#define A5XX_RB_BLEND_CNTL_ENABLE_BLEND__MASK 0x000000ff +#define A5XX_RB_BLEND_CNTL_ENABLE_BLEND__SHIFT 0 +static inline uint32_t A5XX_RB_BLEND_CNTL_ENABLE_BLEND(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_CNTL_ENABLE_BLEND__SHIFT) & A5XX_RB_BLEND_CNTL_ENABLE_BLEND__MASK; +} +#define A5XX_RB_BLEND_CNTL_INDEPENDENT_BLEND 0x00000100 +#define A5XX_RB_BLEND_CNTL_SAMPLE_MASK__MASK 0xffff0000 +#define A5XX_RB_BLEND_CNTL_SAMPLE_MASK__SHIFT 16 +static inline uint32_t A5XX_RB_BLEND_CNTL_SAMPLE_MASK(uint32_t val) +{ + return ((val) << A5XX_RB_BLEND_CNTL_SAMPLE_MASK__SHIFT) & A5XX_RB_BLEND_CNTL_SAMPLE_MASK__MASK; +} + +#define REG_A5XX_RB_DEPTH_PLANE_CNTL 0x0000e1b0 +#define A5XX_RB_DEPTH_PLANE_CNTL_FRAG_WRITES_Z 0x00000001 +#define A5XX_RB_DEPTH_PLANE_CNTL_UNK1 0x00000002 + +#define REG_A5XX_RB_DEPTH_CNTL 0x0000e1b1 +#define A5XX_RB_DEPTH_CNTL_Z_ENABLE 0x00000001 +#define A5XX_RB_DEPTH_CNTL_Z_WRITE_ENABLE 0x00000002 +#define A5XX_RB_DEPTH_CNTL_ZFUNC__MASK 0x0000001c +#define A5XX_RB_DEPTH_CNTL_ZFUNC__SHIFT 2 +static inline uint32_t A5XX_RB_DEPTH_CNTL_ZFUNC(enum adreno_compare_func val) +{ + return ((val) << A5XX_RB_DEPTH_CNTL_ZFUNC__SHIFT) & A5XX_RB_DEPTH_CNTL_ZFUNC__MASK; +} +#define A5XX_RB_DEPTH_CNTL_Z_TEST_ENABLE 0x00000040 + +#define REG_A5XX_RB_DEPTH_BUFFER_INFO 0x0000e1b2 +#define A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT__MASK 0x00000007 +#define A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT__SHIFT 0 +static inline uint32_t A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT(enum a5xx_depth_format val) +{ + return ((val) << A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT__SHIFT) & A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT__MASK; +} + +#define REG_A5XX_RB_DEPTH_BUFFER_BASE_LO 0x0000e1b3 + +#define REG_A5XX_RB_DEPTH_BUFFER_BASE_HI 0x0000e1b4 + +#define REG_A5XX_RB_DEPTH_BUFFER_PITCH 0x0000e1b5 +#define A5XX_RB_DEPTH_BUFFER_PITCH__MASK 0xffffffff +#define A5XX_RB_DEPTH_BUFFER_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_DEPTH_BUFFER_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_DEPTH_BUFFER_PITCH__SHIFT) & A5XX_RB_DEPTH_BUFFER_PITCH__MASK; +} + +#define REG_A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH 0x0000e1b6 +#define A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH__MASK 0xffffffff +#define A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH__SHIFT) & A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_STENCIL_CONTROL 0x0000e1c0 +#define A5XX_RB_STENCIL_CONTROL_STENCIL_ENABLE 0x00000001 +#define A5XX_RB_STENCIL_CONTROL_STENCIL_ENABLE_BF 0x00000002 +#define A5XX_RB_STENCIL_CONTROL_STENCIL_READ 0x00000004 +#define A5XX_RB_STENCIL_CONTROL_FUNC__MASK 0x00000700 +#define A5XX_RB_STENCIL_CONTROL_FUNC__SHIFT 8 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_FUNC(enum adreno_compare_func val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_FUNC__SHIFT) & A5XX_RB_STENCIL_CONTROL_FUNC__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_FAIL__MASK 0x00003800 +#define A5XX_RB_STENCIL_CONTROL_FAIL__SHIFT 11 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_FAIL(enum adreno_stencil_op val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_FAIL__SHIFT) & A5XX_RB_STENCIL_CONTROL_FAIL__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_ZPASS__MASK 0x0001c000 +#define A5XX_RB_STENCIL_CONTROL_ZPASS__SHIFT 14 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_ZPASS(enum adreno_stencil_op val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_ZPASS__SHIFT) & A5XX_RB_STENCIL_CONTROL_ZPASS__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_ZFAIL__MASK 0x000e0000 +#define A5XX_RB_STENCIL_CONTROL_ZFAIL__SHIFT 17 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_ZFAIL(enum adreno_stencil_op val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_ZFAIL__SHIFT) & A5XX_RB_STENCIL_CONTROL_ZFAIL__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_FUNC_BF__MASK 0x00700000 +#define A5XX_RB_STENCIL_CONTROL_FUNC_BF__SHIFT 20 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_FUNC_BF(enum adreno_compare_func val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_FUNC_BF__SHIFT) & A5XX_RB_STENCIL_CONTROL_FUNC_BF__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_FAIL_BF__MASK 0x03800000 +#define A5XX_RB_STENCIL_CONTROL_FAIL_BF__SHIFT 23 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_FAIL_BF(enum adreno_stencil_op val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_FAIL_BF__SHIFT) & A5XX_RB_STENCIL_CONTROL_FAIL_BF__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_ZPASS_BF__MASK 0x1c000000 +#define A5XX_RB_STENCIL_CONTROL_ZPASS_BF__SHIFT 26 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_ZPASS_BF(enum adreno_stencil_op val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_ZPASS_BF__SHIFT) & A5XX_RB_STENCIL_CONTROL_ZPASS_BF__MASK; +} +#define A5XX_RB_STENCIL_CONTROL_ZFAIL_BF__MASK 0xe0000000 +#define A5XX_RB_STENCIL_CONTROL_ZFAIL_BF__SHIFT 29 +static inline uint32_t A5XX_RB_STENCIL_CONTROL_ZFAIL_BF(enum adreno_stencil_op val) +{ + return ((val) << A5XX_RB_STENCIL_CONTROL_ZFAIL_BF__SHIFT) & A5XX_RB_STENCIL_CONTROL_ZFAIL_BF__MASK; +} + +#define REG_A5XX_RB_STENCIL_INFO 0x0000e1c1 +#define A5XX_RB_STENCIL_INFO_SEPARATE_STENCIL 0x00000001 + +#define REG_A5XX_RB_STENCIL_BASE_LO 0x0000e1c2 + +#define REG_A5XX_RB_STENCIL_BASE_HI 0x0000e1c3 + +#define REG_A5XX_RB_STENCIL_PITCH 0x0000e1c4 +#define A5XX_RB_STENCIL_PITCH__MASK 0xffffffff +#define A5XX_RB_STENCIL_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_STENCIL_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_STENCIL_PITCH__SHIFT) & A5XX_RB_STENCIL_PITCH__MASK; +} + +#define REG_A5XX_RB_STENCIL_ARRAY_PITCH 0x0000e1c5 +#define A5XX_RB_STENCIL_ARRAY_PITCH__MASK 0xffffffff +#define A5XX_RB_STENCIL_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_STENCIL_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_STENCIL_ARRAY_PITCH__SHIFT) & A5XX_RB_STENCIL_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_STENCILREFMASK 0x0000e1c6 +#define A5XX_RB_STENCILREFMASK_STENCILREF__MASK 0x000000ff +#define A5XX_RB_STENCILREFMASK_STENCILREF__SHIFT 0 +static inline uint32_t A5XX_RB_STENCILREFMASK_STENCILREF(uint32_t val) +{ + return ((val) << A5XX_RB_STENCILREFMASK_STENCILREF__SHIFT) & A5XX_RB_STENCILREFMASK_STENCILREF__MASK; +} +#define A5XX_RB_STENCILREFMASK_STENCILMASK__MASK 0x0000ff00 +#define A5XX_RB_STENCILREFMASK_STENCILMASK__SHIFT 8 +static inline uint32_t A5XX_RB_STENCILREFMASK_STENCILMASK(uint32_t val) +{ + return ((val) << A5XX_RB_STENCILREFMASK_STENCILMASK__SHIFT) & A5XX_RB_STENCILREFMASK_STENCILMASK__MASK; +} +#define A5XX_RB_STENCILREFMASK_STENCILWRITEMASK__MASK 0x00ff0000 +#define A5XX_RB_STENCILREFMASK_STENCILWRITEMASK__SHIFT 16 +static inline uint32_t A5XX_RB_STENCILREFMASK_STENCILWRITEMASK(uint32_t val) +{ + return ((val) << A5XX_RB_STENCILREFMASK_STENCILWRITEMASK__SHIFT) & A5XX_RB_STENCILREFMASK_STENCILWRITEMASK__MASK; +} + +#define REG_A5XX_RB_STENCILREFMASK_BF 0x0000e1c7 +#define A5XX_RB_STENCILREFMASK_BF_STENCILREF__MASK 0x000000ff +#define A5XX_RB_STENCILREFMASK_BF_STENCILREF__SHIFT 0 +static inline uint32_t A5XX_RB_STENCILREFMASK_BF_STENCILREF(uint32_t val) +{ + return ((val) << A5XX_RB_STENCILREFMASK_BF_STENCILREF__SHIFT) & A5XX_RB_STENCILREFMASK_BF_STENCILREF__MASK; +} +#define A5XX_RB_STENCILREFMASK_BF_STENCILMASK__MASK 0x0000ff00 +#define A5XX_RB_STENCILREFMASK_BF_STENCILMASK__SHIFT 8 +static inline uint32_t A5XX_RB_STENCILREFMASK_BF_STENCILMASK(uint32_t val) +{ + return ((val) << A5XX_RB_STENCILREFMASK_BF_STENCILMASK__SHIFT) & A5XX_RB_STENCILREFMASK_BF_STENCILMASK__MASK; +} +#define A5XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__MASK 0x00ff0000 +#define A5XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__SHIFT 16 +static inline uint32_t A5XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK(uint32_t val) +{ + return ((val) << A5XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__SHIFT) & A5XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__MASK; +} + +#define REG_A5XX_RB_WINDOW_OFFSET 0x0000e1d0 +#define A5XX_RB_WINDOW_OFFSET_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_RB_WINDOW_OFFSET_X__MASK 0x00007fff +#define A5XX_RB_WINDOW_OFFSET_X__SHIFT 0 +static inline uint32_t A5XX_RB_WINDOW_OFFSET_X(uint32_t val) +{ + return ((val) << A5XX_RB_WINDOW_OFFSET_X__SHIFT) & A5XX_RB_WINDOW_OFFSET_X__MASK; +} +#define A5XX_RB_WINDOW_OFFSET_Y__MASK 0x7fff0000 +#define A5XX_RB_WINDOW_OFFSET_Y__SHIFT 16 +static inline uint32_t A5XX_RB_WINDOW_OFFSET_Y(uint32_t val) +{ + return ((val) << A5XX_RB_WINDOW_OFFSET_Y__SHIFT) & A5XX_RB_WINDOW_OFFSET_Y__MASK; +} + +#define REG_A5XX_RB_SAMPLE_COUNT_CONTROL 0x0000e1d1 +#define A5XX_RB_SAMPLE_COUNT_CONTROL_COPY 0x00000002 + +#define REG_A5XX_RB_BLIT_CNTL 0x0000e210 +#define A5XX_RB_BLIT_CNTL_BUF__MASK 0x0000000f +#define A5XX_RB_BLIT_CNTL_BUF__SHIFT 0 +static inline uint32_t A5XX_RB_BLIT_CNTL_BUF(enum a5xx_blit_buf val) +{ + return ((val) << A5XX_RB_BLIT_CNTL_BUF__SHIFT) & A5XX_RB_BLIT_CNTL_BUF__MASK; +} + +#define REG_A5XX_RB_RESOLVE_CNTL_1 0x0000e211 +#define A5XX_RB_RESOLVE_CNTL_1_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_RB_RESOLVE_CNTL_1_X__MASK 0x00007fff +#define A5XX_RB_RESOLVE_CNTL_1_X__SHIFT 0 +static inline uint32_t A5XX_RB_RESOLVE_CNTL_1_X(uint32_t val) +{ + return ((val) << A5XX_RB_RESOLVE_CNTL_1_X__SHIFT) & A5XX_RB_RESOLVE_CNTL_1_X__MASK; +} +#define A5XX_RB_RESOLVE_CNTL_1_Y__MASK 0x7fff0000 +#define A5XX_RB_RESOLVE_CNTL_1_Y__SHIFT 16 +static inline uint32_t A5XX_RB_RESOLVE_CNTL_1_Y(uint32_t val) +{ + return ((val) << A5XX_RB_RESOLVE_CNTL_1_Y__SHIFT) & A5XX_RB_RESOLVE_CNTL_1_Y__MASK; +} + +#define REG_A5XX_RB_RESOLVE_CNTL_2 0x0000e212 +#define A5XX_RB_RESOLVE_CNTL_2_WINDOW_OFFSET_DISABLE 0x80000000 +#define A5XX_RB_RESOLVE_CNTL_2_X__MASK 0x00007fff +#define A5XX_RB_RESOLVE_CNTL_2_X__SHIFT 0 +static inline uint32_t A5XX_RB_RESOLVE_CNTL_2_X(uint32_t val) +{ + return ((val) << A5XX_RB_RESOLVE_CNTL_2_X__SHIFT) & A5XX_RB_RESOLVE_CNTL_2_X__MASK; +} +#define A5XX_RB_RESOLVE_CNTL_2_Y__MASK 0x7fff0000 +#define A5XX_RB_RESOLVE_CNTL_2_Y__SHIFT 16 +static inline uint32_t A5XX_RB_RESOLVE_CNTL_2_Y(uint32_t val) +{ + return ((val) << A5XX_RB_RESOLVE_CNTL_2_Y__SHIFT) & A5XX_RB_RESOLVE_CNTL_2_Y__MASK; +} + +#define REG_A5XX_RB_RESOLVE_CNTL_3 0x0000e213 + +#define REG_A5XX_RB_BLIT_DST_LO 0x0000e214 + +#define REG_A5XX_RB_BLIT_DST_HI 0x0000e215 + +#define REG_A5XX_RB_BLIT_DST_PITCH 0x0000e216 +#define A5XX_RB_BLIT_DST_PITCH__MASK 0xffffffff +#define A5XX_RB_BLIT_DST_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_BLIT_DST_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_BLIT_DST_PITCH__SHIFT) & A5XX_RB_BLIT_DST_PITCH__MASK; +} + +#define REG_A5XX_RB_BLIT_DST_ARRAY_PITCH 0x0000e217 +#define A5XX_RB_BLIT_DST_ARRAY_PITCH__MASK 0xffffffff +#define A5XX_RB_BLIT_DST_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_BLIT_DST_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_BLIT_DST_ARRAY_PITCH__SHIFT) & A5XX_RB_BLIT_DST_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_CLEAR_COLOR_DW0 0x0000e218 + +#define REG_A5XX_RB_CLEAR_COLOR_DW1 0x0000e219 + +#define REG_A5XX_RB_CLEAR_COLOR_DW2 0x0000e21a + +#define REG_A5XX_RB_CLEAR_COLOR_DW3 0x0000e21b + +#define REG_A5XX_RB_CLEAR_CNTL 0x0000e21c +#define A5XX_RB_CLEAR_CNTL_FAST_CLEAR 0x00000002 +#define A5XX_RB_CLEAR_CNTL_MASK__MASK 0x000000f0 +#define A5XX_RB_CLEAR_CNTL_MASK__SHIFT 4 +static inline uint32_t A5XX_RB_CLEAR_CNTL_MASK(uint32_t val) +{ + return ((val) << A5XX_RB_CLEAR_CNTL_MASK__SHIFT) & A5XX_RB_CLEAR_CNTL_MASK__MASK; +} + +#define REG_A5XX_RB_DEPTH_FLAG_BUFFER_BASE_LO 0x0000e240 + +#define REG_A5XX_RB_DEPTH_FLAG_BUFFER_BASE_HI 0x0000e241 + +#define REG_A5XX_RB_DEPTH_FLAG_BUFFER_PITCH 0x0000e242 + +static inline uint32_t REG_A5XX_RB_MRT_FLAG_BUFFER(uint32_t i0) { return 0x0000e243 + 0x4*i0; } + +static inline uint32_t REG_A5XX_RB_MRT_FLAG_BUFFER_ADDR_LO(uint32_t i0) { return 0x0000e243 + 0x4*i0; } + +static inline uint32_t REG_A5XX_RB_MRT_FLAG_BUFFER_ADDR_HI(uint32_t i0) { return 0x0000e244 + 0x4*i0; } + +static inline uint32_t REG_A5XX_RB_MRT_FLAG_BUFFER_PITCH(uint32_t i0) { return 0x0000e245 + 0x4*i0; } +#define A5XX_RB_MRT_FLAG_BUFFER_PITCH__MASK 0xffffffff +#define A5XX_RB_MRT_FLAG_BUFFER_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_MRT_FLAG_BUFFER_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_MRT_FLAG_BUFFER_PITCH__SHIFT) & A5XX_RB_MRT_FLAG_BUFFER_PITCH__MASK; +} + +static inline uint32_t REG_A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH(uint32_t i0) { return 0x0000e246 + 0x4*i0; } +#define A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH__MASK 0xffffffff +#define A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH__SHIFT) & A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_BLIT_FLAG_DST_LO 0x0000e263 + +#define REG_A5XX_RB_BLIT_FLAG_DST_HI 0x0000e264 + +#define REG_A5XX_RB_BLIT_FLAG_DST_PITCH 0x0000e265 +#define A5XX_RB_BLIT_FLAG_DST_PITCH__MASK 0xffffffff +#define A5XX_RB_BLIT_FLAG_DST_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_BLIT_FLAG_DST_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_BLIT_FLAG_DST_PITCH__SHIFT) & A5XX_RB_BLIT_FLAG_DST_PITCH__MASK; +} + +#define REG_A5XX_RB_BLIT_FLAG_DST_ARRAY_PITCH 0x0000e266 +#define A5XX_RB_BLIT_FLAG_DST_ARRAY_PITCH__MASK 0xffffffff +#define A5XX_RB_BLIT_FLAG_DST_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_BLIT_FLAG_DST_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_BLIT_FLAG_DST_ARRAY_PITCH__SHIFT) & A5XX_RB_BLIT_FLAG_DST_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_SAMPLE_COUNT_ADDR_LO 0x0000e267 + +#define REG_A5XX_RB_SAMPLE_COUNT_ADDR_HI 0x0000e268 + +#define REG_A5XX_VPC_CNTL_0 0x0000e280 +#define A5XX_VPC_CNTL_0_STRIDE_IN_VPC__MASK 0x0000007f +#define A5XX_VPC_CNTL_0_STRIDE_IN_VPC__SHIFT 0 +static inline uint32_t A5XX_VPC_CNTL_0_STRIDE_IN_VPC(uint32_t val) +{ + return ((val) << A5XX_VPC_CNTL_0_STRIDE_IN_VPC__SHIFT) & A5XX_VPC_CNTL_0_STRIDE_IN_VPC__MASK; +} +#define A5XX_VPC_CNTL_0_VARYING 0x00000800 + +static inline uint32_t REG_A5XX_VPC_VARYING_INTERP(uint32_t i0) { return 0x0000e282 + 0x1*i0; } + +static inline uint32_t REG_A5XX_VPC_VARYING_INTERP_MODE(uint32_t i0) { return 0x0000e282 + 0x1*i0; } + +static inline uint32_t REG_A5XX_VPC_VARYING_PS_REPL(uint32_t i0) { return 0x0000e28a + 0x1*i0; } + +static inline uint32_t REG_A5XX_VPC_VARYING_PS_REPL_MODE(uint32_t i0) { return 0x0000e28a + 0x1*i0; } + +#define REG_A5XX_UNKNOWN_E292 0x0000e292 + +#define REG_A5XX_UNKNOWN_E293 0x0000e293 + +static inline uint32_t REG_A5XX_VPC_VAR(uint32_t i0) { return 0x0000e294 + 0x1*i0; } + +static inline uint32_t REG_A5XX_VPC_VAR_DISABLE(uint32_t i0) { return 0x0000e294 + 0x1*i0; } + +#define REG_A5XX_VPC_GS_SIV_CNTL 0x0000e298 + +#define REG_A5XX_UNKNOWN_E29A 0x0000e29a + +#define REG_A5XX_VPC_PACK 0x0000e29d +#define A5XX_VPC_PACK_NUMNONPOSVAR__MASK 0x000000ff +#define A5XX_VPC_PACK_NUMNONPOSVAR__SHIFT 0 +static inline uint32_t A5XX_VPC_PACK_NUMNONPOSVAR(uint32_t val) +{ + return ((val) << A5XX_VPC_PACK_NUMNONPOSVAR__SHIFT) & A5XX_VPC_PACK_NUMNONPOSVAR__MASK; +} +#define A5XX_VPC_PACK_PSIZELOC__MASK 0x0000ff00 +#define A5XX_VPC_PACK_PSIZELOC__SHIFT 8 +static inline uint32_t A5XX_VPC_PACK_PSIZELOC(uint32_t val) +{ + return ((val) << A5XX_VPC_PACK_PSIZELOC__SHIFT) & A5XX_VPC_PACK_PSIZELOC__MASK; +} + +#define REG_A5XX_VPC_FS_PRIMITIVEID_CNTL 0x0000e2a0 + +#define REG_A5XX_VPC_SO_BUF_CNTL 0x0000e2a1 +#define A5XX_VPC_SO_BUF_CNTL_BUF0 0x00000001 +#define A5XX_VPC_SO_BUF_CNTL_BUF1 0x00000008 +#define A5XX_VPC_SO_BUF_CNTL_BUF2 0x00000040 +#define A5XX_VPC_SO_BUF_CNTL_BUF3 0x00000200 +#define A5XX_VPC_SO_BUF_CNTL_ENABLE 0x00008000 + +#define REG_A5XX_VPC_SO_OVERRIDE 0x0000e2a2 +#define A5XX_VPC_SO_OVERRIDE_SO_DISABLE 0x00000001 + +#define REG_A5XX_VPC_SO_CNTL 0x0000e2a3 +#define A5XX_VPC_SO_CNTL_ENABLE 0x00010000 + +#define REG_A5XX_VPC_SO_PROG 0x0000e2a4 +#define A5XX_VPC_SO_PROG_A_BUF__MASK 0x00000003 +#define A5XX_VPC_SO_PROG_A_BUF__SHIFT 0 +static inline uint32_t A5XX_VPC_SO_PROG_A_BUF(uint32_t val) +{ + return ((val) << A5XX_VPC_SO_PROG_A_BUF__SHIFT) & A5XX_VPC_SO_PROG_A_BUF__MASK; +} +#define A5XX_VPC_SO_PROG_A_OFF__MASK 0x000007fc +#define A5XX_VPC_SO_PROG_A_OFF__SHIFT 2 +static inline uint32_t A5XX_VPC_SO_PROG_A_OFF(uint32_t val) +{ + assert(!(val & 0x3)); + return ((val >> 2) << A5XX_VPC_SO_PROG_A_OFF__SHIFT) & A5XX_VPC_SO_PROG_A_OFF__MASK; +} +#define A5XX_VPC_SO_PROG_A_EN 0x00000800 +#define A5XX_VPC_SO_PROG_B_BUF__MASK 0x00003000 +#define A5XX_VPC_SO_PROG_B_BUF__SHIFT 12 +static inline uint32_t A5XX_VPC_SO_PROG_B_BUF(uint32_t val) +{ + return ((val) << A5XX_VPC_SO_PROG_B_BUF__SHIFT) & A5XX_VPC_SO_PROG_B_BUF__MASK; +} +#define A5XX_VPC_SO_PROG_B_OFF__MASK 0x007fc000 +#define A5XX_VPC_SO_PROG_B_OFF__SHIFT 14 +static inline uint32_t A5XX_VPC_SO_PROG_B_OFF(uint32_t val) +{ + assert(!(val & 0x3)); + return ((val >> 2) << A5XX_VPC_SO_PROG_B_OFF__SHIFT) & A5XX_VPC_SO_PROG_B_OFF__MASK; +} +#define A5XX_VPC_SO_PROG_B_EN 0x00800000 + +static inline uint32_t REG_A5XX_VPC_SO(uint32_t i0) { return 0x0000e2a7 + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_BUFFER_BASE_LO(uint32_t i0) { return 0x0000e2a7 + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_BUFFER_BASE_HI(uint32_t i0) { return 0x0000e2a8 + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_BUFFER_SIZE(uint32_t i0) { return 0x0000e2a9 + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_NCOMP(uint32_t i0) { return 0x0000e2aa + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_BUFFER_OFFSET(uint32_t i0) { return 0x0000e2ab + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_FLUSH_BASE_LO(uint32_t i0) { return 0x0000e2ac + 0x7*i0; } + +static inline uint32_t REG_A5XX_VPC_SO_FLUSH_BASE_HI(uint32_t i0) { return 0x0000e2ad + 0x7*i0; } + +#define REG_A5XX_PC_PRIMITIVE_CNTL 0x0000e384 +#define A5XX_PC_PRIMITIVE_CNTL_STRIDE_IN_VPC__MASK 0x0000007f +#define A5XX_PC_PRIMITIVE_CNTL_STRIDE_IN_VPC__SHIFT 0 +static inline uint32_t A5XX_PC_PRIMITIVE_CNTL_STRIDE_IN_VPC(uint32_t val) +{ + return ((val) << A5XX_PC_PRIMITIVE_CNTL_STRIDE_IN_VPC__SHIFT) & A5XX_PC_PRIMITIVE_CNTL_STRIDE_IN_VPC__MASK; +} +#define A5XX_PC_PRIMITIVE_CNTL_PRIMITIVE_RESTART 0x00000100 +#define A5XX_PC_PRIMITIVE_CNTL_PROVOKING_VTX_LAST 0x00000400 + +#define REG_A5XX_PC_PRIM_VTX_CNTL 0x0000e385 +#define A5XX_PC_PRIM_VTX_CNTL_PSIZE 0x00000800 + +#define REG_A5XX_PC_RASTER_CNTL 0x0000e388 +#define A5XX_PC_RASTER_CNTL_POLYMODE_FRONT_PTYPE__MASK 0x00000007 +#define A5XX_PC_RASTER_CNTL_POLYMODE_FRONT_PTYPE__SHIFT 0 +static inline uint32_t A5XX_PC_RASTER_CNTL_POLYMODE_FRONT_PTYPE(enum adreno_pa_su_sc_draw val) +{ + return ((val) << A5XX_PC_RASTER_CNTL_POLYMODE_FRONT_PTYPE__SHIFT) & A5XX_PC_RASTER_CNTL_POLYMODE_FRONT_PTYPE__MASK; +} +#define A5XX_PC_RASTER_CNTL_POLYMODE_BACK_PTYPE__MASK 0x00000038 +#define A5XX_PC_RASTER_CNTL_POLYMODE_BACK_PTYPE__SHIFT 3 +static inline uint32_t A5XX_PC_RASTER_CNTL_POLYMODE_BACK_PTYPE(enum adreno_pa_su_sc_draw val) +{ + return ((val) << A5XX_PC_RASTER_CNTL_POLYMODE_BACK_PTYPE__SHIFT) & A5XX_PC_RASTER_CNTL_POLYMODE_BACK_PTYPE__MASK; +} +#define A5XX_PC_RASTER_CNTL_POLYMODE_ENABLE 0x00000040 + +#define REG_A5XX_UNKNOWN_E389 0x0000e389 + +#define REG_A5XX_PC_RESTART_INDEX 0x0000e38c + +#define REG_A5XX_UNKNOWN_E38D 0x0000e38d + +#define REG_A5XX_PC_GS_PARAM 0x0000e38e + +#define REG_A5XX_PC_HS_PARAM 0x0000e38f + +#define REG_A5XX_PC_POWER_CNTL 0x0000e3b0 + +#define REG_A5XX_VFD_CONTROL_0 0x0000e400 +#define A5XX_VFD_CONTROL_0_VTXCNT__MASK 0x0000003f +#define A5XX_VFD_CONTROL_0_VTXCNT__SHIFT 0 +static inline uint32_t A5XX_VFD_CONTROL_0_VTXCNT(uint32_t val) +{ + return ((val) << A5XX_VFD_CONTROL_0_VTXCNT__SHIFT) & A5XX_VFD_CONTROL_0_VTXCNT__MASK; +} + +#define REG_A5XX_VFD_CONTROL_1 0x0000e401 +#define A5XX_VFD_CONTROL_1_REGID4VTX__MASK 0x000000ff +#define A5XX_VFD_CONTROL_1_REGID4VTX__SHIFT 0 +static inline uint32_t A5XX_VFD_CONTROL_1_REGID4VTX(uint32_t val) +{ + return ((val) << A5XX_VFD_CONTROL_1_REGID4VTX__SHIFT) & A5XX_VFD_CONTROL_1_REGID4VTX__MASK; +} +#define A5XX_VFD_CONTROL_1_REGID4INST__MASK 0x0000ff00 +#define A5XX_VFD_CONTROL_1_REGID4INST__SHIFT 8 +static inline uint32_t A5XX_VFD_CONTROL_1_REGID4INST(uint32_t val) +{ + return ((val) << A5XX_VFD_CONTROL_1_REGID4INST__SHIFT) & A5XX_VFD_CONTROL_1_REGID4INST__MASK; +} + +#define REG_A5XX_VFD_CONTROL_2 0x0000e402 + +#define REG_A5XX_VFD_CONTROL_3 0x0000e403 + +#define REG_A5XX_VFD_CONTROL_4 0x0000e404 + +#define REG_A5XX_VFD_CONTROL_5 0x0000e405 + +#define REG_A5XX_VFD_INDEX_OFFSET 0x0000e408 + +#define REG_A5XX_VFD_INSTANCE_START_OFFSET 0x0000e409 + +static inline uint32_t REG_A5XX_VFD_FETCH(uint32_t i0) { return 0x0000e40a + 0x4*i0; } + +static inline uint32_t REG_A5XX_VFD_FETCH_BASE_LO(uint32_t i0) { return 0x0000e40a + 0x4*i0; } + +static inline uint32_t REG_A5XX_VFD_FETCH_BASE_HI(uint32_t i0) { return 0x0000e40b + 0x4*i0; } + +static inline uint32_t REG_A5XX_VFD_FETCH_SIZE(uint32_t i0) { return 0x0000e40c + 0x4*i0; } + +static inline uint32_t REG_A5XX_VFD_FETCH_STRIDE(uint32_t i0) { return 0x0000e40d + 0x4*i0; } + +static inline uint32_t REG_A5XX_VFD_DECODE(uint32_t i0) { return 0x0000e48a + 0x2*i0; } + +static inline uint32_t REG_A5XX_VFD_DECODE_INSTR(uint32_t i0) { return 0x0000e48a + 0x2*i0; } +#define A5XX_VFD_DECODE_INSTR_IDX__MASK 0x0000001f +#define A5XX_VFD_DECODE_INSTR_IDX__SHIFT 0 +static inline uint32_t A5XX_VFD_DECODE_INSTR_IDX(uint32_t val) +{ + return ((val) << A5XX_VFD_DECODE_INSTR_IDX__SHIFT) & A5XX_VFD_DECODE_INSTR_IDX__MASK; +} +#define A5XX_VFD_DECODE_INSTR_INSTANCED 0x00020000 +#define A5XX_VFD_DECODE_INSTR_FORMAT__MASK 0x0ff00000 +#define A5XX_VFD_DECODE_INSTR_FORMAT__SHIFT 20 +static inline uint32_t A5XX_VFD_DECODE_INSTR_FORMAT(enum a5xx_vtx_fmt val) +{ + return ((val) << A5XX_VFD_DECODE_INSTR_FORMAT__SHIFT) & A5XX_VFD_DECODE_INSTR_FORMAT__MASK; +} +#define A5XX_VFD_DECODE_INSTR_SWAP__MASK 0x30000000 +#define A5XX_VFD_DECODE_INSTR_SWAP__SHIFT 28 +static inline uint32_t A5XX_VFD_DECODE_INSTR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_VFD_DECODE_INSTR_SWAP__SHIFT) & A5XX_VFD_DECODE_INSTR_SWAP__MASK; +} +#define A5XX_VFD_DECODE_INSTR_UNK30 0x40000000 +#define A5XX_VFD_DECODE_INSTR_FLOAT 0x80000000 + +static inline uint32_t REG_A5XX_VFD_DECODE_STEP_RATE(uint32_t i0) { return 0x0000e48b + 0x2*i0; } + +static inline uint32_t REG_A5XX_VFD_DEST_CNTL(uint32_t i0) { return 0x0000e4ca + 0x1*i0; } + +static inline uint32_t REG_A5XX_VFD_DEST_CNTL_INSTR(uint32_t i0) { return 0x0000e4ca + 0x1*i0; } +#define A5XX_VFD_DEST_CNTL_INSTR_WRITEMASK__MASK 0x0000000f +#define A5XX_VFD_DEST_CNTL_INSTR_WRITEMASK__SHIFT 0 +static inline uint32_t A5XX_VFD_DEST_CNTL_INSTR_WRITEMASK(uint32_t val) +{ + return ((val) << A5XX_VFD_DEST_CNTL_INSTR_WRITEMASK__SHIFT) & A5XX_VFD_DEST_CNTL_INSTR_WRITEMASK__MASK; +} +#define A5XX_VFD_DEST_CNTL_INSTR_REGID__MASK 0x00000ff0 +#define A5XX_VFD_DEST_CNTL_INSTR_REGID__SHIFT 4 +static inline uint32_t A5XX_VFD_DEST_CNTL_INSTR_REGID(uint32_t val) +{ + return ((val) << A5XX_VFD_DEST_CNTL_INSTR_REGID__SHIFT) & A5XX_VFD_DEST_CNTL_INSTR_REGID__MASK; +} + +#define REG_A5XX_VFD_POWER_CNTL 0x0000e4f0 + +#define REG_A5XX_SP_SP_CNTL 0x0000e580 + +#define REG_A5XX_SP_VS_CONFIG 0x0000e584 +#define A5XX_SP_VS_CONFIG_ENABLED 0x00000001 +#define A5XX_SP_VS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_SP_VS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_SP_VS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_VS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_SP_VS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_SP_VS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_SP_VS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_SP_VS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_VS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_SP_VS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_SP_FS_CONFIG 0x0000e585 +#define A5XX_SP_FS_CONFIG_ENABLED 0x00000001 +#define A5XX_SP_FS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_SP_FS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_SP_FS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_FS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_SP_FS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_SP_FS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_SP_FS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_SP_FS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_FS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_SP_FS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_SP_HS_CONFIG 0x0000e586 +#define A5XX_SP_HS_CONFIG_ENABLED 0x00000001 +#define A5XX_SP_HS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_SP_HS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_SP_HS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_HS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_SP_HS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_SP_HS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_SP_HS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_SP_HS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_HS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_SP_HS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_SP_DS_CONFIG 0x0000e587 +#define A5XX_SP_DS_CONFIG_ENABLED 0x00000001 +#define A5XX_SP_DS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_SP_DS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_SP_DS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_DS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_SP_DS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_SP_DS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_SP_DS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_SP_DS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_DS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_SP_DS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_SP_GS_CONFIG 0x0000e588 +#define A5XX_SP_GS_CONFIG_ENABLED 0x00000001 +#define A5XX_SP_GS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_SP_GS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_SP_GS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_GS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_SP_GS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_SP_GS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_SP_GS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_SP_GS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_GS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_SP_GS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_SP_CS_CONFIG 0x0000e589 +#define A5XX_SP_CS_CONFIG_ENABLED 0x00000001 +#define A5XX_SP_CS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_SP_CS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_SP_CS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_CS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_SP_CS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_SP_CS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_SP_CS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_SP_CS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_SP_CS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_SP_CS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_SP_VS_CONFIG_MAX_CONST 0x0000e58a + +#define REG_A5XX_SP_FS_CONFIG_MAX_CONST 0x0000e58b + +#define REG_A5XX_SP_VS_CTRL_REG0 0x0000e590 +#define A5XX_SP_VS_CTRL_REG0_THREADSIZE__MASK 0x00000008 +#define A5XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT 3 +static inline uint32_t A5XX_SP_VS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A5XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT) & A5XX_SP_VS_CTRL_REG0_THREADSIZE__MASK; +} +#define A5XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0 +#define A5XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4 +static inline uint32_t A5XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val) +{ + return ((val) << A5XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A5XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK; +} +#define A5XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0000fc00 +#define A5XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10 +static inline uint32_t A5XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val) +{ + return ((val) << A5XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A5XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK; +} +#define A5XX_SP_VS_CTRL_REG0_VARYING 0x00010000 +#define A5XX_SP_VS_CTRL_REG0_PIXLODENABLE 0x00100000 +#define A5XX_SP_VS_CTRL_REG0_BRANCHSTACK__MASK 0xfe000000 +#define A5XX_SP_VS_CTRL_REG0_BRANCHSTACK__SHIFT 25 +static inline uint32_t A5XX_SP_VS_CTRL_REG0_BRANCHSTACK(uint32_t val) +{ + return ((val) << A5XX_SP_VS_CTRL_REG0_BRANCHSTACK__SHIFT) & A5XX_SP_VS_CTRL_REG0_BRANCHSTACK__MASK; +} + +#define REG_A5XX_SP_PRIMITIVE_CNTL 0x0000e592 +#define A5XX_SP_PRIMITIVE_CNTL_VSOUT__MASK 0x0000001f +#define A5XX_SP_PRIMITIVE_CNTL_VSOUT__SHIFT 0 +static inline uint32_t A5XX_SP_PRIMITIVE_CNTL_VSOUT(uint32_t val) +{ + return ((val) << A5XX_SP_PRIMITIVE_CNTL_VSOUT__SHIFT) & A5XX_SP_PRIMITIVE_CNTL_VSOUT__MASK; +} + +static inline uint32_t REG_A5XX_SP_VS_OUT(uint32_t i0) { return 0x0000e593 + 0x1*i0; } + +static inline uint32_t REG_A5XX_SP_VS_OUT_REG(uint32_t i0) { return 0x0000e593 + 0x1*i0; } +#define A5XX_SP_VS_OUT_REG_A_REGID__MASK 0x000000ff +#define A5XX_SP_VS_OUT_REG_A_REGID__SHIFT 0 +static inline uint32_t A5XX_SP_VS_OUT_REG_A_REGID(uint32_t val) +{ + return ((val) << A5XX_SP_VS_OUT_REG_A_REGID__SHIFT) & A5XX_SP_VS_OUT_REG_A_REGID__MASK; +} +#define A5XX_SP_VS_OUT_REG_A_COMPMASK__MASK 0x00000f00 +#define A5XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT 8 +static inline uint32_t A5XX_SP_VS_OUT_REG_A_COMPMASK(uint32_t val) +{ + return ((val) << A5XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT) & A5XX_SP_VS_OUT_REG_A_COMPMASK__MASK; +} +#define A5XX_SP_VS_OUT_REG_B_REGID__MASK 0x00ff0000 +#define A5XX_SP_VS_OUT_REG_B_REGID__SHIFT 16 +static inline uint32_t A5XX_SP_VS_OUT_REG_B_REGID(uint32_t val) +{ + return ((val) << A5XX_SP_VS_OUT_REG_B_REGID__SHIFT) & A5XX_SP_VS_OUT_REG_B_REGID__MASK; +} +#define A5XX_SP_VS_OUT_REG_B_COMPMASK__MASK 0x0f000000 +#define A5XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT 24 +static inline uint32_t A5XX_SP_VS_OUT_REG_B_COMPMASK(uint32_t val) +{ + return ((val) << A5XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT) & A5XX_SP_VS_OUT_REG_B_COMPMASK__MASK; +} + +static inline uint32_t REG_A5XX_SP_VS_VPC_DST(uint32_t i0) { return 0x0000e5a3 + 0x1*i0; } + +static inline uint32_t REG_A5XX_SP_VS_VPC_DST_REG(uint32_t i0) { return 0x0000e5a3 + 0x1*i0; } +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK 0x000000ff +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT 0 +static inline uint32_t A5XX_SP_VS_VPC_DST_REG_OUTLOC0(uint32_t val) +{ + return ((val) << A5XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT) & A5XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK; +} +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK 0x0000ff00 +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT 8 +static inline uint32_t A5XX_SP_VS_VPC_DST_REG_OUTLOC1(uint32_t val) +{ + return ((val) << A5XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT) & A5XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK; +} +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK 0x00ff0000 +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT 16 +static inline uint32_t A5XX_SP_VS_VPC_DST_REG_OUTLOC2(uint32_t val) +{ + return ((val) << A5XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT) & A5XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK; +} +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK 0xff000000 +#define A5XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT 24 +static inline uint32_t A5XX_SP_VS_VPC_DST_REG_OUTLOC3(uint32_t val) +{ + return ((val) << A5XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT) & A5XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK; +} + +#define REG_A5XX_UNKNOWN_E5AB 0x0000e5ab + +#define REG_A5XX_SP_VS_OBJ_START_LO 0x0000e5ac + +#define REG_A5XX_SP_VS_OBJ_START_HI 0x0000e5ad + +#define REG_A5XX_SP_FS_CTRL_REG0 0x0000e5c0 +#define A5XX_SP_FS_CTRL_REG0_THREADSIZE__MASK 0x00000008 +#define A5XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT 3 +static inline uint32_t A5XX_SP_FS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A5XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT) & A5XX_SP_FS_CTRL_REG0_THREADSIZE__MASK; +} +#define A5XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0 +#define A5XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4 +static inline uint32_t A5XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val) +{ + return ((val) << A5XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A5XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK; +} +#define A5XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0000fc00 +#define A5XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10 +static inline uint32_t A5XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val) +{ + return ((val) << A5XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A5XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK; +} +#define A5XX_SP_FS_CTRL_REG0_VARYING 0x00010000 +#define A5XX_SP_FS_CTRL_REG0_PIXLODENABLE 0x00100000 +#define A5XX_SP_FS_CTRL_REG0_BRANCHSTACK__MASK 0xfe000000 +#define A5XX_SP_FS_CTRL_REG0_BRANCHSTACK__SHIFT 25 +static inline uint32_t A5XX_SP_FS_CTRL_REG0_BRANCHSTACK(uint32_t val) +{ + return ((val) << A5XX_SP_FS_CTRL_REG0_BRANCHSTACK__SHIFT) & A5XX_SP_FS_CTRL_REG0_BRANCHSTACK__MASK; +} + +#define REG_A5XX_UNKNOWN_E5C2 0x0000e5c2 + +#define REG_A5XX_SP_FS_OBJ_START_LO 0x0000e5c3 + +#define REG_A5XX_SP_FS_OBJ_START_HI 0x0000e5c4 + +#define REG_A5XX_SP_BLEND_CNTL 0x0000e5c9 +#define A5XX_SP_BLEND_CNTL_ENABLED 0x00000001 +#define A5XX_SP_BLEND_CNTL_UNK8 0x00000100 + +#define REG_A5XX_SP_FS_OUTPUT_CNTL 0x0000e5ca +#define A5XX_SP_FS_OUTPUT_CNTL_MRT__MASK 0x0000000f +#define A5XX_SP_FS_OUTPUT_CNTL_MRT__SHIFT 0 +static inline uint32_t A5XX_SP_FS_OUTPUT_CNTL_MRT(uint32_t val) +{ + return ((val) << A5XX_SP_FS_OUTPUT_CNTL_MRT__SHIFT) & A5XX_SP_FS_OUTPUT_CNTL_MRT__MASK; +} +#define A5XX_SP_FS_OUTPUT_CNTL_DEPTH_REGID__MASK 0x00001fe0 +#define A5XX_SP_FS_OUTPUT_CNTL_DEPTH_REGID__SHIFT 5 +static inline uint32_t A5XX_SP_FS_OUTPUT_CNTL_DEPTH_REGID(uint32_t val) +{ + return ((val) << A5XX_SP_FS_OUTPUT_CNTL_DEPTH_REGID__SHIFT) & A5XX_SP_FS_OUTPUT_CNTL_DEPTH_REGID__MASK; +} +#define A5XX_SP_FS_OUTPUT_CNTL_SAMPLEMASK_REGID__MASK 0x001fe000 +#define A5XX_SP_FS_OUTPUT_CNTL_SAMPLEMASK_REGID__SHIFT 13 +static inline uint32_t A5XX_SP_FS_OUTPUT_CNTL_SAMPLEMASK_REGID(uint32_t val) +{ + return ((val) << A5XX_SP_FS_OUTPUT_CNTL_SAMPLEMASK_REGID__SHIFT) & A5XX_SP_FS_OUTPUT_CNTL_SAMPLEMASK_REGID__MASK; +} + +static inline uint32_t REG_A5XX_SP_FS_OUTPUT(uint32_t i0) { return 0x0000e5cb + 0x1*i0; } + +static inline uint32_t REG_A5XX_SP_FS_OUTPUT_REG(uint32_t i0) { return 0x0000e5cb + 0x1*i0; } +#define A5XX_SP_FS_OUTPUT_REG_REGID__MASK 0x000000ff +#define A5XX_SP_FS_OUTPUT_REG_REGID__SHIFT 0 +static inline uint32_t A5XX_SP_FS_OUTPUT_REG_REGID(uint32_t val) +{ + return ((val) << A5XX_SP_FS_OUTPUT_REG_REGID__SHIFT) & A5XX_SP_FS_OUTPUT_REG_REGID__MASK; +} +#define A5XX_SP_FS_OUTPUT_REG_HALF_PRECISION 0x00000100 + +static inline uint32_t REG_A5XX_SP_FS_MRT(uint32_t i0) { return 0x0000e5d3 + 0x1*i0; } + +static inline uint32_t REG_A5XX_SP_FS_MRT_REG(uint32_t i0) { return 0x0000e5d3 + 0x1*i0; } +#define A5XX_SP_FS_MRT_REG_COLOR_FORMAT__MASK 0x000000ff +#define A5XX_SP_FS_MRT_REG_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A5XX_SP_FS_MRT_REG_COLOR_FORMAT(enum a5xx_color_fmt val) +{ + return ((val) << A5XX_SP_FS_MRT_REG_COLOR_FORMAT__SHIFT) & A5XX_SP_FS_MRT_REG_COLOR_FORMAT__MASK; +} +#define A5XX_SP_FS_MRT_REG_COLOR_SINT 0x00000100 +#define A5XX_SP_FS_MRT_REG_COLOR_UINT 0x00000200 +#define A5XX_SP_FS_MRT_REG_COLOR_SRGB 0x00000400 + +#define REG_A5XX_UNKNOWN_E5DB 0x0000e5db + +#define REG_A5XX_UNKNOWN_E5F2 0x0000e5f2 + +#define REG_A5XX_SP_CS_OBJ_START_LO 0x0000e5f3 + +#define REG_A5XX_SP_CS_OBJ_START_HI 0x0000e5f4 + +#define REG_A5XX_SP_CS_CTRL_REG0 0x0000e5f0 +#define A5XX_SP_CS_CTRL_REG0_THREADSIZE__MASK 0x00000008 +#define A5XX_SP_CS_CTRL_REG0_THREADSIZE__SHIFT 3 +static inline uint32_t A5XX_SP_CS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A5XX_SP_CS_CTRL_REG0_THREADSIZE__SHIFT) & A5XX_SP_CS_CTRL_REG0_THREADSIZE__MASK; +} +#define A5XX_SP_CS_CTRL_REG0_HALFREGFOOTPRINT__MASK 0x000003f0 +#define A5XX_SP_CS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT 4 +static inline uint32_t A5XX_SP_CS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val) +{ + return ((val) << A5XX_SP_CS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A5XX_SP_CS_CTRL_REG0_HALFREGFOOTPRINT__MASK; +} +#define A5XX_SP_CS_CTRL_REG0_FULLREGFOOTPRINT__MASK 0x0000fc00 +#define A5XX_SP_CS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT 10 +static inline uint32_t A5XX_SP_CS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val) +{ + return ((val) << A5XX_SP_CS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A5XX_SP_CS_CTRL_REG0_FULLREGFOOTPRINT__MASK; +} +#define A5XX_SP_CS_CTRL_REG0_VARYING 0x00010000 +#define A5XX_SP_CS_CTRL_REG0_PIXLODENABLE 0x00100000 +#define A5XX_SP_CS_CTRL_REG0_BRANCHSTACK__MASK 0xfe000000 +#define A5XX_SP_CS_CTRL_REG0_BRANCHSTACK__SHIFT 25 +static inline uint32_t A5XX_SP_CS_CTRL_REG0_BRANCHSTACK(uint32_t val) +{ + return ((val) << A5XX_SP_CS_CTRL_REG0_BRANCHSTACK__SHIFT) & A5XX_SP_CS_CTRL_REG0_BRANCHSTACK__MASK; +} + +#define REG_A5XX_UNKNOWN_E600 0x0000e600 + +#define REG_A5XX_UNKNOWN_E602 0x0000e602 + +#define REG_A5XX_SP_HS_OBJ_START_LO 0x0000e603 + +#define REG_A5XX_SP_HS_OBJ_START_HI 0x0000e604 + +#define REG_A5XX_UNKNOWN_E62B 0x0000e62b + +#define REG_A5XX_SP_DS_OBJ_START_LO 0x0000e62c + +#define REG_A5XX_SP_DS_OBJ_START_HI 0x0000e62d + +#define REG_A5XX_UNKNOWN_E640 0x0000e640 + +#define REG_A5XX_UNKNOWN_E65B 0x0000e65b + +#define REG_A5XX_SP_GS_OBJ_START_LO 0x0000e65c + +#define REG_A5XX_SP_GS_OBJ_START_HI 0x0000e65d + +#define REG_A5XX_TPL1_TP_RAS_MSAA_CNTL 0x0000e704 +#define A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES__MASK 0x00000003 +#define A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES__SHIFT 0 +static inline uint32_t A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES(enum a3xx_msaa_samples val) +{ + return ((val) << A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES__SHIFT) & A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES__MASK; +} + +#define REG_A5XX_TPL1_TP_DEST_MSAA_CNTL 0x0000e705 +#define A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES__MASK 0x00000003 +#define A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES__SHIFT 0 +static inline uint32_t A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES(enum a3xx_msaa_samples val) +{ + return ((val) << A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES__SHIFT) & A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES__MASK; +} +#define A5XX_TPL1_TP_DEST_MSAA_CNTL_MSAA_DISABLE 0x00000004 + +#define REG_A5XX_TPL1_TP_BORDER_COLOR_BASE_ADDR_LO 0x0000e706 + +#define REG_A5XX_TPL1_TP_BORDER_COLOR_BASE_ADDR_HI 0x0000e707 + +#define REG_A5XX_TPL1_VS_TEX_COUNT 0x0000e700 + +#define REG_A5XX_TPL1_HS_TEX_COUNT 0x0000e701 + +#define REG_A5XX_TPL1_DS_TEX_COUNT 0x0000e702 + +#define REG_A5XX_TPL1_GS_TEX_COUNT 0x0000e703 + +#define REG_A5XX_TPL1_VS_TEX_SAMP_LO 0x0000e722 + +#define REG_A5XX_TPL1_VS_TEX_SAMP_HI 0x0000e723 + +#define REG_A5XX_TPL1_HS_TEX_SAMP_LO 0x0000e724 + +#define REG_A5XX_TPL1_HS_TEX_SAMP_HI 0x0000e725 + +#define REG_A5XX_TPL1_DS_TEX_SAMP_LO 0x0000e726 + +#define REG_A5XX_TPL1_DS_TEX_SAMP_HI 0x0000e727 + +#define REG_A5XX_TPL1_GS_TEX_SAMP_LO 0x0000e728 + +#define REG_A5XX_TPL1_GS_TEX_SAMP_HI 0x0000e729 + +#define REG_A5XX_TPL1_VS_TEX_CONST_LO 0x0000e72a + +#define REG_A5XX_TPL1_VS_TEX_CONST_HI 0x0000e72b + +#define REG_A5XX_TPL1_HS_TEX_CONST_LO 0x0000e72c + +#define REG_A5XX_TPL1_HS_TEX_CONST_HI 0x0000e72d + +#define REG_A5XX_TPL1_DS_TEX_CONST_LO 0x0000e72e + +#define REG_A5XX_TPL1_DS_TEX_CONST_HI 0x0000e72f + +#define REG_A5XX_TPL1_GS_TEX_CONST_LO 0x0000e730 + +#define REG_A5XX_TPL1_GS_TEX_CONST_HI 0x0000e731 + +#define REG_A5XX_TPL1_FS_TEX_COUNT 0x0000e750 + +#define REG_A5XX_TPL1_CS_TEX_COUNT 0x0000e751 + +#define REG_A5XX_TPL1_FS_TEX_SAMP_LO 0x0000e75a + +#define REG_A5XX_TPL1_FS_TEX_SAMP_HI 0x0000e75b + +#define REG_A5XX_TPL1_CS_TEX_SAMP_LO 0x0000e75c + +#define REG_A5XX_TPL1_CS_TEX_SAMP_HI 0x0000e75d + +#define REG_A5XX_TPL1_FS_TEX_CONST_LO 0x0000e75e + +#define REG_A5XX_TPL1_FS_TEX_CONST_HI 0x0000e75f + +#define REG_A5XX_TPL1_CS_TEX_CONST_LO 0x0000e760 + +#define REG_A5XX_TPL1_CS_TEX_CONST_HI 0x0000e761 + +#define REG_A5XX_TPL1_TP_FS_ROTATION_CNTL 0x0000e764 + +#define REG_A5XX_HLSQ_CONTROL_0_REG 0x0000e784 +#define A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK 0x00000001 +#define A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT) & A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK; +} +#define A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE__MASK 0x00000004 +#define A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE__SHIFT 2 +static inline uint32_t A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE(enum a3xx_threadsize val) +{ + return ((val) << A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE__SHIFT) & A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE__MASK; +} + +#define REG_A5XX_HLSQ_CONTROL_1_REG 0x0000e785 +#define A5XX_HLSQ_CONTROL_1_REG_PRIMALLOCTHRESHOLD__MASK 0x0000003f +#define A5XX_HLSQ_CONTROL_1_REG_PRIMALLOCTHRESHOLD__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CONTROL_1_REG_PRIMALLOCTHRESHOLD(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CONTROL_1_REG_PRIMALLOCTHRESHOLD__SHIFT) & A5XX_HLSQ_CONTROL_1_REG_PRIMALLOCTHRESHOLD__MASK; +} + +#define REG_A5XX_HLSQ_CONTROL_2_REG 0x0000e786 +#define A5XX_HLSQ_CONTROL_2_REG_FACEREGID__MASK 0x000000ff +#define A5XX_HLSQ_CONTROL_2_REG_FACEREGID__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CONTROL_2_REG_FACEREGID(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CONTROL_2_REG_FACEREGID__SHIFT) & A5XX_HLSQ_CONTROL_2_REG_FACEREGID__MASK; +} + +#define REG_A5XX_HLSQ_CONTROL_3_REG 0x0000e787 +#define A5XX_HLSQ_CONTROL_3_REG_FRAGCOORDXYREGID__MASK 0x000000ff +#define A5XX_HLSQ_CONTROL_3_REG_FRAGCOORDXYREGID__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CONTROL_3_REG_FRAGCOORDXYREGID(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CONTROL_3_REG_FRAGCOORDXYREGID__SHIFT) & A5XX_HLSQ_CONTROL_3_REG_FRAGCOORDXYREGID__MASK; +} + +#define REG_A5XX_HLSQ_CONTROL_4_REG 0x0000e788 +#define A5XX_HLSQ_CONTROL_4_REG_XYCOORDREGID__MASK 0x00ff0000 +#define A5XX_HLSQ_CONTROL_4_REG_XYCOORDREGID__SHIFT 16 +static inline uint32_t A5XX_HLSQ_CONTROL_4_REG_XYCOORDREGID(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CONTROL_4_REG_XYCOORDREGID__SHIFT) & A5XX_HLSQ_CONTROL_4_REG_XYCOORDREGID__MASK; +} +#define A5XX_HLSQ_CONTROL_4_REG_ZWCOORDREGID__MASK 0xff000000 +#define A5XX_HLSQ_CONTROL_4_REG_ZWCOORDREGID__SHIFT 24 +static inline uint32_t A5XX_HLSQ_CONTROL_4_REG_ZWCOORDREGID(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CONTROL_4_REG_ZWCOORDREGID__SHIFT) & A5XX_HLSQ_CONTROL_4_REG_ZWCOORDREGID__MASK; +} + +#define REG_A5XX_HLSQ_UPDATE_CNTL 0x0000e78a + +#define REG_A5XX_HLSQ_VS_CONFIG 0x0000e78b +#define A5XX_HLSQ_VS_CONFIG_ENABLED 0x00000001 +#define A5XX_HLSQ_VS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_HLSQ_VS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_HLSQ_VS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_VS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_HLSQ_VS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_HLSQ_VS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_HLSQ_VS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_HLSQ_VS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_VS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_HLSQ_VS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_HLSQ_FS_CONFIG 0x0000e78c +#define A5XX_HLSQ_FS_CONFIG_ENABLED 0x00000001 +#define A5XX_HLSQ_FS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_HLSQ_FS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_HLSQ_FS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_FS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_HLSQ_FS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_HLSQ_FS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_HLSQ_FS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_HLSQ_FS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_FS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_HLSQ_FS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_HLSQ_HS_CONFIG 0x0000e78d +#define A5XX_HLSQ_HS_CONFIG_ENABLED 0x00000001 +#define A5XX_HLSQ_HS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_HLSQ_HS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_HLSQ_HS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_HS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_HLSQ_HS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_HLSQ_HS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_HLSQ_HS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_HLSQ_HS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_HS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_HLSQ_HS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_HLSQ_DS_CONFIG 0x0000e78e +#define A5XX_HLSQ_DS_CONFIG_ENABLED 0x00000001 +#define A5XX_HLSQ_DS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_HLSQ_DS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_HLSQ_DS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_DS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_HLSQ_DS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_HLSQ_DS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_HLSQ_DS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_HLSQ_DS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_DS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_HLSQ_DS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_HLSQ_GS_CONFIG 0x0000e78f +#define A5XX_HLSQ_GS_CONFIG_ENABLED 0x00000001 +#define A5XX_HLSQ_GS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_HLSQ_GS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_HLSQ_GS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_GS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_HLSQ_GS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_HLSQ_GS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_HLSQ_GS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_HLSQ_GS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_GS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_HLSQ_GS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_HLSQ_CS_CONFIG 0x0000e790 +#define A5XX_HLSQ_CS_CONFIG_ENABLED 0x00000001 +#define A5XX_HLSQ_CS_CONFIG_CONSTOBJECTOFFSET__MASK 0x000000fe +#define A5XX_HLSQ_CS_CONFIG_CONSTOBJECTOFFSET__SHIFT 1 +static inline uint32_t A5XX_HLSQ_CS_CONFIG_CONSTOBJECTOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CONFIG_CONSTOBJECTOFFSET__SHIFT) & A5XX_HLSQ_CS_CONFIG_CONSTOBJECTOFFSET__MASK; +} +#define A5XX_HLSQ_CS_CONFIG_SHADEROBJOFFSET__MASK 0x00007f00 +#define A5XX_HLSQ_CS_CONFIG_SHADEROBJOFFSET__SHIFT 8 +static inline uint32_t A5XX_HLSQ_CS_CONFIG_SHADEROBJOFFSET(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CONFIG_SHADEROBJOFFSET__SHIFT) & A5XX_HLSQ_CS_CONFIG_SHADEROBJOFFSET__MASK; +} + +#define REG_A5XX_HLSQ_VS_CNTL 0x0000e791 +#define A5XX_HLSQ_VS_CNTL_SSBO_ENABLE 0x00000001 +#define A5XX_HLSQ_VS_CNTL_INSTRLEN__MASK 0xfffffffe +#define A5XX_HLSQ_VS_CNTL_INSTRLEN__SHIFT 1 +static inline uint32_t A5XX_HLSQ_VS_CNTL_INSTRLEN(uint32_t val) +{ + return ((val) << A5XX_HLSQ_VS_CNTL_INSTRLEN__SHIFT) & A5XX_HLSQ_VS_CNTL_INSTRLEN__MASK; +} + +#define REG_A5XX_HLSQ_FS_CNTL 0x0000e792 +#define A5XX_HLSQ_FS_CNTL_SSBO_ENABLE 0x00000001 +#define A5XX_HLSQ_FS_CNTL_INSTRLEN__MASK 0xfffffffe +#define A5XX_HLSQ_FS_CNTL_INSTRLEN__SHIFT 1 +static inline uint32_t A5XX_HLSQ_FS_CNTL_INSTRLEN(uint32_t val) +{ + return ((val) << A5XX_HLSQ_FS_CNTL_INSTRLEN__SHIFT) & A5XX_HLSQ_FS_CNTL_INSTRLEN__MASK; +} + +#define REG_A5XX_HLSQ_HS_CNTL 0x0000e793 +#define A5XX_HLSQ_HS_CNTL_SSBO_ENABLE 0x00000001 +#define A5XX_HLSQ_HS_CNTL_INSTRLEN__MASK 0xfffffffe +#define A5XX_HLSQ_HS_CNTL_INSTRLEN__SHIFT 1 +static inline uint32_t A5XX_HLSQ_HS_CNTL_INSTRLEN(uint32_t val) +{ + return ((val) << A5XX_HLSQ_HS_CNTL_INSTRLEN__SHIFT) & A5XX_HLSQ_HS_CNTL_INSTRLEN__MASK; +} + +#define REG_A5XX_HLSQ_DS_CNTL 0x0000e794 +#define A5XX_HLSQ_DS_CNTL_SSBO_ENABLE 0x00000001 +#define A5XX_HLSQ_DS_CNTL_INSTRLEN__MASK 0xfffffffe +#define A5XX_HLSQ_DS_CNTL_INSTRLEN__SHIFT 1 +static inline uint32_t A5XX_HLSQ_DS_CNTL_INSTRLEN(uint32_t val) +{ + return ((val) << A5XX_HLSQ_DS_CNTL_INSTRLEN__SHIFT) & A5XX_HLSQ_DS_CNTL_INSTRLEN__MASK; +} + +#define REG_A5XX_HLSQ_GS_CNTL 0x0000e795 +#define A5XX_HLSQ_GS_CNTL_SSBO_ENABLE 0x00000001 +#define A5XX_HLSQ_GS_CNTL_INSTRLEN__MASK 0xfffffffe +#define A5XX_HLSQ_GS_CNTL_INSTRLEN__SHIFT 1 +static inline uint32_t A5XX_HLSQ_GS_CNTL_INSTRLEN(uint32_t val) +{ + return ((val) << A5XX_HLSQ_GS_CNTL_INSTRLEN__SHIFT) & A5XX_HLSQ_GS_CNTL_INSTRLEN__MASK; +} + +#define REG_A5XX_HLSQ_CS_CNTL 0x0000e796 +#define A5XX_HLSQ_CS_CNTL_SSBO_ENABLE 0x00000001 +#define A5XX_HLSQ_CS_CNTL_INSTRLEN__MASK 0xfffffffe +#define A5XX_HLSQ_CS_CNTL_INSTRLEN__SHIFT 1 +static inline uint32_t A5XX_HLSQ_CS_CNTL_INSTRLEN(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CNTL_INSTRLEN__SHIFT) & A5XX_HLSQ_CS_CNTL_INSTRLEN__MASK; +} + +#define REG_A5XX_HLSQ_CS_KERNEL_GROUP_X 0x0000e7b9 + +#define REG_A5XX_HLSQ_CS_KERNEL_GROUP_Y 0x0000e7ba + +#define REG_A5XX_HLSQ_CS_KERNEL_GROUP_Z 0x0000e7bb + +#define REG_A5XX_HLSQ_CS_NDRANGE_0 0x0000e7b0 +#define A5XX_HLSQ_CS_NDRANGE_0_KERNELDIM__MASK 0x00000003 +#define A5XX_HLSQ_CS_NDRANGE_0_KERNELDIM__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_0_KERNELDIM(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_0_KERNELDIM__SHIFT) & A5XX_HLSQ_CS_NDRANGE_0_KERNELDIM__MASK; +} +#define A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEX__MASK 0x00000ffc +#define A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEX__SHIFT 2 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEX(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEX__SHIFT) & A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEX__MASK; +} +#define A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEY__MASK 0x003ff000 +#define A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEY__SHIFT 12 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEY(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEY__SHIFT) & A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEY__MASK; +} +#define A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEZ__MASK 0xffc00000 +#define A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEZ__SHIFT 22 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEZ(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEZ__SHIFT) & A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEZ__MASK; +} + +#define REG_A5XX_HLSQ_CS_NDRANGE_1 0x0000e7b1 +#define A5XX_HLSQ_CS_NDRANGE_1_SIZE_X__MASK 0xffffffff +#define A5XX_HLSQ_CS_NDRANGE_1_SIZE_X__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_1_SIZE_X(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_1_SIZE_X__SHIFT) & A5XX_HLSQ_CS_NDRANGE_1_SIZE_X__MASK; +} + +#define REG_A5XX_HLSQ_CS_NDRANGE_2 0x0000e7b2 + +#define REG_A5XX_HLSQ_CS_NDRANGE_3 0x0000e7b3 +#define A5XX_HLSQ_CS_NDRANGE_3_SIZE_Y__MASK 0xffffffff +#define A5XX_HLSQ_CS_NDRANGE_3_SIZE_Y__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_3_SIZE_Y(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_3_SIZE_Y__SHIFT) & A5XX_HLSQ_CS_NDRANGE_3_SIZE_Y__MASK; +} + +#define REG_A5XX_HLSQ_CS_NDRANGE_4 0x0000e7b4 + +#define REG_A5XX_HLSQ_CS_NDRANGE_5 0x0000e7b5 +#define A5XX_HLSQ_CS_NDRANGE_5_SIZE_Z__MASK 0xffffffff +#define A5XX_HLSQ_CS_NDRANGE_5_SIZE_Z__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CS_NDRANGE_5_SIZE_Z(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_NDRANGE_5_SIZE_Z__SHIFT) & A5XX_HLSQ_CS_NDRANGE_5_SIZE_Z__MASK; +} + +#define REG_A5XX_HLSQ_CS_NDRANGE_6 0x0000e7b6 + +#define REG_A5XX_HLSQ_CS_CNTL_0 0x0000e7b7 +#define A5XX_HLSQ_CS_CNTL_0_WGIDCONSTID__MASK 0x000000ff +#define A5XX_HLSQ_CS_CNTL_0_WGIDCONSTID__SHIFT 0 +static inline uint32_t A5XX_HLSQ_CS_CNTL_0_WGIDCONSTID(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CNTL_0_WGIDCONSTID__SHIFT) & A5XX_HLSQ_CS_CNTL_0_WGIDCONSTID__MASK; +} +#define A5XX_HLSQ_CS_CNTL_0_UNK0__MASK 0x0000ff00 +#define A5XX_HLSQ_CS_CNTL_0_UNK0__SHIFT 8 +static inline uint32_t A5XX_HLSQ_CS_CNTL_0_UNK0(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CNTL_0_UNK0__SHIFT) & A5XX_HLSQ_CS_CNTL_0_UNK0__MASK; +} +#define A5XX_HLSQ_CS_CNTL_0_UNK1__MASK 0x00ff0000 +#define A5XX_HLSQ_CS_CNTL_0_UNK1__SHIFT 16 +static inline uint32_t A5XX_HLSQ_CS_CNTL_0_UNK1(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CNTL_0_UNK1__SHIFT) & A5XX_HLSQ_CS_CNTL_0_UNK1__MASK; +} +#define A5XX_HLSQ_CS_CNTL_0_LOCALIDREGID__MASK 0xff000000 +#define A5XX_HLSQ_CS_CNTL_0_LOCALIDREGID__SHIFT 24 +static inline uint32_t A5XX_HLSQ_CS_CNTL_0_LOCALIDREGID(uint32_t val) +{ + return ((val) << A5XX_HLSQ_CS_CNTL_0_LOCALIDREGID__SHIFT) & A5XX_HLSQ_CS_CNTL_0_LOCALIDREGID__MASK; +} + +#define REG_A5XX_HLSQ_CS_CNTL_1 0x0000e7b8 + +#define REG_A5XX_UNKNOWN_E7C0 0x0000e7c0 + +#define REG_A5XX_HLSQ_VS_CONSTLEN 0x0000e7c3 + +#define REG_A5XX_HLSQ_VS_INSTRLEN 0x0000e7c4 + +#define REG_A5XX_UNKNOWN_E7C5 0x0000e7c5 + +#define REG_A5XX_HLSQ_HS_CONSTLEN 0x0000e7c8 + +#define REG_A5XX_HLSQ_HS_INSTRLEN 0x0000e7c9 + +#define REG_A5XX_UNKNOWN_E7CA 0x0000e7ca + +#define REG_A5XX_HLSQ_DS_CONSTLEN 0x0000e7cd + +#define REG_A5XX_HLSQ_DS_INSTRLEN 0x0000e7ce + +#define REG_A5XX_UNKNOWN_E7CF 0x0000e7cf + +#define REG_A5XX_HLSQ_GS_CONSTLEN 0x0000e7d2 + +#define REG_A5XX_HLSQ_GS_INSTRLEN 0x0000e7d3 + +#define REG_A5XX_UNKNOWN_E7D4 0x0000e7d4 + +#define REG_A5XX_HLSQ_FS_CONSTLEN 0x0000e7d7 + +#define REG_A5XX_HLSQ_FS_INSTRLEN 0x0000e7d8 + +#define REG_A5XX_UNKNOWN_E7D9 0x0000e7d9 + +#define REG_A5XX_HLSQ_CS_CONSTLEN 0x0000e7dc + +#define REG_A5XX_HLSQ_CS_INSTRLEN 0x0000e7dd + +#define REG_A5XX_RB_2D_SRC_SOLID_DW0 0x00002101 + +#define REG_A5XX_RB_2D_SRC_SOLID_DW1 0x00002102 + +#define REG_A5XX_RB_2D_SRC_SOLID_DW2 0x00002103 + +#define REG_A5XX_RB_2D_SRC_SOLID_DW3 0x00002104 + +#define REG_A5XX_RB_2D_SRC_INFO 0x00002107 +#define A5XX_RB_2D_SRC_INFO_COLOR_FORMAT__MASK 0x000000ff +#define A5XX_RB_2D_SRC_INFO_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A5XX_RB_2D_SRC_INFO_COLOR_FORMAT(enum a5xx_color_fmt val) +{ + return ((val) << A5XX_RB_2D_SRC_INFO_COLOR_FORMAT__SHIFT) & A5XX_RB_2D_SRC_INFO_COLOR_FORMAT__MASK; +} +#define A5XX_RB_2D_SRC_INFO_COLOR_SWAP__MASK 0x00000c00 +#define A5XX_RB_2D_SRC_INFO_COLOR_SWAP__SHIFT 10 +static inline uint32_t A5XX_RB_2D_SRC_INFO_COLOR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_RB_2D_SRC_INFO_COLOR_SWAP__SHIFT) & A5XX_RB_2D_SRC_INFO_COLOR_SWAP__MASK; +} + +#define REG_A5XX_RB_2D_SRC_LO 0x00002108 + +#define REG_A5XX_RB_2D_SRC_HI 0x00002109 + +#define REG_A5XX_RB_2D_SRC_SIZE 0x0000210a +#define A5XX_RB_2D_SRC_SIZE_PITCH__MASK 0x0000ffff +#define A5XX_RB_2D_SRC_SIZE_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_2D_SRC_SIZE_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_2D_SRC_SIZE_PITCH__SHIFT) & A5XX_RB_2D_SRC_SIZE_PITCH__MASK; +} +#define A5XX_RB_2D_SRC_SIZE_ARRAY_PITCH__MASK 0xffff0000 +#define A5XX_RB_2D_SRC_SIZE_ARRAY_PITCH__SHIFT 16 +static inline uint32_t A5XX_RB_2D_SRC_SIZE_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_2D_SRC_SIZE_ARRAY_PITCH__SHIFT) & A5XX_RB_2D_SRC_SIZE_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_2D_DST_INFO 0x00002110 +#define A5XX_RB_2D_DST_INFO_COLOR_FORMAT__MASK 0x000000ff +#define A5XX_RB_2D_DST_INFO_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A5XX_RB_2D_DST_INFO_COLOR_FORMAT(enum a5xx_color_fmt val) +{ + return ((val) << A5XX_RB_2D_DST_INFO_COLOR_FORMAT__SHIFT) & A5XX_RB_2D_DST_INFO_COLOR_FORMAT__MASK; +} +#define A5XX_RB_2D_DST_INFO_COLOR_SWAP__MASK 0x00000c00 +#define A5XX_RB_2D_DST_INFO_COLOR_SWAP__SHIFT 10 +static inline uint32_t A5XX_RB_2D_DST_INFO_COLOR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_RB_2D_DST_INFO_COLOR_SWAP__SHIFT) & A5XX_RB_2D_DST_INFO_COLOR_SWAP__MASK; +} + +#define REG_A5XX_RB_2D_DST_LO 0x00002111 + +#define REG_A5XX_RB_2D_DST_HI 0x00002112 + +#define REG_A5XX_RB_2D_DST_SIZE 0x00002113 +#define A5XX_RB_2D_DST_SIZE_PITCH__MASK 0x0000ffff +#define A5XX_RB_2D_DST_SIZE_PITCH__SHIFT 0 +static inline uint32_t A5XX_RB_2D_DST_SIZE_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_2D_DST_SIZE_PITCH__SHIFT) & A5XX_RB_2D_DST_SIZE_PITCH__MASK; +} +#define A5XX_RB_2D_DST_SIZE_ARRAY_PITCH__MASK 0xffff0000 +#define A5XX_RB_2D_DST_SIZE_ARRAY_PITCH__SHIFT 16 +static inline uint32_t A5XX_RB_2D_DST_SIZE_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0x3f)); + return ((val >> 6) << A5XX_RB_2D_DST_SIZE_ARRAY_PITCH__SHIFT) & A5XX_RB_2D_DST_SIZE_ARRAY_PITCH__MASK; +} + +#define REG_A5XX_RB_2D_SRC_FLAGS_LO 0x00002140 + +#define REG_A5XX_RB_2D_SRC_FLAGS_HI 0x00002141 + +#define REG_A5XX_RB_2D_DST_FLAGS_LO 0x00002143 + +#define REG_A5XX_RB_2D_DST_FLAGS_HI 0x00002144 + +#define REG_A5XX_GRAS_2D_SRC_INFO 0x00002181 +#define A5XX_GRAS_2D_SRC_INFO_COLOR_FORMAT__MASK 0x000000ff +#define A5XX_GRAS_2D_SRC_INFO_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A5XX_GRAS_2D_SRC_INFO_COLOR_FORMAT(enum a5xx_color_fmt val) +{ + return ((val) << A5XX_GRAS_2D_SRC_INFO_COLOR_FORMAT__SHIFT) & A5XX_GRAS_2D_SRC_INFO_COLOR_FORMAT__MASK; +} +#define A5XX_GRAS_2D_SRC_INFO_COLOR_SWAP__MASK 0x00000c00 +#define A5XX_GRAS_2D_SRC_INFO_COLOR_SWAP__SHIFT 10 +static inline uint32_t A5XX_GRAS_2D_SRC_INFO_COLOR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_GRAS_2D_SRC_INFO_COLOR_SWAP__SHIFT) & A5XX_GRAS_2D_SRC_INFO_COLOR_SWAP__MASK; +} + +#define REG_A5XX_GRAS_2D_DST_INFO 0x00002182 +#define A5XX_GRAS_2D_DST_INFO_COLOR_FORMAT__MASK 0x000000ff +#define A5XX_GRAS_2D_DST_INFO_COLOR_FORMAT__SHIFT 0 +static inline uint32_t A5XX_GRAS_2D_DST_INFO_COLOR_FORMAT(enum a5xx_color_fmt val) +{ + return ((val) << A5XX_GRAS_2D_DST_INFO_COLOR_FORMAT__SHIFT) & A5XX_GRAS_2D_DST_INFO_COLOR_FORMAT__MASK; +} +#define A5XX_GRAS_2D_DST_INFO_COLOR_SWAP__MASK 0x00000c00 +#define A5XX_GRAS_2D_DST_INFO_COLOR_SWAP__SHIFT 10 +static inline uint32_t A5XX_GRAS_2D_DST_INFO_COLOR_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_GRAS_2D_DST_INFO_COLOR_SWAP__SHIFT) & A5XX_GRAS_2D_DST_INFO_COLOR_SWAP__MASK; +} + +#define REG_A5XX_UNKNOWN_2100 0x00002100 + +#define REG_A5XX_UNKNOWN_2180 0x00002180 + +#define REG_A5XX_UNKNOWN_2184 0x00002184 + +#define REG_A5XX_TEX_SAMP_0 0x00000000 +#define A5XX_TEX_SAMP_0_MIPFILTER_LINEAR_NEAR 0x00000001 +#define A5XX_TEX_SAMP_0_XY_MAG__MASK 0x00000006 +#define A5XX_TEX_SAMP_0_XY_MAG__SHIFT 1 +static inline uint32_t A5XX_TEX_SAMP_0_XY_MAG(enum a5xx_tex_filter val) +{ + return ((val) << A5XX_TEX_SAMP_0_XY_MAG__SHIFT) & A5XX_TEX_SAMP_0_XY_MAG__MASK; +} +#define A5XX_TEX_SAMP_0_XY_MIN__MASK 0x00000018 +#define A5XX_TEX_SAMP_0_XY_MIN__SHIFT 3 +static inline uint32_t A5XX_TEX_SAMP_0_XY_MIN(enum a5xx_tex_filter val) +{ + return ((val) << A5XX_TEX_SAMP_0_XY_MIN__SHIFT) & A5XX_TEX_SAMP_0_XY_MIN__MASK; +} +#define A5XX_TEX_SAMP_0_WRAP_S__MASK 0x000000e0 +#define A5XX_TEX_SAMP_0_WRAP_S__SHIFT 5 +static inline uint32_t A5XX_TEX_SAMP_0_WRAP_S(enum a5xx_tex_clamp val) +{ + return ((val) << A5XX_TEX_SAMP_0_WRAP_S__SHIFT) & A5XX_TEX_SAMP_0_WRAP_S__MASK; +} +#define A5XX_TEX_SAMP_0_WRAP_T__MASK 0x00000700 +#define A5XX_TEX_SAMP_0_WRAP_T__SHIFT 8 +static inline uint32_t A5XX_TEX_SAMP_0_WRAP_T(enum a5xx_tex_clamp val) +{ + return ((val) << A5XX_TEX_SAMP_0_WRAP_T__SHIFT) & A5XX_TEX_SAMP_0_WRAP_T__MASK; +} +#define A5XX_TEX_SAMP_0_WRAP_R__MASK 0x00003800 +#define A5XX_TEX_SAMP_0_WRAP_R__SHIFT 11 +static inline uint32_t A5XX_TEX_SAMP_0_WRAP_R(enum a5xx_tex_clamp val) +{ + return ((val) << A5XX_TEX_SAMP_0_WRAP_R__SHIFT) & A5XX_TEX_SAMP_0_WRAP_R__MASK; +} +#define A5XX_TEX_SAMP_0_ANISO__MASK 0x0001c000 +#define A5XX_TEX_SAMP_0_ANISO__SHIFT 14 +static inline uint32_t A5XX_TEX_SAMP_0_ANISO(enum a5xx_tex_aniso val) +{ + return ((val) << A5XX_TEX_SAMP_0_ANISO__SHIFT) & A5XX_TEX_SAMP_0_ANISO__MASK; +} +#define A5XX_TEX_SAMP_0_LOD_BIAS__MASK 0xfff80000 +#define A5XX_TEX_SAMP_0_LOD_BIAS__SHIFT 19 +static inline uint32_t A5XX_TEX_SAMP_0_LOD_BIAS(float val) +{ + return ((((int32_t)(val * 256.0))) << A5XX_TEX_SAMP_0_LOD_BIAS__SHIFT) & A5XX_TEX_SAMP_0_LOD_BIAS__MASK; +} + +#define REG_A5XX_TEX_SAMP_1 0x00000001 +#define A5XX_TEX_SAMP_1_COMPARE_FUNC__MASK 0x0000000e +#define A5XX_TEX_SAMP_1_COMPARE_FUNC__SHIFT 1 +static inline uint32_t A5XX_TEX_SAMP_1_COMPARE_FUNC(enum adreno_compare_func val) +{ + return ((val) << A5XX_TEX_SAMP_1_COMPARE_FUNC__SHIFT) & A5XX_TEX_SAMP_1_COMPARE_FUNC__MASK; +} +#define A5XX_TEX_SAMP_1_CUBEMAPSEAMLESSFILTOFF 0x00000010 +#define A5XX_TEX_SAMP_1_UNNORM_COORDS 0x00000020 +#define A5XX_TEX_SAMP_1_MIPFILTER_LINEAR_FAR 0x00000040 +#define A5XX_TEX_SAMP_1_MAX_LOD__MASK 0x000fff00 +#define A5XX_TEX_SAMP_1_MAX_LOD__SHIFT 8 +static inline uint32_t A5XX_TEX_SAMP_1_MAX_LOD(float val) +{ + return ((((uint32_t)(val * 256.0))) << A5XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A5XX_TEX_SAMP_1_MAX_LOD__MASK; +} +#define A5XX_TEX_SAMP_1_MIN_LOD__MASK 0xfff00000 +#define A5XX_TEX_SAMP_1_MIN_LOD__SHIFT 20 +static inline uint32_t A5XX_TEX_SAMP_1_MIN_LOD(float val) +{ + return ((((uint32_t)(val * 256.0))) << A5XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A5XX_TEX_SAMP_1_MIN_LOD__MASK; +} + +#define REG_A5XX_TEX_SAMP_2 0x00000002 +#define A5XX_TEX_SAMP_2_BCOLOR_OFFSET__MASK 0xfffffff0 +#define A5XX_TEX_SAMP_2_BCOLOR_OFFSET__SHIFT 4 +static inline uint32_t A5XX_TEX_SAMP_2_BCOLOR_OFFSET(uint32_t val) +{ + return ((val) << A5XX_TEX_SAMP_2_BCOLOR_OFFSET__SHIFT) & A5XX_TEX_SAMP_2_BCOLOR_OFFSET__MASK; +} + +#define REG_A5XX_TEX_SAMP_3 0x00000003 + +#define REG_A5XX_TEX_CONST_0 0x00000000 +#define A5XX_TEX_CONST_0_TILE_MODE__MASK 0x00000003 +#define A5XX_TEX_CONST_0_TILE_MODE__SHIFT 0 +static inline uint32_t A5XX_TEX_CONST_0_TILE_MODE(enum a5xx_tile_mode val) +{ + return ((val) << A5XX_TEX_CONST_0_TILE_MODE__SHIFT) & A5XX_TEX_CONST_0_TILE_MODE__MASK; +} +#define A5XX_TEX_CONST_0_SRGB 0x00000004 +#define A5XX_TEX_CONST_0_SWIZ_X__MASK 0x00000070 +#define A5XX_TEX_CONST_0_SWIZ_X__SHIFT 4 +static inline uint32_t A5XX_TEX_CONST_0_SWIZ_X(enum a5xx_tex_swiz val) +{ + return ((val) << A5XX_TEX_CONST_0_SWIZ_X__SHIFT) & A5XX_TEX_CONST_0_SWIZ_X__MASK; +} +#define A5XX_TEX_CONST_0_SWIZ_Y__MASK 0x00000380 +#define A5XX_TEX_CONST_0_SWIZ_Y__SHIFT 7 +static inline uint32_t A5XX_TEX_CONST_0_SWIZ_Y(enum a5xx_tex_swiz val) +{ + return ((val) << A5XX_TEX_CONST_0_SWIZ_Y__SHIFT) & A5XX_TEX_CONST_0_SWIZ_Y__MASK; +} +#define A5XX_TEX_CONST_0_SWIZ_Z__MASK 0x00001c00 +#define A5XX_TEX_CONST_0_SWIZ_Z__SHIFT 10 +static inline uint32_t A5XX_TEX_CONST_0_SWIZ_Z(enum a5xx_tex_swiz val) +{ + return ((val) << A5XX_TEX_CONST_0_SWIZ_Z__SHIFT) & A5XX_TEX_CONST_0_SWIZ_Z__MASK; +} +#define A5XX_TEX_CONST_0_SWIZ_W__MASK 0x0000e000 +#define A5XX_TEX_CONST_0_SWIZ_W__SHIFT 13 +static inline uint32_t A5XX_TEX_CONST_0_SWIZ_W(enum a5xx_tex_swiz val) +{ + return ((val) << A5XX_TEX_CONST_0_SWIZ_W__SHIFT) & A5XX_TEX_CONST_0_SWIZ_W__MASK; +} +#define A5XX_TEX_CONST_0_MIPLVLS__MASK 0x000f0000 +#define A5XX_TEX_CONST_0_MIPLVLS__SHIFT 16 +static inline uint32_t A5XX_TEX_CONST_0_MIPLVLS(uint32_t val) +{ + return ((val) << A5XX_TEX_CONST_0_MIPLVLS__SHIFT) & A5XX_TEX_CONST_0_MIPLVLS__MASK; +} +#define A5XX_TEX_CONST_0_FMT__MASK 0x3fc00000 +#define A5XX_TEX_CONST_0_FMT__SHIFT 22 +static inline uint32_t A5XX_TEX_CONST_0_FMT(enum a5xx_tex_fmt val) +{ + return ((val) << A5XX_TEX_CONST_0_FMT__SHIFT) & A5XX_TEX_CONST_0_FMT__MASK; +} +#define A5XX_TEX_CONST_0_SWAP__MASK 0xc0000000 +#define A5XX_TEX_CONST_0_SWAP__SHIFT 30 +static inline uint32_t A5XX_TEX_CONST_0_SWAP(enum a3xx_color_swap val) +{ + return ((val) << A5XX_TEX_CONST_0_SWAP__SHIFT) & A5XX_TEX_CONST_0_SWAP__MASK; +} + +#define REG_A5XX_TEX_CONST_1 0x00000001 +#define A5XX_TEX_CONST_1_WIDTH__MASK 0x00007fff +#define A5XX_TEX_CONST_1_WIDTH__SHIFT 0 +static inline uint32_t A5XX_TEX_CONST_1_WIDTH(uint32_t val) +{ + return ((val) << A5XX_TEX_CONST_1_WIDTH__SHIFT) & A5XX_TEX_CONST_1_WIDTH__MASK; +} +#define A5XX_TEX_CONST_1_HEIGHT__MASK 0x3fff8000 +#define A5XX_TEX_CONST_1_HEIGHT__SHIFT 15 +static inline uint32_t A5XX_TEX_CONST_1_HEIGHT(uint32_t val) +{ + return ((val) << A5XX_TEX_CONST_1_HEIGHT__SHIFT) & A5XX_TEX_CONST_1_HEIGHT__MASK; +} + +#define REG_A5XX_TEX_CONST_2 0x00000002 +#define A5XX_TEX_CONST_2_FETCHSIZE__MASK 0x0000000f +#define A5XX_TEX_CONST_2_FETCHSIZE__SHIFT 0 +static inline uint32_t A5XX_TEX_CONST_2_FETCHSIZE(enum a5xx_tex_fetchsize val) +{ + return ((val) << A5XX_TEX_CONST_2_FETCHSIZE__SHIFT) & A5XX_TEX_CONST_2_FETCHSIZE__MASK; +} +#define A5XX_TEX_CONST_2_PITCH__MASK 0x1fffff80 +#define A5XX_TEX_CONST_2_PITCH__SHIFT 7 +static inline uint32_t A5XX_TEX_CONST_2_PITCH(uint32_t val) +{ + return ((val) << A5XX_TEX_CONST_2_PITCH__SHIFT) & A5XX_TEX_CONST_2_PITCH__MASK; +} +#define A5XX_TEX_CONST_2_TYPE__MASK 0x60000000 +#define A5XX_TEX_CONST_2_TYPE__SHIFT 29 +static inline uint32_t A5XX_TEX_CONST_2_TYPE(enum a5xx_tex_type val) +{ + return ((val) << A5XX_TEX_CONST_2_TYPE__SHIFT) & A5XX_TEX_CONST_2_TYPE__MASK; +} + +#define REG_A5XX_TEX_CONST_3 0x00000003 +#define A5XX_TEX_CONST_3_ARRAY_PITCH__MASK 0x00003fff +#define A5XX_TEX_CONST_3_ARRAY_PITCH__SHIFT 0 +static inline uint32_t A5XX_TEX_CONST_3_ARRAY_PITCH(uint32_t val) +{ + assert(!(val & 0xfff)); + return ((val >> 12) << A5XX_TEX_CONST_3_ARRAY_PITCH__SHIFT) & A5XX_TEX_CONST_3_ARRAY_PITCH__MASK; +} +#define A5XX_TEX_CONST_3_FLAG 0x10000000 + +#define REG_A5XX_TEX_CONST_4 0x00000004 +#define A5XX_TEX_CONST_4_BASE_LO__MASK 0xffffffe0 +#define A5XX_TEX_CONST_4_BASE_LO__SHIFT 5 +static inline uint32_t A5XX_TEX_CONST_4_BASE_LO(uint32_t val) +{ + assert(!(val & 0x1f)); + return ((val >> 5) << A5XX_TEX_CONST_4_BASE_LO__SHIFT) & A5XX_TEX_CONST_4_BASE_LO__MASK; +} + +#define REG_A5XX_TEX_CONST_5 0x00000005 +#define A5XX_TEX_CONST_5_BASE_HI__MASK 0x0001ffff +#define A5XX_TEX_CONST_5_BASE_HI__SHIFT 0 +static inline uint32_t A5XX_TEX_CONST_5_BASE_HI(uint32_t val) +{ + return ((val) << A5XX_TEX_CONST_5_BASE_HI__SHIFT) & A5XX_TEX_CONST_5_BASE_HI__MASK; +} +#define A5XX_TEX_CONST_5_DEPTH__MASK 0x3ffe0000 +#define A5XX_TEX_CONST_5_DEPTH__SHIFT 17 +static inline uint32_t A5XX_TEX_CONST_5_DEPTH(uint32_t val) +{ + return ((val) << A5XX_TEX_CONST_5_DEPTH__SHIFT) & A5XX_TEX_CONST_5_DEPTH__MASK; +} + +#define REG_A5XX_TEX_CONST_6 0x00000006 + +#define REG_A5XX_TEX_CONST_7 0x00000007 + +#define REG_A5XX_TEX_CONST_8 0x00000008 + +#define REG_A5XX_TEX_CONST_9 0x00000009 + +#define REG_A5XX_TEX_CONST_10 0x0000000a + +#define REG_A5XX_TEX_CONST_11 0x0000000b + + +#endif /* A5XX_XML */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_blend.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_blend.c new file mode 100644 index 000000000..98b6d4498 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_blend.c @@ -0,0 +1,148 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" +#include "util/u_blend.h" +#include "util/u_string.h" +#include "util/u_memory.h" + +#include "fd5_blend.h" +#include "fd5_context.h" +#include "fd5_format.h" + +// XXX move somewhere common.. same across a3xx/a4xx/a5xx.. +static enum a3xx_rb_blend_opcode +blend_func(unsigned func) +{ + switch (func) { + case PIPE_BLEND_ADD: + return BLEND_DST_PLUS_SRC; + case PIPE_BLEND_MIN: + return BLEND_MIN_DST_SRC; + case PIPE_BLEND_MAX: + return BLEND_MAX_DST_SRC; + case PIPE_BLEND_SUBTRACT: + return BLEND_SRC_MINUS_DST; + case PIPE_BLEND_REVERSE_SUBTRACT: + return BLEND_DST_MINUS_SRC; + default: + DBG("invalid blend func: %x", func); + return 0; + } +} + +void * +fd5_blend_state_create(struct pipe_context *pctx, + const struct pipe_blend_state *cso) +{ + struct fd5_blend_stateobj *so; + enum a3xx_rop_code rop = ROP_COPY; + bool reads_dest = false; + unsigned i, mrt_blend = 0; + + if (cso->logicop_enable) { + rop = cso->logicop_func; /* maps 1:1 */ + + switch (cso->logicop_func) { + case PIPE_LOGICOP_NOR: + case PIPE_LOGICOP_AND_INVERTED: + case PIPE_LOGICOP_AND_REVERSE: + case PIPE_LOGICOP_INVERT: + case PIPE_LOGICOP_XOR: + case PIPE_LOGICOP_NAND: + case PIPE_LOGICOP_AND: + case PIPE_LOGICOP_EQUIV: + case PIPE_LOGICOP_NOOP: + case PIPE_LOGICOP_OR_INVERTED: + case PIPE_LOGICOP_OR_REVERSE: + case PIPE_LOGICOP_OR: + reads_dest = true; + break; + } + } + + so = CALLOC_STRUCT(fd5_blend_stateobj); + if (!so) + return NULL; + + so->base = *cso; + + so->lrz_write = true; /* unless blend enabled for any MRT */ + + for (i = 0; i < ARRAY_SIZE(so->rb_mrt); i++) { + const struct pipe_rt_blend_state *rt; + + if (cso->independent_blend_enable) + rt = &cso->rt[i]; + else + rt = &cso->rt[0]; + + so->rb_mrt[i].blend_control_rgb = + A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR(fd_blend_factor(rt->rgb_src_factor)) | + A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE(blend_func(rt->rgb_func)) | + A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR(fd_blend_factor(rt->rgb_dst_factor)); + + so->rb_mrt[i].blend_control_alpha = + A5XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR(fd_blend_factor(rt->alpha_src_factor)) | + A5XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE(blend_func(rt->alpha_func)) | + A5XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR(fd_blend_factor(rt->alpha_dst_factor)); + + so->rb_mrt[i].blend_control_no_alpha_rgb = + A5XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR(fd_blend_factor(util_blend_dst_alpha_to_one(rt->rgb_src_factor))) | + A5XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE(blend_func(rt->rgb_func)) | + A5XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR(fd_blend_factor(util_blend_dst_alpha_to_one(rt->rgb_dst_factor))); + + + so->rb_mrt[i].control = + A5XX_RB_MRT_CONTROL_ROP_CODE(rop) | + COND(cso->logicop_enable, A5XX_RB_MRT_CONTROL_ROP_ENABLE) | + A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE(rt->colormask); + + if (rt->blend_enable) { + so->rb_mrt[i].control |= +// A5XX_RB_MRT_CONTROL_READ_DEST_ENABLE | + A5XX_RB_MRT_CONTROL_BLEND | + A5XX_RB_MRT_CONTROL_BLEND2; + mrt_blend |= (1 << i); + so->lrz_write = false; + } + + if (reads_dest) { +// so->rb_mrt[i].control |= A5XX_RB_MRT_CONTROL_READ_DEST_ENABLE; + mrt_blend |= (1 << i); + } + +// if (cso->dither) +// so->rb_mrt[i].buf_info |= A5XX_RB_MRT_BUF_INFO_DITHER_MODE(DITHER_ALWAYS); + } + + so->rb_blend_cntl = A5XX_RB_BLEND_CNTL_ENABLE_BLEND(mrt_blend) | + COND(cso->independent_blend_enable, A5XX_RB_BLEND_CNTL_INDEPENDENT_BLEND); + so->sp_blend_cntl = A5XX_SP_BLEND_CNTL_UNK8 | + COND(mrt_blend, A5XX_SP_BLEND_CNTL_ENABLED); + + return so; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_blend.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_blend.h new file mode 100644 index 000000000..698549548 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_blend.h @@ -0,0 +1,62 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_BLEND_H_ +#define FD5_BLEND_H_ + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + +#include "freedreno_util.h" + +struct fd5_blend_stateobj { + struct pipe_blend_state base; + + struct { + uint32_t control; + uint32_t buf_info; + /* Blend control bits for color if there is an alpha channel */ + uint32_t blend_control_rgb; + /* Blend control bits for color if there is no alpha channel */ + uint32_t blend_control_no_alpha_rgb; + /* Blend control bits for alpha channel */ + uint32_t blend_control_alpha; + } rb_mrt[A5XX_MAX_RENDER_TARGETS]; + uint32_t rb_blend_cntl; + uint32_t sp_blend_cntl; + bool lrz_write; +}; + +static inline struct fd5_blend_stateobj * +fd5_blend_stateobj(struct pipe_blend_state *blend) +{ + return (struct fd5_blend_stateobj *)blend; +} + +void * fd5_blend_state_create(struct pipe_context *pctx, + const struct pipe_blend_state *cso); + +#endif /* FD5_BLEND_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_compute.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_compute.c new file mode 100644 index 000000000..2efcece41 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_compute.c @@ -0,0 +1,175 @@ +/* + * Copyright (C) 2017 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" + +#include "fd5_compute.h" +#include "fd5_context.h" +#include "fd5_emit.h" + +struct fd5_compute_stateobj { + struct ir3_shader *shader; +}; + + +static void * +fd5_create_compute_state(struct pipe_context *pctx, + const struct pipe_compute_state *cso) +{ + struct fd_context *ctx = fd_context(pctx); + struct ir3_compiler *compiler = ctx->screen->compiler; + struct fd5_compute_stateobj *so = CALLOC_STRUCT(fd5_compute_stateobj); + so->shader = ir3_shader_create_compute(compiler, cso, &ctx->debug); + return so; +} + +static void +fd5_delete_compute_state(struct pipe_context *pctx, void *hwcso) +{ + struct fd5_compute_stateobj *so = hwcso; + ir3_shader_destroy(so->shader); + free(so); +} + +/* maybe move to fd5_program? */ +static void +cs_program_emit(struct fd_ringbuffer *ring, struct ir3_shader_variant *v) +{ + const struct ir3_info *i = &v->info; + enum a3xx_threadsize thrsz; + + /* note: blob uses local_size_x/y/z threshold to choose threadsize: */ + thrsz = FOUR_QUADS; + + OUT_PKT4(ring, REG_A5XX_SP_SP_CNTL, 1); + OUT_RING(ring, 0x00000000); /* SP_SP_CNTL */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_CONTROL_0_REG, 1); + OUT_RING(ring, A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(TWO_QUADS) | + A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE(thrsz) | + 0x00000880 /* XXX */); + + OUT_PKT4(ring, REG_A5XX_SP_CS_CTRL_REG0, 1); + OUT_RING(ring, A5XX_SP_CS_CTRL_REG0_THREADSIZE(thrsz) | + A5XX_SP_CS_CTRL_REG0_HALFREGFOOTPRINT(i->max_half_reg + 1) | + A5XX_SP_CS_CTRL_REG0_FULLREGFOOTPRINT(i->max_reg + 1) | + A5XX_SP_CS_CTRL_REG0_BRANCHSTACK(0x3) | // XXX need to figure this out somehow.. + 0x6 /* XXX */); + + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_CONFIG, 1); + OUT_RING(ring, A5XX_HLSQ_CS_CONFIG_CONSTOBJECTOFFSET(0) | + A5XX_HLSQ_CS_CONFIG_SHADEROBJOFFSET(0) | + A5XX_HLSQ_CS_CONFIG_ENABLED); + + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_CNTL, 1); + OUT_RING(ring, A5XX_HLSQ_CS_CNTL_INSTRLEN(v->instrlen) | + COND(v->has_ssbo, A5XX_HLSQ_CS_CNTL_SSBO_ENABLE)); + + OUT_PKT4(ring, REG_A5XX_SP_CS_CONFIG, 1); + OUT_RING(ring, A5XX_SP_CS_CONFIG_CONSTOBJECTOFFSET(0) | + A5XX_SP_CS_CONFIG_SHADEROBJOFFSET(0) | + A5XX_SP_CS_CONFIG_ENABLED); + + unsigned constlen = align(v->constlen, 4) / 4; + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_CONSTLEN, 2); + OUT_RING(ring, constlen); /* HLSQ_CS_CONSTLEN */ + OUT_RING(ring, v->instrlen); /* HLSQ_CS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_SP_CS_OBJ_START_LO, 2); + OUT_RELOC(ring, v->bo, 0, 0, 0); /* SP_CS_OBJ_START_LO/HI */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_UPDATE_CNTL, 1); + OUT_RING(ring, 0x1f00000); + + uint32_t local_invocation_id, work_group_id; + local_invocation_id = ir3_find_sysval_regid(v, SYSTEM_VALUE_LOCAL_INVOCATION_ID); + work_group_id = ir3_find_sysval_regid(v, SYSTEM_VALUE_WORK_GROUP_ID); + + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_CNTL_0, 2); + OUT_RING(ring, A5XX_HLSQ_CS_CNTL_0_WGIDCONSTID(work_group_id) | + A5XX_HLSQ_CS_CNTL_0_UNK0(regid(63, 0)) | + A5XX_HLSQ_CS_CNTL_0_UNK1(regid(63, 0)) | + A5XX_HLSQ_CS_CNTL_0_LOCALIDREGID(local_invocation_id)); + OUT_RING(ring, 0x1); /* HLSQ_CS_CNTL_1 */ + + fd5_emit_shader(ring, v); +} + +static void +fd5_launch_grid(struct fd_context *ctx, const struct pipe_grid_info *info) +{ + struct fd5_compute_stateobj *so = ctx->compute; + struct ir3_shader_key key = {0}; + struct ir3_shader_variant *v; + struct fd_ringbuffer *ring = ctx->batch->draw; + + if (info->indirect) + return; // TODO + + v = ir3_shader_variant(so->shader, key, &ctx->debug); + + if (ctx->dirty_shader[PIPE_SHADER_COMPUTE] & FD_DIRTY_SHADER_PROG) + cs_program_emit(ring, v); + + fd5_emit_cs_state(ctx, ring, v); + ir3_emit_cs_consts(v, ring, ctx, info); + + const unsigned *local_size = info->block; // v->shader->nir->info->cs.local_size; + const unsigned *num_groups = info->grid; + /* for some reason, mesa/st doesn't set info->work_dim, so just assume 3: */ + const unsigned work_dim = info->work_dim ? info->work_dim : 3; + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_NDRANGE_0, 7); + OUT_RING(ring, A5XX_HLSQ_CS_NDRANGE_0_KERNELDIM(work_dim) | + A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEX(local_size[0] - 1) | + A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEY(local_size[1] - 1) | + A5XX_HLSQ_CS_NDRANGE_0_LOCALSIZEZ(local_size[2] - 1)); + OUT_RING(ring, A5XX_HLSQ_CS_NDRANGE_1_SIZE_X(local_size[0] * num_groups[0])); + OUT_RING(ring, 0); /* HLSQ_CS_NDRANGE_2 */ + OUT_RING(ring, A5XX_HLSQ_CS_NDRANGE_3_SIZE_Y(local_size[1] * num_groups[1])); + OUT_RING(ring, 0); /* HLSQ_CS_NDRANGE_4 */ + OUT_RING(ring, A5XX_HLSQ_CS_NDRANGE_5_SIZE_Z(local_size[2] * num_groups[2])); + OUT_RING(ring, 0); /* HLSQ_CS_NDRANGE_6 */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_KERNEL_GROUP_X, 3); + OUT_RING(ring, 1); /* HLSQ_CS_KERNEL_GROUP_X */ + OUT_RING(ring, 1); /* HLSQ_CS_KERNEL_GROUP_Y */ + OUT_RING(ring, 1); /* HLSQ_CS_KERNEL_GROUP_Z */ + + OUT_PKT7(ring, CP_EXEC_CS, 4); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, CP_EXEC_CS_1_NGROUPS_X(info->grid[0])); + OUT_RING(ring, CP_EXEC_CS_2_NGROUPS_Y(info->grid[1])); + OUT_RING(ring, CP_EXEC_CS_3_NGROUPS_Z(info->grid[2])); +} + +void +fd5_compute_init(struct pipe_context *pctx) +{ + struct fd_context *ctx = fd_context(pctx); + ctx->launch_grid = fd5_launch_grid; + pctx->create_compute_state = fd5_create_compute_state; + pctx->delete_compute_state = fd5_delete_compute_state; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_compute.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_compute.h new file mode 100644 index 000000000..d5cc8b8a0 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_compute.h @@ -0,0 +1,34 @@ +/* + * Copyright (C) 2017 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_COMPUTE_H_ +#define FD5_COMPUTE_H_ + +#include "pipe/p_context.h" + +void fd5_compute_init(struct pipe_context *pctx); + +#endif /* FD5_COMPUTE_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_context.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_context.c new file mode 100644 index 000000000..3632cc522 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_context.c @@ -0,0 +1,120 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "freedreno_query_acc.h" + +#include "fd5_context.h" +#include "fd5_compute.h" +#include "fd5_blend.h" +#include "fd5_draw.h" +#include "fd5_emit.h" +#include "fd5_gmem.h" +#include "fd5_program.h" +#include "fd5_query.h" +#include "fd5_rasterizer.h" +#include "fd5_texture.h" +#include "fd5_zsa.h" + +static void +fd5_context_destroy(struct pipe_context *pctx) +{ + struct fd5_context *fd5_ctx = fd5_context(fd_context(pctx)); + + fd_bo_del(fd5_ctx->vs_pvt_mem); + fd_bo_del(fd5_ctx->fs_pvt_mem); + fd_bo_del(fd5_ctx->vsc_size_mem); + fd_bo_del(fd5_ctx->blit_mem); + + fd_context_cleanup_common_vbos(&fd5_ctx->base); + + u_upload_destroy(fd5_ctx->border_color_uploader); + + fd_context_destroy(pctx); +} + +static const uint8_t primtypes[] = { + [PIPE_PRIM_POINTS] = DI_PT_POINTLIST, + [PIPE_PRIM_LINES] = DI_PT_LINELIST, + [PIPE_PRIM_LINE_STRIP] = DI_PT_LINESTRIP, + [PIPE_PRIM_LINE_LOOP] = DI_PT_LINELOOP, + [PIPE_PRIM_TRIANGLES] = DI_PT_TRILIST, + [PIPE_PRIM_TRIANGLE_STRIP] = DI_PT_TRISTRIP, + [PIPE_PRIM_TRIANGLE_FAN] = DI_PT_TRIFAN, + [PIPE_PRIM_MAX] = DI_PT_RECTLIST, /* internal clear blits */ +}; + +struct pipe_context * +fd5_context_create(struct pipe_screen *pscreen, void *priv, unsigned flags) +{ + struct fd_screen *screen = fd_screen(pscreen); + struct fd5_context *fd5_ctx = CALLOC_STRUCT(fd5_context); + struct pipe_context *pctx; + + if (!fd5_ctx) + return NULL; + + pctx = &fd5_ctx->base.base; + + fd5_ctx->base.dev = fd_device_ref(screen->dev); + fd5_ctx->base.screen = fd_screen(pscreen); + + pctx->destroy = fd5_context_destroy; + pctx->create_blend_state = fd5_blend_state_create; + pctx->create_rasterizer_state = fd5_rasterizer_state_create; + pctx->create_depth_stencil_alpha_state = fd5_zsa_state_create; + + fd5_draw_init(pctx); + fd5_compute_init(pctx); + fd5_gmem_init(pctx); + fd5_texture_init(pctx); + fd5_prog_init(pctx); + fd5_emit_init(pctx); + + pctx = fd_context_init(&fd5_ctx->base, pscreen, primtypes, priv); + if (!pctx) + return NULL; + + fd5_ctx->vs_pvt_mem = fd_bo_new(screen->dev, 0x2000, + DRM_FREEDRENO_GEM_TYPE_KMEM); + + fd5_ctx->fs_pvt_mem = fd_bo_new(screen->dev, 0x2000, + DRM_FREEDRENO_GEM_TYPE_KMEM); + + fd5_ctx->vsc_size_mem = fd_bo_new(screen->dev, 0x1000, + DRM_FREEDRENO_GEM_TYPE_KMEM); + + fd5_ctx->blit_mem = fd_bo_new(screen->dev, 0x1000, + DRM_FREEDRENO_GEM_TYPE_KMEM); + + fd_context_setup_common_vbos(&fd5_ctx->base); + + fd5_query_context_init(pctx); + + fd5_ctx->border_color_uploader = u_upload_create(pctx, 4096, 0, + PIPE_USAGE_STREAM); + + return pctx; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_context.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_context.h new file mode 100644 index 000000000..f6de6ca2a --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_context.h @@ -0,0 +1,100 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_CONTEXT_H_ +#define FD5_CONTEXT_H_ + +#include "util/u_upload_mgr.h" + +#include "freedreno_drmif.h" + +#include "freedreno_context.h" + +#include "ir3_shader.h" + +struct fd5_context { + struct fd_context base; + + struct fd_bo *vs_pvt_mem, *fs_pvt_mem; + + /* This only needs to be 4 * num_of_pipes bytes (ie. 32 bytes). We + * could combine it with another allocation. + * + * (upper area used as scratch bo.. see fd5_query) + * + * XXX remove if unneeded after binning r/e.. + */ + struct fd_bo *vsc_size_mem; + + /* TODO not sure what this is for.. probably similar to + * CACHE_FLUSH_TS on kernel side, where value gets written + * to this address synchronized w/ 3d (ie. a way to + * synchronize when the CP is running far ahead) + */ + struct fd_bo *blit_mem; + + struct u_upload_mgr *border_color_uploader; + struct pipe_resource *border_color_buf; + + /* if *any* of bits are set in {v,f}saturate_{s,t,r} */ + bool vsaturate, fsaturate; + + /* bitmask of sampler which needs coords clamped for vertex + * shader: + */ + uint16_t vsaturate_s, vsaturate_t, vsaturate_r; + + /* bitmask of sampler which needs coords clamped for frag + * shader: + */ + uint16_t fsaturate_s, fsaturate_t, fsaturate_r; + + /* bitmask of samplers which need astc srgb workaround: */ + uint16_t vastc_srgb, fastc_srgb; + + /* some state changes require a different shader variant. Keep + * track of this so we know when we need to re-emit shader state + * due to variant change. See fixup_shader_state() + */ + struct ir3_shader_key last_key; + + /* number of active samples-passed queries: */ + int samples_passed_queries; + + /* cached state about current emitted shader program (3d): */ + unsigned max_loc; +}; + +static inline struct fd5_context * +fd5_context(struct fd_context *ctx) +{ + return (struct fd5_context *)ctx; +} + +struct pipe_context * +fd5_context_create(struct pipe_screen *pscreen, void *priv, unsigned flags); + +#endif /* FD5_CONTEXT_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_draw.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_draw.c new file mode 100644 index 000000000..d1f1d039b --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_draw.c @@ -0,0 +1,394 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" +#include "util/u_prim.h" + +#include "freedreno_state.h" +#include "freedreno_resource.h" + +#include "fd5_draw.h" +#include "fd5_context.h" +#include "fd5_emit.h" +#include "fd5_program.h" +#include "fd5_format.h" +#include "fd5_zsa.h" + + +static void +draw_impl(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct fd5_emit *emit, unsigned index_offset) +{ + const struct pipe_draw_info *info = emit->info; + enum pc_di_primtype primtype = ctx->primtypes[info->mode]; + + fd5_emit_state(ctx, ring, emit); + + if (emit->dirty & (FD_DIRTY_VTXBUF | FD_DIRTY_VTXSTATE)) + fd5_emit_vertex_bufs(ring, emit); + + OUT_PKT4(ring, REG_A5XX_VFD_INDEX_OFFSET, 2); + OUT_RING(ring, info->index_size ? info->index_bias : info->start); /* VFD_INDEX_OFFSET */ + OUT_RING(ring, info->start_instance); /* ??? UNKNOWN_2209 */ + + OUT_PKT4(ring, REG_A5XX_PC_RESTART_INDEX, 1); + OUT_RING(ring, info->primitive_restart ? /* PC_RESTART_INDEX */ + info->restart_index : 0xffffffff); + + fd5_emit_render_cntl(ctx, false, emit->key.binning_pass); + fd5_draw_emit(ctx->batch, ring, primtype, + emit->key.binning_pass ? IGNORE_VISIBILITY : USE_VISIBILITY, + info, index_offset); +} + +/* fixup dirty shader state in case some "unrelated" (from the state- + * tracker's perspective) state change causes us to switch to a + * different variant. + */ +static void +fixup_shader_state(struct fd_context *ctx, struct ir3_shader_key *key) +{ + struct fd5_context *fd5_ctx = fd5_context(ctx); + struct ir3_shader_key *last_key = &fd5_ctx->last_key; + + if (!ir3_shader_key_equal(last_key, key)) { + if (ir3_shader_key_changes_fs(last_key, key)) { + ctx->dirty_shader[PIPE_SHADER_FRAGMENT] |= FD_DIRTY_SHADER_PROG; + ctx->dirty |= FD_DIRTY_PROG; + } + + if (ir3_shader_key_changes_vs(last_key, key)) { + ctx->dirty_shader[PIPE_SHADER_VERTEX] |= FD_DIRTY_SHADER_PROG; + ctx->dirty |= FD_DIRTY_PROG; + } + + fd5_ctx->last_key = *key; + } +} + +static bool +fd5_draw_vbo(struct fd_context *ctx, const struct pipe_draw_info *info, + unsigned index_offset) +{ + struct fd5_context *fd5_ctx = fd5_context(ctx); + struct fd5_emit emit = { + .debug = &ctx->debug, + .vtx = &ctx->vtx, + .prog = &ctx->prog, + .info = info, + .key = { + .color_two_side = ctx->rasterizer->light_twoside, + .vclamp_color = ctx->rasterizer->clamp_vertex_color, + .fclamp_color = ctx->rasterizer->clamp_fragment_color, + .rasterflat = ctx->rasterizer->flatshade, + .half_precision = ctx->in_blit && + fd_half_precision(&ctx->batch->framebuffer), + .ucp_enables = ctx->rasterizer->clip_plane_enable, + .has_per_samp = (fd5_ctx->fsaturate || fd5_ctx->vsaturate || + fd5_ctx->fastc_srgb || fd5_ctx->vastc_srgb), + .vsaturate_s = fd5_ctx->vsaturate_s, + .vsaturate_t = fd5_ctx->vsaturate_t, + .vsaturate_r = fd5_ctx->vsaturate_r, + .fsaturate_s = fd5_ctx->fsaturate_s, + .fsaturate_t = fd5_ctx->fsaturate_t, + .fsaturate_r = fd5_ctx->fsaturate_r, + .vastc_srgb = fd5_ctx->vastc_srgb, + .fastc_srgb = fd5_ctx->fastc_srgb, + }, + .rasterflat = ctx->rasterizer->flatshade, + .sprite_coord_enable = ctx->rasterizer->sprite_coord_enable, + .sprite_coord_mode = ctx->rasterizer->sprite_coord_mode, + }; + + fixup_shader_state(ctx, &emit.key); + + unsigned dirty = ctx->dirty; + const struct ir3_shader_variant *vp = fd5_emit_get_vp(&emit); + const struct ir3_shader_variant *fp = fd5_emit_get_fp(&emit); + + /* do regular pass first, since that is more likely to fail compiling: */ + + if (!vp || !fp) + return false; + + /* figure out whether we need to disable LRZ write for binning + * pass using draw pass's fp: + */ + emit.no_lrz_write = fp->writes_pos || fp->has_kill; + + emit.key.binning_pass = false; + emit.dirty = dirty; + + draw_impl(ctx, ctx->batch->draw, &emit, index_offset); + + /* and now binning pass: */ + emit.key.binning_pass = true; + emit.dirty = dirty & ~(FD_DIRTY_BLEND); + emit.vp = NULL; /* we changed key so need to refetch vp */ + emit.fp = NULL; + draw_impl(ctx, ctx->batch->binning, &emit, index_offset); + + if (emit.streamout_mask) { + struct fd_ringbuffer *ring = ctx->batch->draw; + + for (unsigned i = 0; i < PIPE_MAX_SO_BUFFERS; i++) { + if (emit.streamout_mask & (1 << i)) { + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, FLUSH_SO_0 + i); + } + } + } + + fd_context_all_clean(ctx); + + return true; +} + +static bool is_z32(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT: + case PIPE_FORMAT_Z32_UNORM: + case PIPE_FORMAT_Z32_FLOAT: + return true; + default: + return false; + } +} + +static void +fd5_clear_lrz(struct fd_batch *batch, struct fd_resource *zsbuf, double depth) +{ + struct fd_ringbuffer *ring; + uint32_t clear = util_pack_z(PIPE_FORMAT_Z16_UNORM, depth); + + // TODO mid-frame clears (ie. app doing crazy stuff)?? Maybe worth + // splitting both clear and lrz clear out into their own rb's. And + // just throw away any draws prior to clear. (Anything not fullscreen + // clear, just fallback to generic path that treats it as a normal + // draw + + if (!batch->lrz_clear) { + batch->lrz_clear = fd_ringbuffer_new(batch->ctx->screen->pipe, 0x1000); + fd_ringbuffer_set_parent(batch->lrz_clear, batch->gmem); + } + + ring = batch->lrz_clear; + + OUT_WFI5(ring); + + OUT_PKT4(ring, REG_A5XX_RB_CCU_CNTL, 1); + OUT_RING(ring, 0x10000000); + + OUT_PKT4(ring, REG_A5XX_HLSQ_UPDATE_CNTL, 1); + OUT_RING(ring, 0x20fffff); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_CNTL, 1); + OUT_RING(ring, A5XX_GRAS_SU_CNTL_LINEHALFWIDTH(0.0)); + + OUT_PKT4(ring, REG_A5XX_GRAS_CNTL, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_GRAS_CL_CNTL, 1); + OUT_RING(ring, 0x00000181); + + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_CNTL, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_RB_MRT_BUF_INFO(0), 5); + OUT_RING(ring, A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT(RB5_R16_UNORM) | + A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE(TILE5_LINEAR) | + A5XX_RB_MRT_BUF_INFO_COLOR_SWAP(WZYX)); + OUT_RING(ring, A5XX_RB_MRT_PITCH(zsbuf->lrz_pitch * 2)); + OUT_RING(ring, A5XX_RB_MRT_ARRAY_PITCH(fd_bo_size(zsbuf->lrz))); + OUT_RELOCW(ring, zsbuf->lrz, 0x1000, 0, 0); + + OUT_PKT4(ring, REG_A5XX_RB_RENDER_CNTL, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_RB_DEST_MSAA_CNTL, 1); + OUT_RING(ring, A5XX_RB_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE)); + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_CNTL, 1); + OUT_RING(ring, A5XX_RB_BLIT_CNTL_BUF(BLIT_MRT0)); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_CNTL, 1); + OUT_RING(ring, A5XX_RB_CLEAR_CNTL_FAST_CLEAR | + A5XX_RB_CLEAR_CNTL_MASK(0xf)); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_COLOR_DW0, 1); + OUT_RING(ring, clear); /* RB_CLEAR_COLOR_DW0 */ + + OUT_PKT4(ring, REG_A5XX_VSC_RESOLVE_CNTL, 2); + OUT_RING(ring, A5XX_VSC_RESOLVE_CNTL_X(zsbuf->lrz_width) | + A5XX_VSC_RESOLVE_CNTL_Y(zsbuf->lrz_height)); + OUT_RING(ring, 0x00000000); // XXX UNKNOWN_0CDE + + OUT_PKT4(ring, REG_A5XX_RB_CNTL, 1); + OUT_RING(ring, A5XX_RB_CNTL_BYPASS); + + OUT_PKT4(ring, REG_A5XX_RB_RESOLVE_CNTL_1, 2); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_1_X(0) | + A5XX_RB_RESOLVE_CNTL_1_Y(0)); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_2_X(zsbuf->lrz_width - 1) | + A5XX_RB_RESOLVE_CNTL_2_Y(zsbuf->lrz_height - 1)); + + fd5_emit_blit(batch->ctx, ring); +} + +static bool +fd5_clear(struct fd_context *ctx, unsigned buffers, + const union pipe_color_union *color, double depth, unsigned stencil) +{ + struct fd_ringbuffer *ring = ctx->batch->draw; + struct pipe_framebuffer_state *pfb = &ctx->batch->framebuffer; + struct pipe_scissor_state *scissor = fd_context_get_scissor(ctx); + + if ((buffers & (PIPE_CLEAR_DEPTH | PIPE_CLEAR_STENCIL)) && + is_z32(pfb->zsbuf->format)) + return false; + + ctx->batch->max_scissor.minx = MIN2(ctx->batch->max_scissor.minx, scissor->minx); + ctx->batch->max_scissor.miny = MIN2(ctx->batch->max_scissor.miny, scissor->miny); + ctx->batch->max_scissor.maxx = MAX2(ctx->batch->max_scissor.maxx, scissor->maxx); + ctx->batch->max_scissor.maxy = MAX2(ctx->batch->max_scissor.maxy, scissor->maxy); + + fd5_emit_render_cntl(ctx, true, false); + + if (buffers & PIPE_CLEAR_COLOR) { + for (int i = 0; i < pfb->nr_cbufs; i++) { + union util_color uc = {0}; + + if (!pfb->cbufs[i]) + continue; + + if (!(buffers & (PIPE_CLEAR_COLOR0 << i))) + continue; + + enum pipe_format pfmt = pfb->cbufs[i]->format; + + // XXX I think RB_CLEAR_COLOR_DWn wants to take into account SWAP?? + union pipe_color_union swapped; + switch (fd5_pipe2swap(pfmt)) { + case WZYX: + swapped.ui[0] = color->ui[0]; + swapped.ui[1] = color->ui[1]; + swapped.ui[2] = color->ui[2]; + swapped.ui[3] = color->ui[3]; + break; + case WXYZ: + swapped.ui[2] = color->ui[0]; + swapped.ui[1] = color->ui[1]; + swapped.ui[0] = color->ui[2]; + swapped.ui[3] = color->ui[3]; + break; + case ZYXW: + swapped.ui[3] = color->ui[0]; + swapped.ui[0] = color->ui[1]; + swapped.ui[1] = color->ui[2]; + swapped.ui[2] = color->ui[3]; + break; + case XYZW: + swapped.ui[3] = color->ui[0]; + swapped.ui[2] = color->ui[1]; + swapped.ui[1] = color->ui[2]; + swapped.ui[0] = color->ui[3]; + break; + } + + if (util_format_is_pure_uint(pfmt)) { + util_format_write_4ui(pfmt, swapped.ui, 0, &uc, 0, 0, 0, 1, 1); + } else if (util_format_is_pure_sint(pfmt)) { + util_format_write_4i(pfmt, swapped.i, 0, &uc, 0, 0, 0, 1, 1); + } else { + util_pack_color(swapped.f, pfmt, &uc); + } + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_CNTL, 1); + OUT_RING(ring, A5XX_RB_BLIT_CNTL_BUF(BLIT_MRT0 + i)); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_CNTL, 1); + OUT_RING(ring, A5XX_RB_CLEAR_CNTL_FAST_CLEAR | + A5XX_RB_CLEAR_CNTL_MASK(0xf)); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_COLOR_DW0, 4); + OUT_RING(ring, uc.ui[0]); /* RB_CLEAR_COLOR_DW0 */ + OUT_RING(ring, uc.ui[1]); /* RB_CLEAR_COLOR_DW1 */ + OUT_RING(ring, uc.ui[2]); /* RB_CLEAR_COLOR_DW2 */ + OUT_RING(ring, uc.ui[3]); /* RB_CLEAR_COLOR_DW3 */ + + fd5_emit_blit(ctx, ring); + } + } + + if (pfb->zsbuf && (buffers & (PIPE_CLEAR_DEPTH | PIPE_CLEAR_STENCIL))) { + uint32_t clear = + util_pack_z_stencil(pfb->zsbuf->format, depth, stencil); + uint32_t mask = 0; + + if (buffers & PIPE_CLEAR_DEPTH) + mask |= 0x1; + + if (buffers & PIPE_CLEAR_STENCIL) + mask |= 0x2; + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_CNTL, 1); + OUT_RING(ring, A5XX_RB_BLIT_CNTL_BUF(BLIT_ZS)); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_CNTL, 1); + OUT_RING(ring, A5XX_RB_CLEAR_CNTL_FAST_CLEAR | + A5XX_RB_CLEAR_CNTL_MASK(mask)); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_COLOR_DW0, 1); + OUT_RING(ring, clear); /* RB_CLEAR_COLOR_DW0 */ + + fd5_emit_blit(ctx, ring); + + if (pfb->zsbuf && (buffers & PIPE_CLEAR_DEPTH)) { + struct fd_resource *zsbuf = fd_resource(pfb->zsbuf->texture); + if (zsbuf->lrz) { + zsbuf->lrz_valid = true; + fd5_clear_lrz(ctx->batch, zsbuf, depth); + } + } + } + + /* disable fast clear to not interfere w/ gmem->mem, etc.. */ + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_CNTL, 1); + OUT_RING(ring, 0x00000000); /* RB_CLEAR_CNTL */ + + return true; +} + +void +fd5_draw_init(struct pipe_context *pctx) +{ + struct fd_context *ctx = fd_context(pctx); + ctx->draw_vbo = fd5_draw_vbo; + ctx->clear = fd5_clear; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_draw.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_draw.h new file mode 100644 index 000000000..de210e456 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_draw.h @@ -0,0 +1,112 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_DRAW_H_ +#define FD5_DRAW_H_ + +#include "pipe/p_context.h" + +#include "freedreno_draw.h" + +/* some bits in common w/ a4xx: */ +#include "a4xx/fd4_draw.h" + +void fd5_draw_init(struct pipe_context *pctx); + +static inline void +fd5_draw(struct fd_batch *batch, struct fd_ringbuffer *ring, + enum pc_di_primtype primtype, + enum pc_di_vis_cull_mode vismode, + enum pc_di_src_sel src_sel, uint32_t count, + uint32_t instances, enum a4xx_index_size idx_type, + uint32_t idx_size, uint32_t idx_offset, + struct pipe_resource *idx_buffer) +{ + /* for debug after a lock up, write a unique counter value + * to scratch7 for each draw, to make it easier to match up + * register dumps to cmdstream. The combination of IB + * (scratch6) and DRAW is enough to "triangulate" the + * particular draw that caused lockup. + */ + emit_marker5(ring, 7); + + OUT_PKT7(ring, CP_DRAW_INDX_OFFSET, idx_buffer ? 7 : 3); + if (vismode == USE_VISIBILITY) { + /* leave vis mode blank for now, it will be patched up when + * we know if we are binning or not + */ + OUT_RINGP(ring, DRAW4(primtype, src_sel, idx_type, 0), + &batch->draw_patches); + } else { + OUT_RING(ring, DRAW4(primtype, src_sel, idx_type, vismode)); + } + OUT_RING(ring, instances); /* NumInstances */ + OUT_RING(ring, count); /* NumIndices */ + if (idx_buffer) { + OUT_RING(ring, 0x0); /* XXX */ + OUT_RELOC(ring, fd_resource(idx_buffer)->bo, idx_offset, 0, 0); + OUT_RING (ring, idx_size); + } + + emit_marker5(ring, 7); + + fd_reset_wfi(batch); +} + +static inline void +fd5_draw_emit(struct fd_batch *batch, struct fd_ringbuffer *ring, + enum pc_di_primtype primtype, + enum pc_di_vis_cull_mode vismode, + const struct pipe_draw_info *info, + unsigned index_offset) +{ + struct pipe_resource *idx_buffer = NULL; + enum a4xx_index_size idx_type; + enum pc_di_src_sel src_sel; + uint32_t idx_size, idx_offset; + + if (info->index_size) { + assert(!info->has_user_indices); + + idx_buffer = info->index.resource; + idx_type = fd4_size2indextype(info->index_size); + idx_size = info->index_size * info->count; + idx_offset = index_offset + info->start * info->index_size; + src_sel = DI_SRC_SEL_DMA; + } else { + idx_buffer = NULL; + idx_type = INDEX4_SIZE_32_BIT; + idx_size = 0; + idx_offset = 0; + src_sel = DI_SRC_SEL_AUTO_INDEX; + } + + fd5_draw(batch, ring, primtype, vismode, src_sel, + info->count, info->instance_count, + idx_type, idx_size, idx_offset, idx_buffer); +} + +#endif /* FD5_DRAW_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_emit.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_emit.c new file mode 100644 index 000000000..21931e9df --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_emit.c @@ -0,0 +1,1069 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" +#include "util/u_helpers.h" +#include "util/u_format.h" +#include "util/u_viewport.h" + +#include "freedreno_resource.h" +#include "freedreno_query_hw.h" + +#include "fd5_emit.h" +#include "fd5_blend.h" +#include "fd5_context.h" +#include "fd5_program.h" +#include "fd5_rasterizer.h" +#include "fd5_texture.h" +#include "fd5_format.h" +#include "fd5_zsa.h" + +/* regid: base const register + * prsc or dwords: buffer containing constant values + * sizedwords: size of const value buffer + */ +static void +fd5_emit_const(struct fd_ringbuffer *ring, enum shader_t type, + uint32_t regid, uint32_t offset, uint32_t sizedwords, + const uint32_t *dwords, struct pipe_resource *prsc) +{ + uint32_t i, sz; + enum a4xx_state_src src; + + debug_assert((regid % 4) == 0); + debug_assert((sizedwords % 4) == 0); + + if (prsc) { + sz = 0; + src = SS4_INDIRECT; + } else { + sz = sizedwords; + src = SS4_DIRECT; + } + + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + sz); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(regid/4) | + CP_LOAD_STATE4_0_STATE_SRC(src) | + CP_LOAD_STATE4_0_STATE_BLOCK(fd4_stage2shadersb(type)) | + CP_LOAD_STATE4_0_NUM_UNIT(sizedwords/4)); + if (prsc) { + struct fd_bo *bo = fd_resource(prsc)->bo; + OUT_RELOC(ring, bo, offset, + CP_LOAD_STATE4_1_STATE_TYPE(ST4_CONSTANTS), 0); + } else { + OUT_RING(ring, CP_LOAD_STATE4_1_EXT_SRC_ADDR(0) | + CP_LOAD_STATE4_1_STATE_TYPE(ST4_CONSTANTS)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + dwords = (uint32_t *)&((uint8_t *)dwords)[offset]; + } + for (i = 0; i < sz; i++) { + OUT_RING(ring, dwords[i]); + } +} + +static void +fd5_emit_const_bo(struct fd_ringbuffer *ring, enum shader_t type, boolean write, + uint32_t regid, uint32_t num, struct pipe_resource **prscs, uint32_t *offsets) +{ + uint32_t anum = align(num, 2); + uint32_t i; + + debug_assert((regid % 4) == 0); + + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + (2 * anum)); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(regid/4) | + CP_LOAD_STATE4_0_STATE_SRC(SS4_DIRECT) | + CP_LOAD_STATE4_0_STATE_BLOCK(fd4_stage2shadersb(type)) | + CP_LOAD_STATE4_0_NUM_UNIT(anum/2)); + OUT_RING(ring, CP_LOAD_STATE4_1_EXT_SRC_ADDR(0) | + CP_LOAD_STATE4_1_STATE_TYPE(ST4_CONSTANTS)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + + for (i = 0; i < num; i++) { + if (prscs[i]) { + if (write) { + OUT_RELOCW(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0); + } else { + OUT_RELOC(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0); + } + } else { + OUT_RING(ring, 0xbad00000 | (i << 16)); + OUT_RING(ring, 0xbad00000 | (i << 16)); + } + } + + for (; i < anum; i++) { + OUT_RING(ring, 0xffffffff); + OUT_RING(ring, 0xffffffff); + } +} + +/* Border color layout is diff from a4xx/a5xx.. if it turns out to be + * the same as a6xx then move this somewhere common ;-) + * + * Entry layout looks like (total size, 0x60 bytes): + */ + +struct PACKED bcolor_entry { + uint32_t fp32[4]; + uint16_t ui16[4]; + int16_t si16[4]; + uint16_t fp16[4]; + uint16_t rgb565; + uint16_t rgb5a1; + uint16_t rgba4; + uint8_t __pad0[2]; + uint8_t ui8[4]; + int8_t si8[4]; + uint32_t rgb10a2; + uint32_t z24; /* also s8? */ + uint8_t __pad1[32]; +}; + +#define FD5_BORDER_COLOR_SIZE 0x60 +#define FD5_BORDER_COLOR_UPLOAD_SIZE (2 * PIPE_MAX_SAMPLERS * FD5_BORDER_COLOR_SIZE) + +static void +setup_border_colors(struct fd_texture_stateobj *tex, struct bcolor_entry *entries) +{ + unsigned i, j; + STATIC_ASSERT(sizeof(struct bcolor_entry) == FD5_BORDER_COLOR_SIZE); + + for (i = 0; i < tex->num_samplers; i++) { + struct bcolor_entry *e = &entries[i]; + struct pipe_sampler_state *sampler = tex->samplers[i]; + union pipe_color_union *bc; + + if (!sampler) + continue; + + bc = &sampler->border_color; + + /* + * XXX HACK ALERT XXX + * + * The border colors need to be swizzled in a particular + * format-dependent order. Even though samplers don't know about + * formats, we can assume that with a GL state tracker, there's a + * 1:1 correspondence between sampler and texture. Take advantage + * of that knowledge. + */ + if ((i >= tex->num_textures) || !tex->textures[i]) + continue; + + const struct util_format_description *desc = + util_format_description(tex->textures[i]->format); + + e->rgb565 = 0; + e->rgb5a1 = 0; + e->rgba4 = 0; + e->rgb10a2 = 0; + e->z24 = 0; + + for (j = 0; j < 4; j++) { + int c = desc->swizzle[j]; + + if (c >= 4) + continue; + + if (desc->channel[c].pure_integer) { + uint16_t clamped; + switch (desc->channel[c].size) { + case 2: + assert(desc->channel[c].type == UTIL_FORMAT_TYPE_UNSIGNED); + clamped = CLAMP(bc->ui[j], 0, 0x3); + break; + case 8: + if (desc->channel[c].type == UTIL_FORMAT_TYPE_SIGNED) + clamped = CLAMP(bc->i[j], -128, 127); + else + clamped = CLAMP(bc->ui[j], 0, 255); + break; + case 10: + assert(desc->channel[c].type == UTIL_FORMAT_TYPE_UNSIGNED); + clamped = CLAMP(bc->ui[j], 0, 0x3ff); + break; + case 16: + if (desc->channel[c].type == UTIL_FORMAT_TYPE_SIGNED) + clamped = CLAMP(bc->i[j], -32768, 32767); + else + clamped = CLAMP(bc->ui[j], 0, 65535); + break; + default: + assert(!"Unexpected bit size"); + case 32: + clamped = 0; + break; + } + e->fp32[c] = bc->ui[j]; + e->fp16[c] = clamped; + } else { + float f = bc->f[j]; + float f_u = CLAMP(f, 0, 1); + float f_s = CLAMP(f, -1, 1); + + e->fp32[c] = fui(f); + e->fp16[c] = util_float_to_half(f); + e->ui16[c] = f_u * 0xffff; + e->si16[c] = f_s * 0x7fff; + e->ui8[c] = f_u * 0xff; + e->si8[c] = f_s * 0x7f; + if (c == 1) + e->rgb565 |= (int)(f_u * 0x3f) << 5; + else if (c < 3) + e->rgb565 |= (int)(f_u * 0x1f) << (c ? 11 : 0); + if (c == 3) + e->rgb5a1 |= (f_u > 0.5) ? 0x8000 : 0; + else + e->rgb5a1 |= (int)(f_u * 0x1f) << (c * 5); + if (c == 3) + e->rgb10a2 |= (int)(f_u * 0x3) << 30; + else + e->rgb10a2 |= (int)(f_u * 0x3ff) << (c * 10); + e->rgba4 |= (int)(f_u * 0xf) << (c * 4); + if (c == 0) + e->z24 = f_u * 0xffffff; + } + } + +#ifdef DEBUG + memset(&e->__pad0, 0, sizeof(e->__pad0)); + memset(&e->__pad1, 0, sizeof(e->__pad1)); +#endif + } +} + +static void +emit_border_color(struct fd_context *ctx, struct fd_ringbuffer *ring) +{ + struct fd5_context *fd5_ctx = fd5_context(ctx); + struct bcolor_entry *entries; + unsigned off; + void *ptr; + + STATIC_ASSERT(sizeof(struct bcolor_entry) == FD5_BORDER_COLOR_SIZE); + + u_upload_alloc(fd5_ctx->border_color_uploader, + 0, FD5_BORDER_COLOR_UPLOAD_SIZE, + FD5_BORDER_COLOR_UPLOAD_SIZE, &off, + &fd5_ctx->border_color_buf, + &ptr); + + entries = ptr; + + setup_border_colors(&ctx->tex[PIPE_SHADER_VERTEX], &entries[0]); + setup_border_colors(&ctx->tex[PIPE_SHADER_FRAGMENT], + &entries[ctx->tex[PIPE_SHADER_VERTEX].num_samplers]); + + OUT_PKT4(ring, REG_A5XX_TPL1_TP_BORDER_COLOR_BASE_ADDR_LO, 2); + OUT_RELOC(ring, fd_resource(fd5_ctx->border_color_buf)->bo, off, 0, 0); + + u_upload_unmap(fd5_ctx->border_color_uploader); +} + +static bool +emit_textures(struct fd_context *ctx, struct fd_ringbuffer *ring, + enum a4xx_state_block sb, struct fd_texture_stateobj *tex) +{ + bool needs_border = false; + unsigned bcolor_offset = (sb == SB4_FS_TEX) ? ctx->tex[PIPE_SHADER_VERTEX].num_samplers : 0; + unsigned i; + + if (tex->num_samplers > 0) { + /* output sampler state: */ + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + (4 * tex->num_samplers)); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(0) | + CP_LOAD_STATE4_0_STATE_SRC(SS4_DIRECT) | + CP_LOAD_STATE4_0_STATE_BLOCK(sb) | + CP_LOAD_STATE4_0_NUM_UNIT(tex->num_samplers)); + OUT_RING(ring, CP_LOAD_STATE4_1_STATE_TYPE(ST4_SHADER) | + CP_LOAD_STATE4_1_EXT_SRC_ADDR(0)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + for (i = 0; i < tex->num_samplers; i++) { + static const struct fd5_sampler_stateobj dummy_sampler = {}; + const struct fd5_sampler_stateobj *sampler = tex->samplers[i] ? + fd5_sampler_stateobj(tex->samplers[i]) : + &dummy_sampler; + OUT_RING(ring, sampler->texsamp0); + OUT_RING(ring, sampler->texsamp1); + OUT_RING(ring, sampler->texsamp2 | + A5XX_TEX_SAMP_2_BCOLOR_OFFSET(bcolor_offset)); + OUT_RING(ring, sampler->texsamp3); + + needs_border |= sampler->needs_border; + } + } + + if (tex->num_textures > 0) { + unsigned num_textures = tex->num_textures; + + /* emit texture state: */ + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + (12 * num_textures)); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(0) | + CP_LOAD_STATE4_0_STATE_SRC(SS4_DIRECT) | + CP_LOAD_STATE4_0_STATE_BLOCK(sb) | + CP_LOAD_STATE4_0_NUM_UNIT(num_textures)); + OUT_RING(ring, CP_LOAD_STATE4_1_STATE_TYPE(ST4_CONSTANTS) | + CP_LOAD_STATE4_1_EXT_SRC_ADDR(0)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + for (i = 0; i < tex->num_textures; i++) { + static const struct fd5_pipe_sampler_view dummy_view = {}; + const struct fd5_pipe_sampler_view *view = tex->textures[i] ? + fd5_pipe_sampler_view(tex->textures[i]) : + &dummy_view; + + OUT_RING(ring, view->texconst0); + OUT_RING(ring, view->texconst1); + OUT_RING(ring, view->texconst2); + OUT_RING(ring, view->texconst3); + if (view->base.texture) { + struct fd_resource *rsc = fd_resource(view->base.texture); + OUT_RELOC(ring, rsc->bo, view->offset, + (uint64_t)view->texconst5 << 32, 0); + } else { + OUT_RING(ring, 0x00000000); + OUT_RING(ring, view->texconst5); + } + OUT_RING(ring, view->texconst6); + OUT_RING(ring, view->texconst7); + OUT_RING(ring, view->texconst8); + OUT_RING(ring, view->texconst9); + OUT_RING(ring, view->texconst10); + OUT_RING(ring, view->texconst11); + } + } + + return needs_border; +} + +static void +emit_ssbos(struct fd_context *ctx, struct fd_ringbuffer *ring, + enum a4xx_state_block sb, struct fd_shaderbuf_stateobj *so) +{ + unsigned count = util_last_bit(so->enabled_mask); + + if (count == 0) + return; + + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + (4 * count)); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(0) | + CP_LOAD_STATE4_0_STATE_SRC(SS4_DIRECT) | + CP_LOAD_STATE4_0_STATE_BLOCK(sb) | + CP_LOAD_STATE4_0_NUM_UNIT(count)); + OUT_RING(ring, CP_LOAD_STATE4_1_STATE_TYPE(0) | + CP_LOAD_STATE4_1_EXT_SRC_ADDR(0)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + for (unsigned i = 0; i < count; i++) { + struct pipe_shader_buffer *buf = &so->sb[i]; + if (buf->buffer) { + struct fd_resource *rsc = fd_resource(buf->buffer); + OUT_RELOCW(ring, rsc->bo, 0, 0, 0); + } else { + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + } + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + } + + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + (2 * count)); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(0) | + CP_LOAD_STATE4_0_STATE_SRC(SS4_DIRECT) | + CP_LOAD_STATE4_0_STATE_BLOCK(sb) | + CP_LOAD_STATE4_0_NUM_UNIT(count)); + OUT_RING(ring, CP_LOAD_STATE4_1_STATE_TYPE(1) | + CP_LOAD_STATE4_1_EXT_SRC_ADDR(0)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + for (unsigned i = 0; i < count; i++) { + struct pipe_shader_buffer *buf = &so->sb[i]; + + // TODO maybe offset encoded somewhere here?? + OUT_RING(ring, (buf->buffer_size << 16)); + OUT_RING(ring, 0x00000000); + } + + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + (2 * count)); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(0) | + CP_LOAD_STATE4_0_STATE_SRC(SS4_DIRECT) | + CP_LOAD_STATE4_0_STATE_BLOCK(sb) | + CP_LOAD_STATE4_0_NUM_UNIT(count)); + OUT_RING(ring, CP_LOAD_STATE4_1_STATE_TYPE(2) | + CP_LOAD_STATE4_1_EXT_SRC_ADDR(0)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + for (unsigned i = 0; i < count; i++) { + struct pipe_shader_buffer *buf = &so->sb[i]; + if (buf->buffer) { + struct fd_resource *rsc = fd_resource(buf->buffer); + OUT_RELOCW(ring, rsc->bo, 0, 0, 0); + } else { + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + } + } +} + +void +fd5_emit_vertex_bufs(struct fd_ringbuffer *ring, struct fd5_emit *emit) +{ + int32_t i, j; + const struct fd_vertex_state *vtx = emit->vtx; + const struct ir3_shader_variant *vp = fd5_emit_get_vp(emit); + + for (i = 0, j = 0; i <= vp->inputs_count; i++) { + if (vp->inputs[i].sysval) + continue; + if (vp->inputs[i].compmask) { + struct pipe_vertex_element *elem = &vtx->vtx->pipe[i]; + const struct pipe_vertex_buffer *vb = + &vtx->vertexbuf.vb[elem->vertex_buffer_index]; + struct fd_resource *rsc = fd_resource(vb->buffer.resource); + enum pipe_format pfmt = elem->src_format; + enum a5xx_vtx_fmt fmt = fd5_pipe2vtx(pfmt); + bool isint = util_format_is_pure_integer(pfmt); + uint32_t off = vb->buffer_offset + elem->src_offset; + uint32_t size = fd_bo_size(rsc->bo) - off; + debug_assert(fmt != ~0); + + OUT_PKT4(ring, REG_A5XX_VFD_FETCH(j), 4); + OUT_RELOC(ring, rsc->bo, off, 0, 0); + OUT_RING(ring, size); /* VFD_FETCH[j].SIZE */ + OUT_RING(ring, vb->stride); /* VFD_FETCH[j].STRIDE */ + + OUT_PKT4(ring, REG_A5XX_VFD_DECODE(j), 2); + OUT_RING(ring, A5XX_VFD_DECODE_INSTR_IDX(j) | + A5XX_VFD_DECODE_INSTR_FORMAT(fmt) | + COND(elem->instance_divisor, A5XX_VFD_DECODE_INSTR_INSTANCED) | + A5XX_VFD_DECODE_INSTR_SWAP(fd5_pipe2swap(pfmt)) | + A5XX_VFD_DECODE_INSTR_UNK30 | + COND(!isint, A5XX_VFD_DECODE_INSTR_FLOAT)); + OUT_RING(ring, MAX2(1, elem->instance_divisor)); /* VFD_DECODE[j].STEP_RATE */ + + OUT_PKT4(ring, REG_A5XX_VFD_DEST_CNTL(j), 1); + OUT_RING(ring, A5XX_VFD_DEST_CNTL_INSTR_WRITEMASK(vp->inputs[i].compmask) | + A5XX_VFD_DEST_CNTL_INSTR_REGID(vp->inputs[i].regid)); + + j++; + } + } + + OUT_PKT4(ring, REG_A5XX_VFD_CONTROL_0, 1); + OUT_RING(ring, A5XX_VFD_CONTROL_0_VTXCNT(j)); +} + +void +fd5_emit_state(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct fd5_emit *emit) +{ + struct pipe_framebuffer_state *pfb = &ctx->batch->framebuffer; + const struct ir3_shader_variant *vp = fd5_emit_get_vp(emit); + const struct ir3_shader_variant *fp = fd5_emit_get_fp(emit); + const enum fd_dirty_3d_state dirty = emit->dirty; + bool needs_border = false; + + emit_marker5(ring, 5); + + if ((dirty & FD_DIRTY_FRAMEBUFFER) && !emit->key.binning_pass) { + unsigned char mrt_comp[A5XX_MAX_RENDER_TARGETS] = {0}; + + for (unsigned i = 0; i < A5XX_MAX_RENDER_TARGETS; i++) { + mrt_comp[i] = ((i < pfb->nr_cbufs) && pfb->cbufs[i]) ? 0xf : 0; + } + + OUT_PKT4(ring, REG_A5XX_RB_RENDER_COMPONENTS, 1); + OUT_RING(ring, A5XX_RB_RENDER_COMPONENTS_RT0(mrt_comp[0]) | + A5XX_RB_RENDER_COMPONENTS_RT1(mrt_comp[1]) | + A5XX_RB_RENDER_COMPONENTS_RT2(mrt_comp[2]) | + A5XX_RB_RENDER_COMPONENTS_RT3(mrt_comp[3]) | + A5XX_RB_RENDER_COMPONENTS_RT4(mrt_comp[4]) | + A5XX_RB_RENDER_COMPONENTS_RT5(mrt_comp[5]) | + A5XX_RB_RENDER_COMPONENTS_RT6(mrt_comp[6]) | + A5XX_RB_RENDER_COMPONENTS_RT7(mrt_comp[7])); + } + + if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_FRAMEBUFFER)) { + struct fd5_zsa_stateobj *zsa = fd5_zsa_stateobj(ctx->zsa); + uint32_t rb_alpha_control = zsa->rb_alpha_control; + + if (util_format_is_pure_integer(pipe_surface_format(pfb->cbufs[0]))) + rb_alpha_control &= ~A5XX_RB_ALPHA_CONTROL_ALPHA_TEST; + + OUT_PKT4(ring, REG_A5XX_RB_ALPHA_CONTROL, 1); + OUT_RING(ring, rb_alpha_control); + + OUT_PKT4(ring, REG_A5XX_RB_STENCIL_CONTROL, 1); + OUT_RING(ring, zsa->rb_stencil_control); + } + + if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_BLEND | FD_DIRTY_PROG)) { + struct fd5_blend_stateobj *blend = fd5_blend_stateobj(ctx->blend); + struct fd5_zsa_stateobj *zsa = fd5_zsa_stateobj(ctx->zsa); + + if (pfb->zsbuf) { + struct fd_resource *rsc = fd_resource(pfb->zsbuf->texture); + uint32_t gras_lrz_cntl = zsa->gras_lrz_cntl; + + if (emit->no_lrz_write || !rsc->lrz || !rsc->lrz_valid) + gras_lrz_cntl = 0; + else if (emit->key.binning_pass && blend->lrz_write && zsa->lrz_write) + gras_lrz_cntl |= A5XX_GRAS_LRZ_CNTL_LRZ_WRITE; + + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_CNTL, 1); + OUT_RING(ring, gras_lrz_cntl); + } + } + + if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_STENCIL_REF)) { + struct fd5_zsa_stateobj *zsa = fd5_zsa_stateobj(ctx->zsa); + struct pipe_stencil_ref *sr = &ctx->stencil_ref; + + OUT_PKT4(ring, REG_A5XX_RB_STENCILREFMASK, 2); + OUT_RING(ring, zsa->rb_stencilrefmask | + A5XX_RB_STENCILREFMASK_STENCILREF(sr->ref_value[0])); + OUT_RING(ring, zsa->rb_stencilrefmask_bf | + A5XX_RB_STENCILREFMASK_BF_STENCILREF(sr->ref_value[1])); + } + + if (dirty & (FD_DIRTY_ZSA | FD_DIRTY_RASTERIZER | FD_DIRTY_PROG)) { + struct fd5_zsa_stateobj *zsa = fd5_zsa_stateobj(ctx->zsa); + bool fragz = fp->has_kill | fp->writes_pos; + + OUT_PKT4(ring, REG_A5XX_RB_DEPTH_CNTL, 1); + OUT_RING(ring, zsa->rb_depth_cntl); + + OUT_PKT4(ring, REG_A5XX_RB_DEPTH_PLANE_CNTL, 1); + OUT_RING(ring, COND(fragz, A5XX_RB_DEPTH_PLANE_CNTL_FRAG_WRITES_Z) | + COND(fragz && fp->frag_coord, A5XX_RB_DEPTH_PLANE_CNTL_UNK1)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_DEPTH_PLANE_CNTL, 1); + OUT_RING(ring, COND(fragz, A5XX_GRAS_SU_DEPTH_PLANE_CNTL_FRAG_WRITES_Z) | + COND(fragz && fp->frag_coord, A5XX_GRAS_SU_DEPTH_PLANE_CNTL_UNK1)); + } + + if (dirty & FD_DIRTY_SCISSOR) { + struct pipe_scissor_state *scissor = fd_context_get_scissor(ctx); + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0, 2); + OUT_RING(ring, A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X(scissor->minx) | + A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y(scissor->miny)); + OUT_RING(ring, A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_X(scissor->maxx - 1) | + A5XX_GRAS_SC_SCREEN_SCISSOR_TL_0_Y(scissor->maxy - 1)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0, 2); + OUT_RING(ring, A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X(scissor->minx) | + A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y(scissor->miny)); + OUT_RING(ring, A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_X(scissor->maxx - 1) | + A5XX_GRAS_SC_VIEWPORT_SCISSOR_TL_0_Y(scissor->maxy - 1)); + + ctx->batch->max_scissor.minx = MIN2(ctx->batch->max_scissor.minx, scissor->minx); + ctx->batch->max_scissor.miny = MIN2(ctx->batch->max_scissor.miny, scissor->miny); + ctx->batch->max_scissor.maxx = MAX2(ctx->batch->max_scissor.maxx, scissor->maxx); + ctx->batch->max_scissor.maxy = MAX2(ctx->batch->max_scissor.maxy, scissor->maxy); + } + + if (dirty & FD_DIRTY_VIEWPORT) { + fd_wfi(ctx->batch, ring); + OUT_PKT4(ring, REG_A5XX_GRAS_CL_VPORT_XOFFSET_0, 6); + OUT_RING(ring, A5XX_GRAS_CL_VPORT_XOFFSET_0(ctx->viewport.translate[0])); + OUT_RING(ring, A5XX_GRAS_CL_VPORT_XSCALE_0(ctx->viewport.scale[0])); + OUT_RING(ring, A5XX_GRAS_CL_VPORT_YOFFSET_0(ctx->viewport.translate[1])); + OUT_RING(ring, A5XX_GRAS_CL_VPORT_YSCALE_0(ctx->viewport.scale[1])); + OUT_RING(ring, A5XX_GRAS_CL_VPORT_ZOFFSET_0(ctx->viewport.translate[2])); + OUT_RING(ring, A5XX_GRAS_CL_VPORT_ZSCALE_0(ctx->viewport.scale[2])); + } + + if (dirty & FD_DIRTY_PROG) + fd5_program_emit(ctx, ring, emit); + + if (dirty & FD_DIRTY_RASTERIZER) { + struct fd5_rasterizer_stateobj *rasterizer = + fd5_rasterizer_stateobj(ctx->rasterizer); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_CNTL, 1); + OUT_RING(ring, rasterizer->gras_su_cntl); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_POINT_MINMAX, 2); + OUT_RING(ring, rasterizer->gras_su_point_minmax); + OUT_RING(ring, rasterizer->gras_su_point_size); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_POLY_OFFSET_SCALE, 3); + OUT_RING(ring, rasterizer->gras_su_poly_offset_scale); + OUT_RING(ring, rasterizer->gras_su_poly_offset_offset); + OUT_RING(ring, rasterizer->gras_su_poly_offset_clamp); + + OUT_PKT4(ring, REG_A5XX_PC_RASTER_CNTL, 1); + OUT_RING(ring, rasterizer->pc_raster_cntl); + + OUT_PKT4(ring, REG_A5XX_GRAS_CL_CNTL, 1); + OUT_RING(ring, rasterizer->gras_cl_clip_cntl); + } + + /* note: must come after program emit.. because there is some overlap + * in registers, ex. PC_PRIMITIVE_CNTL and we rely on some cached + * values from fd5_program_emit() to avoid having to re-emit the prog + * every time rast state changes. + * + * Since the primitive restart state is not part of a tracked object, we + * re-emit this register every time. + */ + if (emit->info && ctx->rasterizer) { + struct fd5_rasterizer_stateobj *rasterizer = + fd5_rasterizer_stateobj(ctx->rasterizer); + unsigned max_loc = fd5_context(ctx)->max_loc; + + OUT_PKT4(ring, REG_A5XX_PC_PRIMITIVE_CNTL, 1); + OUT_RING(ring, rasterizer->pc_primitive_cntl | + A5XX_PC_PRIMITIVE_CNTL_STRIDE_IN_VPC(max_loc) | + COND(emit->info->primitive_restart && emit->info->index_size, + A5XX_PC_PRIMITIVE_CNTL_PRIMITIVE_RESTART)); + } + + if (dirty & (FD_DIRTY_FRAMEBUFFER | FD_DIRTY_RASTERIZER | FD_DIRTY_PROG)) { + uint32_t posz_regid = ir3_find_output_regid(fp, FRAG_RESULT_DEPTH); + unsigned nr = pfb->nr_cbufs; + + if (emit->key.binning_pass) + nr = 0; + else if (ctx->rasterizer->rasterizer_discard) + nr = 0; + + OUT_PKT4(ring, REG_A5XX_RB_FS_OUTPUT_CNTL, 1); + OUT_RING(ring, A5XX_RB_FS_OUTPUT_CNTL_MRT(nr) | + COND(fp->writes_pos, A5XX_RB_FS_OUTPUT_CNTL_FRAG_WRITES_Z)); + + OUT_PKT4(ring, REG_A5XX_SP_FS_OUTPUT_CNTL, 1); + OUT_RING(ring, A5XX_SP_FS_OUTPUT_CNTL_MRT(nr) | + A5XX_SP_FS_OUTPUT_CNTL_DEPTH_REGID(posz_regid) | + A5XX_SP_FS_OUTPUT_CNTL_SAMPLEMASK_REGID(regid(63, 0))); + } + + if (emit->prog == &ctx->prog) { /* evil hack to deal sanely with clear path */ + ir3_emit_vs_consts(vp, ring, ctx, emit->info); + if (!emit->key.binning_pass) + ir3_emit_fs_consts(fp, ring, ctx); + + struct pipe_stream_output_info *info = &vp->shader->stream_output; + if (info->num_outputs) { + struct fd_streamout_stateobj *so = &ctx->streamout; + + for (unsigned i = 0; i < so->num_targets; i++) { + struct pipe_stream_output_target *target = so->targets[i]; + + if (!target) + continue; + + unsigned offset = (so->offsets[i] * info->stride[i] * 4) + + target->buffer_offset; + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_BASE_LO(i), 3); + /* VPC_SO[i].BUFFER_BASE_LO: */ + OUT_RELOCW(ring, fd_resource(target->buffer)->bo, 0, 0, 0); + OUT_RING(ring, target->buffer_size + offset); + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_OFFSET(i), 3); + OUT_RING(ring, offset); + /* VPC_SO[i].FLUSH_BASE_LO/HI: */ + // TODO just give hw a dummy addr for now.. we should + // be using this an then CP_MEM_TO_REG to set the + // VPC_SO[i].BUFFER_OFFSET for the next draw.. + OUT_RELOCW(ring, fd5_context(ctx)->blit_mem, 0x100, 0, 0); + + emit->streamout_mask |= (1 << i); + } + } + } + + if ((dirty & FD_DIRTY_BLEND)) { + struct fd5_blend_stateobj *blend = fd5_blend_stateobj(ctx->blend); + uint32_t i; + + for (i = 0; i < A5XX_MAX_RENDER_TARGETS; i++) { + enum pipe_format format = pipe_surface_format(pfb->cbufs[i]); + bool is_int = util_format_is_pure_integer(format); + bool has_alpha = util_format_has_alpha(format); + uint32_t control = blend->rb_mrt[i].control; + uint32_t blend_control = blend->rb_mrt[i].blend_control_alpha; + + if (is_int) { + control &= A5XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK; + control |= A5XX_RB_MRT_CONTROL_ROP_CODE(ROP_COPY); + } + + if (has_alpha) { + blend_control |= blend->rb_mrt[i].blend_control_rgb; + } else { + blend_control |= blend->rb_mrt[i].blend_control_no_alpha_rgb; + control &= ~A5XX_RB_MRT_CONTROL_BLEND2; + } + + OUT_PKT4(ring, REG_A5XX_RB_MRT_CONTROL(i), 1); + OUT_RING(ring, control); + + OUT_PKT4(ring, REG_A5XX_RB_MRT_BLEND_CONTROL(i), 1); + OUT_RING(ring, blend_control); + } + + OUT_PKT4(ring, REG_A5XX_RB_BLEND_CNTL, 1); + OUT_RING(ring, blend->rb_blend_cntl | + A5XX_RB_BLEND_CNTL_SAMPLE_MASK(0xffff)); + + OUT_PKT4(ring, REG_A5XX_SP_BLEND_CNTL, 1); + OUT_RING(ring, blend->sp_blend_cntl); + } + + if (dirty & FD_DIRTY_BLEND_COLOR) { + struct pipe_blend_color *bcolor = &ctx->blend_color; + + OUT_PKT4(ring, REG_A5XX_RB_BLEND_RED, 8); + OUT_RING(ring, A5XX_RB_BLEND_RED_FLOAT(bcolor->color[0]) | + A5XX_RB_BLEND_RED_UINT(bcolor->color[0] * 0xff) | + A5XX_RB_BLEND_RED_SINT(bcolor->color[0] * 0x7f)); + OUT_RING(ring, A5XX_RB_BLEND_RED_F32(bcolor->color[0])); + OUT_RING(ring, A5XX_RB_BLEND_GREEN_FLOAT(bcolor->color[1]) | + A5XX_RB_BLEND_GREEN_UINT(bcolor->color[1] * 0xff) | + A5XX_RB_BLEND_GREEN_SINT(bcolor->color[1] * 0x7f)); + OUT_RING(ring, A5XX_RB_BLEND_RED_F32(bcolor->color[1])); + OUT_RING(ring, A5XX_RB_BLEND_BLUE_FLOAT(bcolor->color[2]) | + A5XX_RB_BLEND_BLUE_UINT(bcolor->color[2] * 0xff) | + A5XX_RB_BLEND_BLUE_SINT(bcolor->color[2] * 0x7f)); + OUT_RING(ring, A5XX_RB_BLEND_BLUE_F32(bcolor->color[2])); + OUT_RING(ring, A5XX_RB_BLEND_ALPHA_FLOAT(bcolor->color[3]) | + A5XX_RB_BLEND_ALPHA_UINT(bcolor->color[3] * 0xff) | + A5XX_RB_BLEND_ALPHA_SINT(bcolor->color[3] * 0x7f)); + OUT_RING(ring, A5XX_RB_BLEND_ALPHA_F32(bcolor->color[3])); + } + + if (ctx->dirty_shader[PIPE_SHADER_VERTEX] & FD_DIRTY_SHADER_TEX) { + needs_border |= emit_textures(ctx, ring, SB4_VS_TEX, + &ctx->tex[PIPE_SHADER_VERTEX]); + OUT_PKT4(ring, REG_A5XX_TPL1_VS_TEX_COUNT, 1); + OUT_RING(ring, ctx->tex[PIPE_SHADER_VERTEX].num_textures); + } + + if (ctx->dirty_shader[PIPE_SHADER_FRAGMENT] & FD_DIRTY_SHADER_TEX) { + needs_border |= emit_textures(ctx, ring, SB4_FS_TEX, + &ctx->tex[PIPE_SHADER_FRAGMENT]); + OUT_PKT4(ring, REG_A5XX_TPL1_FS_TEX_COUNT, 1); + OUT_RING(ring, ctx->tex[PIPE_SHADER_FRAGMENT].num_textures); + } + + OUT_PKT4(ring, REG_A5XX_TPL1_CS_TEX_COUNT, 1); + OUT_RING(ring, 0); + + if (needs_border) + emit_border_color(ctx, ring); + + if (ctx->dirty_shader[PIPE_SHADER_FRAGMENT] & FD_DIRTY_SHADER_SSBO) + emit_ssbos(ctx, ring, SB4_SSBO, &ctx->shaderbuf[PIPE_SHADER_FRAGMENT]); +} + +void +fd5_emit_cs_state(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct ir3_shader_variant *cp) +{ + enum fd_dirty_shader_state dirty = ctx->dirty_shader[PIPE_SHADER_COMPUTE]; + + if (dirty & FD_DIRTY_SHADER_TEX) { + bool needs_border = false; + needs_border |= emit_textures(ctx, ring, SB4_CS_TEX, + &ctx->tex[PIPE_SHADER_COMPUTE]); + + if (needs_border) + emit_border_color(ctx, ring); + + OUT_PKT4(ring, REG_A5XX_TPL1_VS_TEX_COUNT, 1); + OUT_RING(ring, 0); + + OUT_PKT4(ring, REG_A5XX_TPL1_HS_TEX_COUNT, 1); + OUT_RING(ring, 0); + + OUT_PKT4(ring, REG_A5XX_TPL1_DS_TEX_COUNT, 1); + OUT_RING(ring, 0); + + OUT_PKT4(ring, REG_A5XX_TPL1_GS_TEX_COUNT, 1); + OUT_RING(ring, 0); + + OUT_PKT4(ring, REG_A5XX_TPL1_FS_TEX_COUNT, 1); + OUT_RING(ring, 0); + + OUT_PKT4(ring, REG_A5XX_TPL1_CS_TEX_COUNT, 1); + OUT_RING(ring, ctx->tex[PIPE_SHADER_COMPUTE].num_textures); + } + + if (dirty & FD_DIRTY_SHADER_SSBO) + emit_ssbos(ctx, ring, SB4_CS_SSBO, &ctx->shaderbuf[PIPE_SHADER_COMPUTE]); +} + +/* emit setup at begin of new cmdstream buffer (don't rely on previous + * state, there could have been a context switch between ioctls): + */ +void +fd5_emit_restore(struct fd_batch *batch, struct fd_ringbuffer *ring) +{ + struct fd_context *ctx = batch->ctx; + + fd5_set_render_mode(ctx, ring, BYPASS); + fd5_cache_flush(batch, ring); + + OUT_PKT4(ring, REG_A5XX_HLSQ_UPDATE_CNTL, 1); + OUT_RING(ring, 0xfffff); + +/* +t7 opcode: CP_PERFCOUNTER_ACTION (50) (4 dwords) +0000000500024048: 70d08003 00000000 001c5000 00000005 +t7 opcode: CP_PERFCOUNTER_ACTION (50) (4 dwords) +0000000500024058: 70d08003 00000010 001c7000 00000005 + +t7 opcode: CP_WAIT_FOR_IDLE (26) (1 dwords) +0000000500024068: 70268000 +*/ + + OUT_PKT4(ring, REG_A5XX_PC_RESTART_INDEX, 1); + OUT_RING(ring, 0xffffffff); + + OUT_PKT4(ring, REG_A5XX_PC_RASTER_CNTL, 1); + OUT_RING(ring, 0x00000012); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_POINT_MINMAX, 2); + OUT_RING(ring, A5XX_GRAS_SU_POINT_MINMAX_MIN(1.0) | + A5XX_GRAS_SU_POINT_MINMAX_MAX(4092.0)); + OUT_RING(ring, A5XX_GRAS_SU_POINT_SIZE(0.5)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_CONSERVATIVE_RAS_CNTL, 1); + OUT_RING(ring, 0x00000000); /* GRAS_SU_CONSERVATIVE_RAS_CNTL */ + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_SCREEN_SCISSOR_CNTL, 1); + OUT_RING(ring, 0x00000000); /* GRAS_SC_SCREEN_SCISSOR_CNTL */ + + OUT_PKT4(ring, REG_A5XX_SP_VS_CONFIG_MAX_CONST, 1); + OUT_RING(ring, 0); /* SP_VS_CONFIG_MAX_CONST */ + + OUT_PKT4(ring, REG_A5XX_SP_FS_CONFIG_MAX_CONST, 1); + OUT_RING(ring, 0); /* SP_FS_CONFIG_MAX_CONST */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E292, 2); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E292 */ + OUT_RING(ring, 0x00000000); /* UNKNOWN_E293 */ + + OUT_PKT4(ring, REG_A5XX_RB_MODE_CNTL, 1); + OUT_RING(ring, 0x00000044); /* RB_MODE_CNTL */ + + OUT_PKT4(ring, REG_A5XX_RB_DBG_ECO_CNTL, 1); + OUT_RING(ring, 0x00100000); /* RB_DBG_ECO_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VFD_MODE_CNTL, 1); + OUT_RING(ring, 0x00000000); /* VFD_MODE_CNTL */ + + OUT_PKT4(ring, REG_A5XX_PC_MODE_CNTL, 1); + OUT_RING(ring, 0x0000001f); /* PC_MODE_CNTL */ + + OUT_PKT4(ring, REG_A5XX_SP_MODE_CNTL, 1); + OUT_RING(ring, 0x0000001e); /* SP_MODE_CNTL */ + + OUT_PKT4(ring, REG_A5XX_SP_DBG_ECO_CNTL, 1); + OUT_RING(ring, 0x40000800); /* SP_DBG_ECO_CNTL */ + + OUT_PKT4(ring, REG_A5XX_TPL1_MODE_CNTL, 1); + OUT_RING(ring, 0x00000544); /* TPL1_MODE_CNTL */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_TIMEOUT_THRESHOLD_0, 2); + OUT_RING(ring, 0x00000080); /* HLSQ_TIMEOUT_THRESHOLD_0 */ + OUT_RING(ring, 0x00000000); /* HLSQ_TIMEOUT_THRESHOLD_1 */ + + OUT_PKT4(ring, REG_A5XX_VPC_DBG_ECO_CNTL, 1); + OUT_RING(ring, 0x00000400); /* VPC_DBG_ECO_CNTL */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_MODE_CNTL, 1); + OUT_RING(ring, 0x00000001); /* HLSQ_MODE_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VPC_MODE_CNTL, 1); + OUT_RING(ring, 0x00000000); /* VPC_MODE_CNTL */ + + /* we don't use this yet.. probably best to disable.. */ + OUT_PKT7(ring, CP_SET_DRAW_STATE, 3); + OUT_RING(ring, CP_SET_DRAW_STATE__0_COUNT(0) | + CP_SET_DRAW_STATE__0_DISABLE_ALL_GROUPS | + CP_SET_DRAW_STATE__0_GROUP_ID(0)); + OUT_RING(ring, CP_SET_DRAW_STATE__1_ADDR_LO(0)); + OUT_RING(ring, CP_SET_DRAW_STATE__2_ADDR_HI(0)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_CONSERVATIVE_RAS_CNTL, 1); + OUT_RING(ring, 0x00000000); /* GRAS_SU_CONSERVATIVE_RAS_CNTL */ + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_BIN_CNTL, 1); + OUT_RING(ring, 0x00000000); /* GRAS_SC_BIN_CNTL */ + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_BIN_CNTL, 1); + OUT_RING(ring, 0x00000000); /* GRAS_SC_BIN_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VPC_FS_PRIMITIVEID_CNTL, 1); + OUT_RING(ring, 0x000000ff); /* VPC_FS_PRIMITIVEID_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VPC_SO_OVERRIDE, 1); + OUT_RING(ring, A5XX_VPC_SO_OVERRIDE_SO_DISABLE); + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_BASE_LO(0), 3); + OUT_RING(ring, 0x00000000); /* VPC_SO_BUFFER_BASE_LO_0 */ + OUT_RING(ring, 0x00000000); /* VPC_SO_BUFFER_BASE_HI_0 */ + OUT_RING(ring, 0x00000000); /* VPC_SO_BUFFER_SIZE_0 */ + + OUT_PKT4(ring, REG_A5XX_VPC_SO_FLUSH_BASE_LO(0), 2); + OUT_RING(ring, 0x00000000); /* VPC_SO_FLUSH_BASE_LO_0 */ + OUT_RING(ring, 0x00000000); /* VPC_SO_FLUSH_BASE_HI_0 */ + + OUT_PKT4(ring, REG_A5XX_PC_GS_PARAM, 1); + OUT_RING(ring, 0x00000000); /* PC_GS_PARAM */ + + OUT_PKT4(ring, REG_A5XX_PC_HS_PARAM, 1); + OUT_RING(ring, 0x00000000); /* PC_HS_PARAM */ + + OUT_PKT4(ring, REG_A5XX_TPL1_TP_FS_ROTATION_CNTL, 1); + OUT_RING(ring, 0x00000000); /* TPL1_TP_FS_ROTATION_CNTL */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E001, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E001 */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E004, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E004 */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E093, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E093 */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E29A, 1); + OUT_RING(ring, 0x00ffff00); /* UNKNOWN_E29A */ + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUF_CNTL, 1); + OUT_RING(ring, 0x00000000); /* VPC_SO_BUF_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_OFFSET(0), 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E2AB */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E389, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E389 */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E38D, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E38D */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E5AB, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E5AB */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E5C2, 1); + OUT_RING(ring, 0x00000000); /* UNKNOWN_E5C2 */ + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_BASE_LO(1), 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_OFFSET(1), 6); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_OFFSET(2), 6); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_VPC_SO_BUFFER_OFFSET(3), 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E5DB, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E600, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E640, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_TPL1_VS_TEX_COUNT, 4); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_TPL1_FS_TEX_COUNT, 2); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E7C0, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E7C5, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E7CA, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E7CF, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E7D4, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_E7D9, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_RB_CLEAR_CNTL, 1); + OUT_RING(ring, 0x00000000); +} + +static void +fd5_emit_ib(struct fd_ringbuffer *ring, struct fd_ringbuffer *target) +{ + __OUT_IB5(ring, target); +} + +void +fd5_emit_init(struct pipe_context *pctx) +{ + struct fd_context *ctx = fd_context(pctx); + ctx->emit_const = fd5_emit_const; + ctx->emit_const_bo = fd5_emit_const_bo; + ctx->emit_ib = fd5_emit_ib; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_emit.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_emit.h new file mode 100644 index 000000000..2d8a0fd09 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_emit.h @@ -0,0 +1,199 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_EMIT_H +#define FD5_EMIT_H + +#include "pipe/p_context.h" + +#include "freedreno_context.h" +#include "fd5_context.h" +#include "fd5_format.h" +#include "fd5_program.h" +#include "ir3_shader.h" + +struct fd_ringbuffer; + +/* grouped together emit-state for prog/vertex/state emit: */ +struct fd5_emit { + struct pipe_debug_callback *debug; + const struct fd_vertex_state *vtx; + const struct fd_program_stateobj *prog; + const struct pipe_draw_info *info; + struct ir3_shader_key key; + enum fd_dirty_3d_state dirty; + + uint32_t sprite_coord_enable; /* bitmask */ + bool sprite_coord_mode; + bool rasterflat; + bool no_decode_srgb; + + /* in binning pass, we don't have real frag shader, so we + * don't know if real draw disqualifies lrz write. So just + * figure that out up-front and stash it in the emit. + */ + bool no_lrz_write; + + /* cached to avoid repeated lookups of same variants: */ + const struct ir3_shader_variant *vp, *fp; + /* TODO: other shader stages.. */ + + unsigned streamout_mask; +}; + +static inline enum a5xx_color_fmt fd5_emit_format(struct pipe_surface *surf) +{ + if (!surf) + return 0; + return fd5_pipe2color(surf->format); +} + +static inline const struct ir3_shader_variant * +fd5_emit_get_vp(struct fd5_emit *emit) +{ + if (!emit->vp) { + struct fd5_shader_stateobj *so = emit->prog->vp; + emit->vp = ir3_shader_variant(so->shader, emit->key, emit->debug); + } + return emit->vp; +} + +static inline const struct ir3_shader_variant * +fd5_emit_get_fp(struct fd5_emit *emit) +{ + if (!emit->fp) { + if (emit->key.binning_pass) { + /* use dummy stateobj to simplify binning vs non-binning: */ + static const struct ir3_shader_variant binning_fp = {}; + emit->fp = &binning_fp; + } else { + struct fd5_shader_stateobj *so = emit->prog->fp; + emit->fp = ir3_shader_variant(so->shader, emit->key, emit->debug); + } + } + return emit->fp; +} + +static inline void +fd5_cache_flush(struct fd_batch *batch, struct fd_ringbuffer *ring) +{ + fd_reset_wfi(batch); + OUT_PKT4(ring, REG_A5XX_UCHE_CACHE_INVALIDATE_MIN_LO, 5); + OUT_RING(ring, 0x00000000); /* UCHE_CACHE_INVALIDATE_MIN_LO */ + OUT_RING(ring, 0x00000000); /* UCHE_CACHE_INVALIDATE_MIN_HI */ + OUT_RING(ring, 0x00000000); /* UCHE_CACHE_INVALIDATE_MAX_LO */ + OUT_RING(ring, 0x00000000); /* UCHE_CACHE_INVALIDATE_MAX_HI */ + OUT_RING(ring, 0x00000012); /* UCHE_CACHE_INVALIDATE */ + fd_wfi(batch, ring); +} + +static inline void +fd5_set_render_mode(struct fd_context *ctx, struct fd_ringbuffer *ring, + enum render_mode_cmd mode) +{ + /* TODO add preemption support, gmem bypass, etc */ + emit_marker5(ring, 7); + OUT_PKT7(ring, CP_SET_RENDER_MODE, 5); + OUT_RING(ring, CP_SET_RENDER_MODE_0_MODE(mode)); + OUT_RING(ring, 0x00000000); /* ADDR_LO */ + OUT_RING(ring, 0x00000000); /* ADDR_HI */ + OUT_RING(ring, COND(mode == GMEM, CP_SET_RENDER_MODE_3_GMEM_ENABLE) | + COND(mode == BINNING, CP_SET_RENDER_MODE_3_VSC_ENABLE)); + OUT_RING(ring, 0x00000000); + emit_marker5(ring, 7); +} + +static inline void +fd5_emit_blit(struct fd_context *ctx, struct fd_ringbuffer *ring) +{ + struct fd5_context *fd5_ctx = fd5_context(ctx); + + emit_marker5(ring, 7); + + OUT_PKT7(ring, CP_EVENT_WRITE, 4); + OUT_RING(ring, CP_EVENT_WRITE_0_EVENT(BLIT)); + OUT_RELOCW(ring, fd5_ctx->blit_mem, 0, 0, 0); /* ADDR_LO/HI */ + OUT_RING(ring, 0x00000000); + + emit_marker5(ring, 7); +} + +static inline void +fd5_emit_render_cntl(struct fd_context *ctx, bool blit, bool binning) +{ + struct fd_ringbuffer *ring = binning ? ctx->batch->binning : ctx->batch->draw; + + /* TODO eventually this partially depends on the pfb state, ie. + * which of the cbuf(s)/zsbuf has an UBWC flag buffer.. that part + * we could probably cache and just regenerate if framebuffer + * state is dirty (or something like that).. + * + * Other bits seem to depend on query state, like if samples-passed + * query is active. + */ + bool samples_passed = (fd5_context(ctx)->samples_passed_queries > 0); + OUT_PKT4(ring, REG_A5XX_RB_RENDER_CNTL, 1); + OUT_RING(ring, 0x00000000 | /* RB_RENDER_CNTL */ + COND(binning, A5XX_RB_RENDER_CNTL_BINNING_PASS) | + COND(binning, A5XX_RB_RENDER_CNTL_DISABLE_COLOR_PIPE) | + COND(samples_passed, A5XX_RB_RENDER_CNTL_SAMPLES_PASSED) | + COND(!blit, 0x8)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_CNTL, 1); + OUT_RING(ring, 0x00000008 | /* GRAS_SC_CNTL */ + COND(binning, A5XX_GRAS_SC_CNTL_BINNING_PASS) | + COND(samples_passed, A5XX_GRAS_SC_CNTL_SAMPLES_PASSED)); +} + +static inline void +fd5_emit_lrz_flush(struct fd_ringbuffer *ring) +{ + /* TODO I think the extra writes to GRAS_LRZ_CNTL are probably + * a workaround and not needed on all a5xx. + */ + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_CNTL, 1); + OUT_RING(ring, A5XX_GRAS_LRZ_CNTL_ENABLE); + + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, LRZ_FLUSH); + + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_CNTL, 1); + OUT_RING(ring, 0x0); +} + +void fd5_emit_vertex_bufs(struct fd_ringbuffer *ring, struct fd5_emit *emit); + +void fd5_emit_state(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct fd5_emit *emit); + +void fd5_emit_cs_state(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct ir3_shader_variant *cp); + +void fd5_emit_restore(struct fd_batch *batch, struct fd_ringbuffer *ring); + +void fd5_emit_init(struct pipe_context *pctx); + +#endif /* FD5_EMIT_H */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_format.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_format.c new file mode 100644 index 000000000..ae5cc839f --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_format.c @@ -0,0 +1,445 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_defines.h" +#include "util/u_format.h" + +#include "fd5_format.h" + + +/* Specifies the table of all the formats and their features. Also supplies + * the helpers that look up various data in those tables. + */ + +struct fd5_format { + enum a5xx_vtx_fmt vtx; + enum a5xx_tex_fmt tex; + enum a5xx_color_fmt rb; + enum a3xx_color_swap swap; + boolean present; +}; + +#define RB5_NONE ~0 + +/* vertex + texture */ +#define VT(pipe, fmt, rbfmt, swapfmt) \ + [PIPE_FORMAT_ ## pipe] = { \ + .present = 1, \ + .vtx = VFMT5_ ## fmt, \ + .tex = TFMT5_ ## fmt, \ + .rb = RB5_ ## rbfmt, \ + .swap = swapfmt \ + } + +/* texture-only */ +#define _T(pipe, fmt, rbfmt, swapfmt) \ + [PIPE_FORMAT_ ## pipe] = { \ + .present = 1, \ + .vtx = ~0, \ + .tex = TFMT5_ ## fmt, \ + .rb = RB5_ ## rbfmt, \ + .swap = swapfmt \ + } + +/* vertex-only */ +#define V_(pipe, fmt, rbfmt, swapfmt) \ + [PIPE_FORMAT_ ## pipe] = { \ + .present = 1, \ + .vtx = VFMT5_ ## fmt, \ + .tex = ~0, \ + .rb = RB5_ ## rbfmt, \ + .swap = swapfmt \ + } + +static struct fd5_format formats[PIPE_FORMAT_COUNT] = { + /* 8-bit */ + VT(R8_UNORM, 8_UNORM, R8_UNORM, WZYX), + VT(R8_SNORM, 8_SNORM, R8_SNORM, WZYX), + VT(R8_UINT, 8_UINT, R8_UINT, WZYX), + VT(R8_SINT, 8_SINT, R8_SINT, WZYX), + V_(R8_USCALED, 8_UINT, NONE, WZYX), + V_(R8_SSCALED, 8_UINT, NONE, WZYX), + + _T(A8_UNORM, 8_UNORM, A8_UNORM, WZYX), + _T(L8_UNORM, 8_UNORM, R8_UNORM, WZYX), + _T(I8_UNORM, 8_UNORM, NONE, WZYX), + + _T(A8_UINT, 8_UINT, NONE, WZYX), + _T(A8_SINT, 8_SINT, NONE, WZYX), + _T(L8_UINT, 8_UINT, NONE, WZYX), + _T(L8_SINT, 8_SINT, NONE, WZYX), + _T(I8_UINT, 8_UINT, NONE, WZYX), + _T(I8_SINT, 8_SINT, NONE, WZYX), + + _T(S8_UINT, 8_UINT, R8_UNORM, WZYX), + + /* 16-bit */ + VT(R16_UNORM, 16_UNORM, R16_UNORM, WZYX), + VT(R16_SNORM, 16_SNORM, R16_SNORM, WZYX), + VT(R16_UINT, 16_UINT, R16_UINT, WZYX), + VT(R16_SINT, 16_SINT, R16_SINT, WZYX), + V_(R16_USCALED, 16_UINT, NONE, WZYX), + V_(R16_SSCALED, 16_UINT, NONE, WZYX), + VT(R16_FLOAT, 16_FLOAT, R16_FLOAT, WZYX), + + _T(A16_UNORM, 16_UNORM, NONE, WZYX), + _T(A16_SNORM, 16_SNORM, NONE, WZYX), + _T(A16_UINT, 16_UINT, NONE, WZYX), + _T(A16_SINT, 16_SINT, NONE, WZYX), + _T(L16_UNORM, 16_UNORM, NONE, WZYX), + _T(L16_SNORM, 16_SNORM, NONE, WZYX), + _T(L16_UINT, 16_UINT, NONE, WZYX), + _T(L16_SINT, 16_SINT, NONE, WZYX), + _T(I16_UNORM, 16_UNORM, NONE, WZYX), + _T(I16_SNORM, 16_SNORM, NONE, WZYX), + _T(I16_UINT, 16_UINT, NONE, WZYX), + _T(I16_SINT, 16_SINT, NONE, WZYX), + + VT(R8G8_UNORM, 8_8_UNORM, R8G8_UNORM, WZYX), + VT(R8G8_SNORM, 8_8_SNORM, R8G8_SNORM, WZYX), + VT(R8G8_UINT, 8_8_UINT, R8G8_UINT, WZYX), + VT(R8G8_SINT, 8_8_SINT, R8G8_SINT, WZYX), + V_(R8G8_USCALED, 8_8_UINT, NONE, WZYX), + V_(R8G8_SSCALED, 8_8_SINT, NONE, WZYX), + + _T(L8A8_UINT, 8_8_UINT, NONE, WZYX), + _T(L8A8_SINT, 8_8_SINT, NONE, WZYX), + + _T(B5G6R5_UNORM, 5_6_5_UNORM, R5G6B5_UNORM, WXYZ), + _T(B5G5R5A1_UNORM, 5_5_5_1_UNORM, R5G5B5A1_UNORM, WXYZ), + _T(B5G5R5X1_UNORM, 5_5_5_1_UNORM, R5G5B5A1_UNORM, WXYZ), + _T(B4G4R4A4_UNORM, 4_4_4_4_UNORM, R4G4B4A4_UNORM, WXYZ), + + /* 24-bit */ + V_(R8G8B8_UNORM, 8_8_8_UNORM, NONE, WZYX), + V_(R8G8B8_SNORM, 8_8_8_SNORM, NONE, WZYX), + V_(R8G8B8_UINT, 8_8_8_UINT, NONE, WZYX), + V_(R8G8B8_SINT, 8_8_8_SINT, NONE, WZYX), + V_(R8G8B8_USCALED, 8_8_8_UINT, NONE, WZYX), + V_(R8G8B8_SSCALED, 8_8_8_SINT, NONE, WZYX), + + /* 32-bit */ + VT(R32_UINT, 32_UINT, R32_UINT, WZYX), + VT(R32_SINT, 32_SINT, R32_SINT, WZYX), + V_(R32_USCALED, 32_UINT, NONE, WZYX), + V_(R32_SSCALED, 32_UINT, NONE, WZYX), + VT(R32_FLOAT, 32_FLOAT, R32_FLOAT,WZYX), + V_(R32_FIXED, 32_FIXED, NONE, WZYX), + + _T(A32_UINT, 32_UINT, NONE, WZYX), + _T(A32_SINT, 32_SINT, NONE, WZYX), + _T(L32_UINT, 32_UINT, NONE, WZYX), + _T(L32_SINT, 32_SINT, NONE, WZYX), + _T(I32_UINT, 32_UINT, NONE, WZYX), + _T(I32_SINT, 32_SINT, NONE, WZYX), + + VT(R16G16_UNORM, 16_16_UNORM, R16G16_UNORM, WZYX), + VT(R16G16_SNORM, 16_16_SNORM, R16G16_SNORM, WZYX), + VT(R16G16_UINT, 16_16_UINT, R16G16_UINT, WZYX), + VT(R16G16_SINT, 16_16_SINT, R16G16_SINT, WZYX), + VT(R16G16_USCALED, 16_16_UINT, NONE, WZYX), + VT(R16G16_SSCALED, 16_16_SINT, NONE, WZYX), + VT(R16G16_FLOAT, 16_16_FLOAT, R16G16_FLOAT, WZYX), + + _T(L16A16_UNORM, 16_16_UNORM, NONE, WZYX), + _T(L16A16_SNORM, 16_16_SNORM, NONE, WZYX), + _T(L16A16_UINT, 16_16_UINT, NONE, WZYX), + _T(L16A16_SINT, 16_16_SINT, NONE, WZYX), + + VT(R8G8B8A8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WZYX), + _T(R8G8B8X8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WZYX), + _T(R8G8B8A8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WZYX), + _T(R8G8B8X8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WZYX), + VT(R8G8B8A8_SNORM, 8_8_8_8_SNORM, R8G8B8A8_SNORM, WZYX), + VT(R8G8B8A8_UINT, 8_8_8_8_UINT, R8G8B8A8_UINT, WZYX), + VT(R8G8B8A8_SINT, 8_8_8_8_SINT, R8G8B8A8_SINT, WZYX), + V_(R8G8B8A8_USCALED, 8_8_8_8_UINT, NONE, WZYX), + V_(R8G8B8A8_SSCALED, 8_8_8_8_SINT, NONE, WZYX), + + VT(B8G8R8A8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WXYZ), + _T(B8G8R8X8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WXYZ), + VT(B8G8R8A8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WXYZ), + _T(B8G8R8X8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, WXYZ), + + VT(A8B8G8R8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, XYZW), + _T(X8B8G8R8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, XYZW), + _T(A8B8G8R8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, XYZW), + _T(X8B8G8R8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, XYZW), + + VT(A8R8G8B8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, ZYXW), + _T(X8R8G8B8_UNORM, 8_8_8_8_UNORM, R8G8B8A8_UNORM, ZYXW), + _T(A8R8G8B8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, ZYXW), + _T(X8R8G8B8_SRGB, 8_8_8_8_UNORM, R8G8B8A8_UNORM, ZYXW), + + VT(R10G10B10A2_UNORM, 10_10_10_2_UNORM, R10G10B10A2_UNORM, WZYX), + VT(B10G10R10A2_UNORM, 10_10_10_2_UNORM, R10G10B10A2_UNORM, WXYZ), + _T(B10G10R10X2_UNORM, 10_10_10_2_UNORM, R10G10B10A2_UNORM, WXYZ), + V_(R10G10B10A2_SNORM, 10_10_10_2_SNORM, NONE, WZYX), + V_(B10G10R10A2_SNORM, 10_10_10_2_SNORM, NONE, WXYZ), + VT(R10G10B10A2_UINT, 10_10_10_2_UINT, R10G10B10A2_UINT, WZYX), + VT(B10G10R10A2_UINT, 10_10_10_2_UINT, R10G10B10A2_UINT, WXYZ), + V_(R10G10B10A2_USCALED, 10_10_10_2_UINT, NONE, WZYX), + V_(B10G10R10A2_USCALED, 10_10_10_2_UINT, NONE, WXYZ), + V_(R10G10B10A2_SSCALED, 10_10_10_2_SINT, NONE, WZYX), + V_(B10G10R10A2_SSCALED, 10_10_10_2_SINT, NONE, WXYZ), + + VT(R11G11B10_FLOAT, 11_11_10_FLOAT, R11G11B10_FLOAT, WZYX), + _T(R9G9B9E5_FLOAT, 9_9_9_E5_FLOAT, NONE, WZYX), + + _T(Z24X8_UNORM, X8Z24_UNORM, R8G8B8A8_UNORM, WZYX), + _T(Z24_UNORM_S8_UINT, X8Z24_UNORM, R8G8B8A8_UNORM, WZYX), + _T(Z32_FLOAT, 32_FLOAT, R8G8B8A8_UNORM, WZYX), + _T(Z32_FLOAT_S8X24_UINT, 32_FLOAT,R8G8B8A8_UNORM, WZYX), + + /* 48-bit */ + V_(R16G16B16_UNORM, 16_16_16_UNORM, NONE, WZYX), + V_(R16G16B16_SNORM, 16_16_16_SNORM, NONE, WZYX), + V_(R16G16B16_UINT, 16_16_16_UINT, NONE, WZYX), + V_(R16G16B16_SINT, 16_16_16_SINT, NONE, WZYX), + V_(R16G16B16_USCALED, 16_16_16_UINT, NONE, WZYX), + V_(R16G16B16_SSCALED, 16_16_16_SINT, NONE, WZYX), + V_(R16G16B16_FLOAT, 16_16_16_FLOAT, NONE, WZYX), + + /* 64-bit */ + VT(R16G16B16A16_UNORM, 16_16_16_16_UNORM, R16G16B16A16_UNORM, WZYX), + VT(R16G16B16X16_UNORM, 16_16_16_16_UNORM, R16G16B16A16_UNORM, WZYX), + VT(R16G16B16A16_SNORM, 16_16_16_16_SNORM, R16G16B16A16_SNORM, WZYX), + VT(R16G16B16X16_SNORM, 16_16_16_16_SNORM, R16G16B16A16_SNORM, WZYX), + VT(R16G16B16A16_UINT, 16_16_16_16_UINT, R16G16B16A16_UINT, WZYX), + VT(R16G16B16X16_UINT, 16_16_16_16_UINT, R16G16B16A16_UINT, WZYX), + VT(R16G16B16A16_SINT, 16_16_16_16_SINT, R16G16B16A16_SINT, WZYX), + VT(R16G16B16X16_SINT, 16_16_16_16_SINT, R16G16B16A16_SINT, WZYX), + VT(R16G16B16A16_USCALED, 16_16_16_16_UINT, NONE, WZYX), + VT(R16G16B16A16_SSCALED, 16_16_16_16_SINT, NONE, WZYX), + VT(R16G16B16A16_FLOAT, 16_16_16_16_FLOAT, R16G16B16A16_FLOAT, WZYX), + VT(R16G16B16X16_FLOAT, 16_16_16_16_FLOAT, R16G16B16A16_FLOAT, WZYX), + + VT(R32G32_UINT, 32_32_UINT, R32G32_UINT, WZYX), + VT(R32G32_SINT, 32_32_SINT, R32G32_SINT, WZYX), + V_(R32G32_USCALED, 32_32_UINT, NONE, WZYX), + V_(R32G32_SSCALED, 32_32_SINT, NONE, WZYX), + VT(R32G32_FLOAT, 32_32_FLOAT, R32G32_FLOAT,WZYX), + V_(R32G32_FIXED, 32_32_FIXED, NONE, WZYX), + + _T(L32A32_UINT, 32_32_UINT, NONE, WZYX), + _T(L32A32_SINT, 32_32_SINT, NONE, WZYX), + + /* 96-bit */ + VT(R32G32B32_UINT, 32_32_32_UINT, NONE, WZYX), + VT(R32G32B32_SINT, 32_32_32_SINT, NONE, WZYX), + V_(R32G32B32_USCALED, 32_32_32_UINT, NONE, WZYX), + V_(R32G32B32_SSCALED, 32_32_32_SINT, NONE, WZYX), + VT(R32G32B32_FLOAT, 32_32_32_FLOAT, NONE, WZYX), + V_(R32G32B32_FIXED, 32_32_32_FIXED, NONE, WZYX), + + /* 128-bit */ + VT(R32G32B32A32_UINT, 32_32_32_32_UINT, R32G32B32A32_UINT, WZYX), + _T(R32G32B32X32_UINT, 32_32_32_32_UINT, R32G32B32A32_UINT, WZYX), + VT(R32G32B32A32_SINT, 32_32_32_32_SINT, R32G32B32A32_SINT, WZYX), + _T(R32G32B32X32_SINT, 32_32_32_32_SINT, R32G32B32A32_SINT, WZYX), + V_(R32G32B32A32_USCALED, 32_32_32_32_UINT, NONE, WZYX), + V_(R32G32B32A32_SSCALED, 32_32_32_32_SINT, NONE, WZYX), + VT(R32G32B32A32_FLOAT, 32_32_32_32_FLOAT, R32G32B32A32_FLOAT, WZYX), + _T(R32G32B32X32_FLOAT, 32_32_32_32_FLOAT, R32G32B32A32_FLOAT, WZYX), + V_(R32G32B32A32_FIXED, 32_32_32_32_FIXED, NONE, WZYX), + + /* compressed */ + _T(ETC1_RGB8, ETC1, NONE, WZYX), + _T(ETC2_RGB8, ETC2_RGB8, NONE, WZYX), + _T(ETC2_SRGB8, ETC2_RGB8, NONE, WZYX), + _T(ETC2_RGB8A1, ETC2_RGB8A1, NONE, WZYX), + _T(ETC2_SRGB8A1, ETC2_RGB8A1, NONE, WZYX), + _T(ETC2_RGBA8, ETC2_RGBA8, NONE, WZYX), + _T(ETC2_SRGBA8, ETC2_RGBA8, NONE, WZYX), + _T(ETC2_R11_UNORM, ETC2_R11_UNORM, NONE, WZYX), + _T(ETC2_R11_SNORM, ETC2_R11_SNORM, NONE, WZYX), + _T(ETC2_RG11_UNORM, ETC2_RG11_UNORM, NONE, WZYX), + _T(ETC2_RG11_SNORM, ETC2_RG11_SNORM, NONE, WZYX), + + _T(DXT1_RGB, DXT1, NONE, WZYX), + _T(DXT1_SRGB, DXT1, NONE, WZYX), + _T(DXT1_RGBA, DXT1, NONE, WZYX), + _T(DXT1_SRGBA, DXT1, NONE, WZYX), + _T(DXT3_RGBA, DXT3, NONE, WZYX), + _T(DXT3_SRGBA, DXT3, NONE, WZYX), + _T(DXT5_RGBA, DXT5, NONE, WZYX), + _T(DXT5_SRGBA, DXT5, NONE, WZYX), + + _T(BPTC_RGBA_UNORM, BPTC, NONE, WZYX), + _T(BPTC_SRGBA, BPTC, NONE, WZYX), + _T(BPTC_RGB_FLOAT, BPTC_FLOAT, NONE, WZYX), + _T(BPTC_RGB_UFLOAT, BPTC_UFLOAT, NONE, WZYX), + + _T(RGTC1_UNORM, RGTC1_UNORM, NONE, WZYX), + _T(RGTC1_SNORM, RGTC1_SNORM, NONE, WZYX), + _T(RGTC2_UNORM, RGTC2_UNORM, NONE, WZYX), + _T(RGTC2_SNORM, RGTC2_SNORM, NONE, WZYX), + _T(LATC1_UNORM, RGTC1_UNORM, NONE, WZYX), + _T(LATC1_SNORM, RGTC1_SNORM, NONE, WZYX), + _T(LATC2_UNORM, RGTC2_UNORM, NONE, WZYX), + _T(LATC2_SNORM, RGTC2_SNORM, NONE, WZYX), + + _T(ASTC_4x4, ASTC_4x4, NONE, WZYX), + _T(ASTC_5x4, ASTC_5x4, NONE, WZYX), + _T(ASTC_5x5, ASTC_5x5, NONE, WZYX), + _T(ASTC_6x5, ASTC_6x5, NONE, WZYX), + _T(ASTC_6x6, ASTC_6x6, NONE, WZYX), + _T(ASTC_8x5, ASTC_8x5, NONE, WZYX), + _T(ASTC_8x6, ASTC_8x6, NONE, WZYX), + _T(ASTC_8x8, ASTC_8x8, NONE, WZYX), + _T(ASTC_10x5, ASTC_10x5, NONE, WZYX), + _T(ASTC_10x6, ASTC_10x6, NONE, WZYX), + _T(ASTC_10x8, ASTC_10x8, NONE, WZYX), + _T(ASTC_10x10, ASTC_10x10, NONE, WZYX), + _T(ASTC_12x10, ASTC_12x10, NONE, WZYX), + _T(ASTC_12x12, ASTC_12x12, NONE, WZYX), + + _T(ASTC_4x4_SRGB, ASTC_4x4, NONE, WZYX), + _T(ASTC_5x4_SRGB, ASTC_5x4, NONE, WZYX), + _T(ASTC_5x5_SRGB, ASTC_5x5, NONE, WZYX), + _T(ASTC_6x5_SRGB, ASTC_6x5, NONE, WZYX), + _T(ASTC_6x6_SRGB, ASTC_6x6, NONE, WZYX), + _T(ASTC_8x5_SRGB, ASTC_8x5, NONE, WZYX), + _T(ASTC_8x6_SRGB, ASTC_8x6, NONE, WZYX), + _T(ASTC_8x8_SRGB, ASTC_8x8, NONE, WZYX), + _T(ASTC_10x5_SRGB, ASTC_10x5, NONE, WZYX), + _T(ASTC_10x6_SRGB, ASTC_10x6, NONE, WZYX), + _T(ASTC_10x8_SRGB, ASTC_10x8, NONE, WZYX), + _T(ASTC_10x10_SRGB, ASTC_10x10, NONE, WZYX), + _T(ASTC_12x10_SRGB, ASTC_12x10, NONE, WZYX), + _T(ASTC_12x12_SRGB, ASTC_12x12, NONE, WZYX), +}; + +/* convert pipe format to vertex buffer format: */ +enum a5xx_vtx_fmt +fd5_pipe2vtx(enum pipe_format format) +{ + if (!formats[format].present) + return ~0; + return formats[format].vtx; +} + +/* convert pipe format to texture sampler format: */ +enum a5xx_tex_fmt +fd5_pipe2tex(enum pipe_format format) +{ + if (!formats[format].present) + return ~0; + return formats[format].tex; +} + +/* convert pipe format to MRT / copydest format used for render-target: */ +enum a5xx_color_fmt +fd5_pipe2color(enum pipe_format format) +{ + if (!formats[format].present) + return ~0; + return formats[format].rb; +} + +enum a3xx_color_swap +fd5_pipe2swap(enum pipe_format format) +{ + if (!formats[format].present) + return WZYX; + return formats[format].swap; +} + +// XXX possibly same as a4xx.. +enum a5xx_tex_fetchsize +fd5_pipe2fetchsize(enum pipe_format format) +{ + if (format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT) + format = PIPE_FORMAT_Z32_FLOAT; + + if (util_format_description(format)->layout == UTIL_FORMAT_LAYOUT_ASTC) + return TFETCH5_16_BYTE; + + switch (util_format_get_blocksizebits(format) / util_format_get_blockwidth(format)) { + case 8: return TFETCH5_1_BYTE; + case 16: return TFETCH5_2_BYTE; + case 32: return TFETCH5_4_BYTE; + case 64: return TFETCH5_8_BYTE; + case 96: return TFETCH5_1_BYTE; /* Does this matter? */ + case 128: return TFETCH5_16_BYTE; + default: + debug_printf("Unknown block size for format %s: %d\n", + util_format_name(format), + util_format_get_blocksizebits(format)); + return TFETCH5_1_BYTE; + } +} + +enum a5xx_depth_format +fd5_pipe2depth(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_Z16_UNORM: + return DEPTH5_16; + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_Z24_UNORM_S8_UINT: + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8_UINT_Z24_UNORM: + return DEPTH5_24_8; + case PIPE_FORMAT_Z32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT: + return DEPTH5_32; + default: + return ~0; + } +} + +static inline enum a5xx_tex_swiz +tex_swiz(unsigned swiz) +{ + switch (swiz) { + default: + case PIPE_SWIZZLE_X: return A5XX_TEX_X; + case PIPE_SWIZZLE_Y: return A5XX_TEX_Y; + case PIPE_SWIZZLE_Z: return A5XX_TEX_Z; + case PIPE_SWIZZLE_W: return A5XX_TEX_W; + case PIPE_SWIZZLE_0: return A5XX_TEX_ZERO; + case PIPE_SWIZZLE_1: return A5XX_TEX_ONE; + } +} + +uint32_t +fd5_tex_swiz(enum pipe_format format, unsigned swizzle_r, unsigned swizzle_g, + unsigned swizzle_b, unsigned swizzle_a) +{ + const struct util_format_description *desc = + util_format_description(format); + unsigned char swiz[4] = { + swizzle_r, swizzle_g, swizzle_b, swizzle_a, + }, rswiz[4]; + + util_format_compose_swizzles(desc->swizzle, swiz, rswiz); + + return A5XX_TEX_CONST_0_SWIZ_X(tex_swiz(rswiz[0])) | + A5XX_TEX_CONST_0_SWIZ_Y(tex_swiz(rswiz[1])) | + A5XX_TEX_CONST_0_SWIZ_Z(tex_swiz(rswiz[2])) | + A5XX_TEX_CONST_0_SWIZ_W(tex_swiz(rswiz[3])); +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_format.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_format.h new file mode 100644 index 000000000..b052aa529 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_format.h @@ -0,0 +1,44 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_UTIL_H_ +#define FD5_UTIL_H_ + +#include "freedreno_util.h" + +#include "a5xx.xml.h" + +enum a5xx_vtx_fmt fd5_pipe2vtx(enum pipe_format format); +enum a5xx_tex_fmt fd5_pipe2tex(enum pipe_format format); +enum a5xx_color_fmt fd5_pipe2color(enum pipe_format format); +enum a3xx_color_swap fd5_pipe2swap(enum pipe_format format); +enum a5xx_tex_fetchsize fd5_pipe2fetchsize(enum pipe_format format); +enum a5xx_depth_format fd5_pipe2depth(enum pipe_format format); + +uint32_t fd5_tex_swiz(enum pipe_format format, unsigned swizzle_r, + unsigned swizzle_g, unsigned swizzle_b, unsigned swizzle_a); + +#endif /* FD5_UTIL_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_gmem.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_gmem.c new file mode 100644 index 000000000..c623b572b --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_gmem.c @@ -0,0 +1,774 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" +#include "util/u_inlines.h" +#include "util/u_format.h" + +#include "freedreno_draw.h" +#include "freedreno_state.h" +#include "freedreno_resource.h" + +#include "fd5_gmem.h" +#include "fd5_context.h" +#include "fd5_draw.h" +#include "fd5_emit.h" +#include "fd5_program.h" +#include "fd5_format.h" +#include "fd5_zsa.h" + +static void +emit_mrt(struct fd_ringbuffer *ring, unsigned nr_bufs, + struct pipe_surface **bufs, struct fd_gmem_stateobj *gmem) +{ + enum a5xx_tile_mode tile_mode; + unsigned i; + + if (gmem) { + tile_mode = TILE5_2; + } else { + tile_mode = TILE5_LINEAR; + } + + for (i = 0; i < A5XX_MAX_RENDER_TARGETS; i++) { + enum a5xx_color_fmt format = 0; + enum a3xx_color_swap swap = WZYX; + bool srgb = false, sint = false, uint = false; + struct fd_resource *rsc = NULL; + struct fd_resource_slice *slice = NULL; + uint32_t stride = 0; + uint32_t size = 0; + uint32_t base = 0; + uint32_t offset = 0; + + if ((i < nr_bufs) && bufs[i]) { + struct pipe_surface *psurf = bufs[i]; + enum pipe_format pformat = psurf->format; + + rsc = fd_resource(psurf->texture); + + slice = fd_resource_slice(rsc, psurf->u.tex.level); + format = fd5_pipe2color(pformat); + swap = fd5_pipe2swap(pformat); + srgb = util_format_is_srgb(pformat); + sint = util_format_is_pure_sint(pformat); + uint = util_format_is_pure_uint(pformat); + + debug_assert(psurf->u.tex.first_layer == psurf->u.tex.last_layer); + + offset = fd_resource_offset(rsc, psurf->u.tex.level, + psurf->u.tex.first_layer); + + if (gmem) { + stride = gmem->bin_w * rsc->cpp; + size = stride * gmem->bin_h; + base = gmem->cbuf_base[i]; + } else { + stride = slice->pitch * rsc->cpp; + size = slice->size0; + } + } + + OUT_PKT4(ring, REG_A5XX_RB_MRT_BUF_INFO(i), 5); + OUT_RING(ring, A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT(format) | + A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE(tile_mode) | + A5XX_RB_MRT_BUF_INFO_COLOR_SWAP(swap) | + COND(gmem, 0x800) | /* XXX 0x1000 for RECTLIST clear, 0x0 for BLIT.. */ + COND(srgb, A5XX_RB_MRT_BUF_INFO_COLOR_SRGB)); + OUT_RING(ring, A5XX_RB_MRT_PITCH(stride)); + OUT_RING(ring, A5XX_RB_MRT_ARRAY_PITCH(size)); + if (gmem || (i >= nr_bufs) || !bufs[i]) { + OUT_RING(ring, base); /* RB_MRT[i].BASE_LO */ + OUT_RING(ring, 0x00000000); /* RB_MRT[i].BASE_HI */ + } else { + debug_assert((offset + size) <= fd_bo_size(rsc->bo)); + OUT_RELOCW(ring, rsc->bo, offset, 0, 0); /* BASE_LO/HI */ + } + + OUT_PKT4(ring, REG_A5XX_SP_FS_MRT_REG(i), 1); + OUT_RING(ring, A5XX_SP_FS_MRT_REG_COLOR_FORMAT(format) | + COND(sint, A5XX_SP_FS_MRT_REG_COLOR_SINT) | + COND(uint, A5XX_SP_FS_MRT_REG_COLOR_UINT) | + COND(srgb, A5XX_SP_FS_MRT_REG_COLOR_SRGB)); + + /* when we support UBWC, these would be the system memory + * addr/pitch/etc: + */ + OUT_PKT4(ring, REG_A5XX_RB_MRT_FLAG_BUFFER(i), 4); + OUT_RING(ring, 0x00000000); /* RB_MRT_FLAG_BUFFER[i].ADDR_LO */ + OUT_RING(ring, 0x00000000); /* RB_MRT_FLAG_BUFFER[i].ADDR_HI */ + OUT_RING(ring, A5XX_RB_MRT_FLAG_BUFFER_PITCH(0)); + OUT_RING(ring, A5XX_RB_MRT_FLAG_BUFFER_ARRAY_PITCH(0)); + } +} + +static void +emit_zs(struct fd_ringbuffer *ring, struct pipe_surface *zsbuf, + struct fd_gmem_stateobj *gmem) +{ + if (zsbuf) { + struct fd_resource *rsc = fd_resource(zsbuf->texture); + enum a5xx_depth_format fmt = fd5_pipe2depth(zsbuf->format); + uint32_t cpp = rsc->cpp; + uint32_t stride = 0; + uint32_t size = 0; + + if (gmem) { + stride = cpp * gmem->bin_w; + size = stride * gmem->bin_h; + } else { + struct fd_resource_slice *slice = fd_resource_slice(rsc, 0); + stride = slice->pitch * rsc->cpp; + size = slice->size0; + } + + OUT_PKT4(ring, REG_A5XX_RB_DEPTH_BUFFER_INFO, 5); + OUT_RING(ring, A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT(fmt)); + if (gmem) { + OUT_RING(ring, gmem->zsbuf_base[0]); /* RB_DEPTH_BUFFER_BASE_LO */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_BUFFER_BASE_HI */ + } else { + OUT_RELOCW(ring, rsc->bo, 0, 0, 0); /* RB_DEPTH_BUFFER_BASE_LO/HI */ + } + OUT_RING(ring, A5XX_RB_DEPTH_BUFFER_PITCH(stride)); + OUT_RING(ring, A5XX_RB_DEPTH_BUFFER_ARRAY_PITCH(size)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_DEPTH_BUFFER_INFO, 1); + OUT_RING(ring, A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT(fmt)); + + OUT_PKT4(ring, REG_A5XX_RB_DEPTH_FLAG_BUFFER_BASE_LO, 3); + OUT_RING(ring, 0x00000000); /* RB_DEPTH_FLAG_BUFFER_BASE_LO */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_FLAG_BUFFER_BASE_HI */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_FLAG_BUFFER_PITCH */ + + if (rsc->lrz) { + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_BUFFER_BASE_LO, 3); + OUT_RELOCW(ring, rsc->lrz, 0x1000, 0, 0); + OUT_RING(ring, A5XX_GRAS_LRZ_BUFFER_PITCH(rsc->lrz_pitch)); + + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_FAST_CLEAR_BUFFER_BASE_LO, 2); + OUT_RELOCW(ring, rsc->lrz, 0, 0, 0); + } else { + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_BUFFER_BASE_LO, 3); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); /* GRAS_LRZ_BUFFER_PITCH */ + + OUT_PKT4(ring, REG_A5XX_GRAS_LRZ_FAST_CLEAR_BUFFER_BASE_LO, 2); + OUT_RING(ring, 0x00000000); + OUT_RING(ring, 0x00000000); + } + + if (rsc->stencil) { + if (gmem) { + stride = 1 * gmem->bin_w; + size = stride * gmem->bin_h; + } else { + struct fd_resource_slice *slice = fd_resource_slice(rsc->stencil, 0); + stride = slice->pitch * rsc->cpp; + size = slice->size0; + } + + OUT_PKT4(ring, REG_A5XX_RB_STENCIL_INFO, 5); + OUT_RING(ring, A5XX_RB_STENCIL_INFO_SEPARATE_STENCIL); + if (gmem) { + OUT_RING(ring, gmem->zsbuf_base[1]); /* RB_STENCIL_BASE_LO */ + OUT_RING(ring, 0x00000000); /* RB_STENCIL_BASE_HI */ + } else { + OUT_RELOCW(ring, rsc->stencil->bo, 0, 0, 0); /* RB_STENCIL_BASE_LO/HI */ + } + OUT_RING(ring, A5XX_RB_STENCIL_PITCH(stride)); + OUT_RING(ring, A5XX_RB_STENCIL_ARRAY_PITCH(size)); + } else { + OUT_PKT4(ring, REG_A5XX_RB_STENCIL_INFO, 1); + OUT_RING(ring, 0x00000000); /* RB_STENCIL_INFO */ + } + } else { + OUT_PKT4(ring, REG_A5XX_RB_DEPTH_BUFFER_INFO, 5); + OUT_RING(ring, A5XX_RB_DEPTH_BUFFER_INFO_DEPTH_FORMAT(DEPTH5_NONE)); + OUT_RING(ring, 0x00000000); /* RB_DEPTH_BUFFER_BASE_LO */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_BUFFER_BASE_HI */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_BUFFER_PITCH */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_BUFFER_ARRAY_PITCH */ + + OUT_PKT4(ring, REG_A5XX_GRAS_SU_DEPTH_BUFFER_INFO, 1); + OUT_RING(ring, A5XX_GRAS_SU_DEPTH_BUFFER_INFO_DEPTH_FORMAT(DEPTH5_NONE)); + + OUT_PKT4(ring, REG_A5XX_RB_DEPTH_FLAG_BUFFER_BASE_LO, 3); + OUT_RING(ring, 0x00000000); /* RB_DEPTH_FLAG_BUFFER_BASE_LO */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_FLAG_BUFFER_BASE_HI */ + OUT_RING(ring, 0x00000000); /* RB_DEPTH_FLAG_BUFFER_PITCH */ + + OUT_PKT4(ring, REG_A5XX_RB_STENCIL_INFO, 1); + OUT_RING(ring, 0x00000000); /* RB_STENCIL_INFO */ + } +} + +static bool +use_hw_binning(struct fd_batch *batch) +{ + struct fd_gmem_stateobj *gmem = &batch->ctx->gmem; + + if ((gmem->maxpw * gmem->maxph) > 32) + return false; + + if ((gmem->maxpw > 15) || (gmem->maxph > 15)) + return false; + + return fd_binning_enabled && ((gmem->nbins_x * gmem->nbins_y) > 2) && + (batch->num_draws > 0); +} + +static void +patch_draws(struct fd_batch *batch, enum pc_di_vis_cull_mode vismode) +{ + unsigned i; + for (i = 0; i < fd_patch_num_elements(&batch->draw_patches); i++) { + struct fd_cs_patch *patch = fd_patch_element(&batch->draw_patches, i); + *patch->cs = patch->val | DRAW4(0, 0, 0, vismode); + } + util_dynarray_resize(&batch->draw_patches, 0); +} + +static void +update_vsc_pipe(struct fd_batch *batch) +{ + struct fd_context *ctx = batch->ctx; + struct fd5_context *fd5_ctx = fd5_context(ctx); + struct fd_gmem_stateobj *gmem = &batch->ctx->gmem; + struct fd_ringbuffer *ring = batch->gmem; + int i; + + OUT_PKT4(ring, REG_A5XX_VSC_BIN_SIZE, 3); + OUT_RING(ring, A5XX_VSC_BIN_SIZE_WIDTH(gmem->bin_w) | + A5XX_VSC_BIN_SIZE_HEIGHT(gmem->bin_h)); + OUT_RELOCW(ring, fd5_ctx->vsc_size_mem, 0, 0, 0); /* VSC_SIZE_ADDRESS_LO/HI */ + + OUT_PKT4(ring, REG_A5XX_UNKNOWN_0BC5, 2); + OUT_RING(ring, 0x00000000); /* UNKNOWN_0BC5 */ + OUT_RING(ring, 0x00000000); /* UNKNOWN_0BC6 */ + + OUT_PKT4(ring, REG_A5XX_VSC_PIPE_CONFIG_REG(0), 16); + for (i = 0; i < 16; i++) { + struct fd_vsc_pipe *pipe = &ctx->pipe[i]; + OUT_RING(ring, A5XX_VSC_PIPE_CONFIG_REG_X(pipe->x) | + A5XX_VSC_PIPE_CONFIG_REG_Y(pipe->y) | + A5XX_VSC_PIPE_CONFIG_REG_W(pipe->w) | + A5XX_VSC_PIPE_CONFIG_REG_H(pipe->h)); + } + + OUT_PKT4(ring, REG_A5XX_VSC_PIPE_DATA_ADDRESS_LO(0), 32); + for (i = 0; i < 16; i++) { + struct fd_vsc_pipe *pipe = &ctx->pipe[i]; + if (!pipe->bo) { + pipe->bo = fd_bo_new(ctx->dev, 0x20000, + DRM_FREEDRENO_GEM_TYPE_KMEM); + } + OUT_RELOCW(ring, pipe->bo, 0, 0, 0); /* VSC_PIPE_DATA_ADDRESS[i].LO/HI */ + } + + OUT_PKT4(ring, REG_A5XX_VSC_PIPE_DATA_LENGTH_REG(0), 16); + for (i = 0; i < 16; i++) { + struct fd_vsc_pipe *pipe = &ctx->pipe[i]; + OUT_RING(ring, fd_bo_size(pipe->bo) - 32); /* VSC_PIPE_DATA_LENGTH[i] */ + } +} + +static void +emit_binning_pass(struct fd_batch *batch) +{ + struct fd_context *ctx = batch->ctx; + struct fd_ringbuffer *ring = batch->gmem; + struct fd_gmem_stateobj *gmem = &batch->ctx->gmem; + + uint32_t x1 = gmem->minx; + uint32_t y1 = gmem->miny; + uint32_t x2 = gmem->minx + gmem->width - 1; + uint32_t y2 = gmem->miny + gmem->height - 1; + + fd5_set_render_mode(batch->ctx, ring, BINNING); + + OUT_PKT4(ring, REG_A5XX_RB_CNTL, 1); + OUT_RING(ring, A5XX_RB_CNTL_WIDTH(gmem->bin_w) | + A5XX_RB_CNTL_HEIGHT(gmem->bin_h)); + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_WINDOW_SCISSOR_TL, 2); + OUT_RING(ring, A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X(x1) | + A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(y1)); + OUT_RING(ring, A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X(x2) | + A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(y2)); + + OUT_PKT4(ring, REG_A5XX_RB_RESOLVE_CNTL_1, 2); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_1_X(x1) | + A5XX_RB_RESOLVE_CNTL_1_Y(y1)); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_2_X(x2) | + A5XX_RB_RESOLVE_CNTL_2_Y(y2)); + + update_vsc_pipe(batch); + + OUT_PKT4(ring, REG_A5XX_VPC_MODE_CNTL, 1); + OUT_RING(ring, A5XX_VPC_MODE_CNTL_BINNING_PASS); + + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, UNK_2C); + + OUT_PKT4(ring, REG_A5XX_RB_WINDOW_OFFSET, 1); + OUT_RING(ring, A5XX_RB_WINDOW_OFFSET_X(0) | + A5XX_RB_WINDOW_OFFSET_Y(0)); + + /* emit IB to binning drawcmds: */ + ctx->emit_ib(ring, batch->binning); + + fd_reset_wfi(batch); + + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, UNK_2D); + + OUT_PKT7(ring, CP_EVENT_WRITE, 4); + OUT_RING(ring, CACHE_FLUSH_TS); + OUT_RELOCW(ring, fd5_context(ctx)->blit_mem, 0, 0, 0); /* ADDR_LO/HI */ + OUT_RING(ring, 0x00000000); + + // TODO CP_COND_WRITE's for all the vsc buffers (check for overflow??) + + fd_wfi(batch, ring); + + OUT_PKT4(ring, REG_A5XX_VPC_MODE_CNTL, 1); + OUT_RING(ring, 0x0); +} + +/* before first tile */ +static void +fd5_emit_tile_init(struct fd_batch *batch) +{ + struct fd_context *ctx = batch->ctx; + struct fd_ringbuffer *ring = batch->gmem; + struct pipe_framebuffer_state *pfb = &batch->framebuffer; + + fd5_emit_restore(batch, ring); + + if (batch->lrz_clear) + ctx->emit_ib(ring, batch->lrz_clear); + + fd5_emit_lrz_flush(ring); + + OUT_PKT4(ring, REG_A5XX_GRAS_CL_CNTL, 1); + OUT_RING(ring, 0x00000080); /* GRAS_CL_CNTL */ + + OUT_PKT7(ring, CP_SKIP_IB2_ENABLE_GLOBAL, 1); + OUT_RING(ring, 0x0); + + OUT_PKT4(ring, REG_A5XX_PC_POWER_CNTL, 1); + OUT_RING(ring, 0x00000003); /* PC_POWER_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VFD_POWER_CNTL, 1); + OUT_RING(ring, 0x00000003); /* VFD_POWER_CNTL */ + + /* 0x10000000 for BYPASS.. 0x7c13c080 for GMEM: */ + fd_wfi(batch, ring); + OUT_PKT4(ring, REG_A5XX_RB_CCU_CNTL, 1); + OUT_RING(ring, 0x7c13c080); /* RB_CCU_CNTL */ + + emit_zs(ring, pfb->zsbuf, &ctx->gmem); + emit_mrt(ring, pfb->nr_cbufs, pfb->cbufs, &ctx->gmem); + + if (use_hw_binning(batch)) { + emit_binning_pass(batch); + fd5_emit_lrz_flush(ring); + patch_draws(batch, USE_VISIBILITY); + } else { + patch_draws(batch, IGNORE_VISIBILITY); + } + + fd5_set_render_mode(batch->ctx, ring, GMEM); +} + +/* before mem2gmem */ +static void +fd5_emit_tile_prep(struct fd_batch *batch, struct fd_tile *tile) +{ + struct fd_context *ctx = batch->ctx; + struct fd5_context *fd5_ctx = fd5_context(ctx); + struct fd_ringbuffer *ring = batch->gmem; + + uint32_t x1 = tile->xoff; + uint32_t y1 = tile->yoff; + uint32_t x2 = tile->xoff + tile->bin_w - 1; + uint32_t y2 = tile->yoff + tile->bin_h - 1; + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_WINDOW_SCISSOR_TL, 2); + OUT_RING(ring, A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X(x1) | + A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(y1)); + OUT_RING(ring, A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X(x2) | + A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(y2)); + + OUT_PKT4(ring, REG_A5XX_RB_RESOLVE_CNTL_1, 2); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_1_X(x1) | + A5XX_RB_RESOLVE_CNTL_1_Y(y1)); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_2_X(x2) | + A5XX_RB_RESOLVE_CNTL_2_Y(y2)); + + if (use_hw_binning(batch)) { + struct fd_vsc_pipe *pipe = &ctx->pipe[tile->p]; + + OUT_PKT7(ring, CP_WAIT_FOR_ME, 0); + + OUT_PKT7(ring, CP_SET_VISIBILITY_OVERRIDE, 1); + OUT_RING(ring, 0x0); + + OUT_PKT7(ring, CP_SET_BIN_DATA5, 5); + OUT_RING(ring, CP_SET_BIN_DATA5_0_VSC_SIZE(pipe->w * pipe->h) | + CP_SET_BIN_DATA5_0_VSC_N(tile->n)); + OUT_RELOC(ring, pipe->bo, 0, 0, 0); /* VSC_PIPE[p].DATA_ADDRESS */ + OUT_RELOC(ring, fd5_ctx->vsc_size_mem, /* VSC_SIZE_ADDRESS + (p * 4) */ + (tile->p * 4), 0, 0); + } else { + OUT_PKT7(ring, CP_SET_VISIBILITY_OVERRIDE, 1); + OUT_RING(ring, 0x1); + } + + OUT_PKT4(ring, REG_A5XX_RB_WINDOW_OFFSET, 1); + OUT_RING(ring, A5XX_RB_WINDOW_OFFSET_X(x1) | + A5XX_RB_WINDOW_OFFSET_Y(y1)); +} + + +/* + * transfer from system memory to gmem + */ + +static void +emit_mem2gmem_surf(struct fd_batch *batch, uint32_t base, + struct pipe_surface *psurf, enum a5xx_blit_buf buf) +{ + struct fd_ringbuffer *ring = batch->gmem; + struct fd_gmem_stateobj *gmem = &batch->ctx->gmem; + struct fd_resource *rsc = fd_resource(psurf->texture); + uint32_t stride, size; + + debug_assert(psurf->u.tex.first_layer == psurf->u.tex.last_layer); + + stride = gmem->bin_w * rsc->cpp; + size = stride * gmem->bin_h; + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_FLAG_DST_LO, 4); + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_LO */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_HI */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_PITCH */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_ARRAY_PITCH */ + + OUT_PKT4(ring, REG_A5XX_RB_RESOLVE_CNTL_3, 5); + OUT_RING(ring, 0x00000000); /* RB_RESOLVE_CNTL_3 */ + OUT_RING(ring, base); /* RB_BLIT_DST_LO */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_DST_HI */ + OUT_RING(ring, A5XX_RB_BLIT_DST_PITCH(stride)); + OUT_RING(ring, A5XX_RB_BLIT_DST_ARRAY_PITCH(size)); + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_CNTL, 1); + OUT_RING(ring, A5XX_RB_BLIT_CNTL_BUF(buf)); + + fd5_emit_blit(batch->ctx, ring); +} + +static void +fd5_emit_tile_mem2gmem(struct fd_batch *batch, struct fd_tile *tile) +{ + struct fd_ringbuffer *ring = batch->gmem; + struct fd_context *ctx = batch->ctx; + struct fd_gmem_stateobj *gmem = &ctx->gmem; + struct pipe_framebuffer_state *pfb = &batch->framebuffer; + + /* + * setup mrt and zs with system memory base addresses: + */ + + emit_mrt(ring, pfb->nr_cbufs, pfb->cbufs, NULL); +// emit_zs(ring, pfb->zsbuf, NULL); + + OUT_PKT4(ring, REG_A5XX_RB_CNTL, 1); + OUT_RING(ring, A5XX_RB_CNTL_WIDTH(gmem->bin_w) | + A5XX_RB_CNTL_HEIGHT(gmem->bin_h) | + A5XX_RB_CNTL_BYPASS); + + if (fd_gmem_needs_restore(batch, tile, FD_BUFFER_COLOR)) { + unsigned i; + for (i = 0; i < pfb->nr_cbufs; i++) { + if (!pfb->cbufs[i]) + continue; + if (!(batch->restore & (PIPE_CLEAR_COLOR0 << i))) + continue; + emit_mem2gmem_surf(batch, gmem->cbuf_base[i], + pfb->cbufs[i], BLIT_MRT0 + i); + } + } + + if (fd_gmem_needs_restore(batch, tile, FD_BUFFER_DEPTH | FD_BUFFER_STENCIL)) { + struct fd_resource *rsc = fd_resource(pfb->zsbuf->texture); + // XXX BLIT_ZS vs BLIT_Z32 .. need some more cmdstream traces + // with z32_x24s8.. + + // XXX hack import via BLIT_MRT0 instead of BLIT_ZS, since I don't + // know otherwise how to go from linear in sysmem to tiled in gmem. + // possibly we want to flip this around gmem2mem and keep depth + // tiled in sysmem (and fixup sampler state to assume tiled).. this + // might be required for doing depth/stencil in bypass mode? + struct fd_resource_slice *slice = fd_resource_slice(rsc, 0); + enum a5xx_color_fmt format = + fd5_pipe2color(fd_gmem_restore_format(pfb->zsbuf->format)); + + OUT_PKT4(ring, REG_A5XX_RB_MRT_BUF_INFO(0), 5); + OUT_RING(ring, A5XX_RB_MRT_BUF_INFO_COLOR_FORMAT(format) | + A5XX_RB_MRT_BUF_INFO_COLOR_TILE_MODE(TILE5_LINEAR) | + A5XX_RB_MRT_BUF_INFO_COLOR_SWAP(WZYX)); + OUT_RING(ring, A5XX_RB_MRT_PITCH(slice->pitch * rsc->cpp)); + OUT_RING(ring, A5XX_RB_MRT_ARRAY_PITCH(slice->size0)); + OUT_RELOCW(ring, rsc->bo, 0, 0, 0); /* BASE_LO/HI */ + + emit_mem2gmem_surf(batch, ctx->gmem.zsbuf_base[0], pfb->zsbuf, BLIT_MRT0); + } +} + + +/* before IB to rendering cmds: */ +static void +fd5_emit_tile_renderprep(struct fd_batch *batch, struct fd_tile *tile) +{ + struct fd_ringbuffer *ring = batch->gmem; + struct fd_gmem_stateobj *gmem = &batch->ctx->gmem; + struct pipe_framebuffer_state *pfb = &batch->framebuffer; + + OUT_PKT4(ring, REG_A5XX_RB_CNTL, 1); + OUT_RING(ring, A5XX_RB_CNTL_WIDTH(gmem->bin_w) | + A5XX_RB_CNTL_HEIGHT(gmem->bin_h)); + + emit_zs(ring, pfb->zsbuf, gmem); + emit_mrt(ring, pfb->nr_cbufs, pfb->cbufs, gmem); + + // TODO MSAA + OUT_PKT4(ring, REG_A5XX_TPL1_TP_RAS_MSAA_CNTL, 2); + OUT_RING(ring, A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES(MSAA_ONE)); + OUT_RING(ring, A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE) | + A5XX_TPL1_TP_DEST_MSAA_CNTL_MSAA_DISABLE); + + OUT_PKT4(ring, REG_A5XX_RB_RAS_MSAA_CNTL, 2); + OUT_RING(ring, A5XX_RB_RAS_MSAA_CNTL_SAMPLES(MSAA_ONE)); + OUT_RING(ring, A5XX_RB_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE) | + A5XX_RB_DEST_MSAA_CNTL_MSAA_DISABLE); + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_RAS_MSAA_CNTL, 2); + OUT_RING(ring, A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES(MSAA_ONE)); + OUT_RING(ring, A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE) | + A5XX_GRAS_SC_DEST_MSAA_CNTL_MSAA_DISABLE); +} + + +/* + * transfer from gmem to system memory (ie. normal RAM) + */ + +static void +emit_gmem2mem_surf(struct fd_batch *batch, uint32_t base, + struct pipe_surface *psurf, enum a5xx_blit_buf buf) +{ + struct fd_ringbuffer *ring = batch->gmem; + struct fd_resource *rsc = fd_resource(psurf->texture); + struct fd_resource_slice *slice; + uint32_t offset; + + slice = fd_resource_slice(rsc, psurf->u.tex.level); + offset = fd_resource_offset(rsc, psurf->u.tex.level, + psurf->u.tex.first_layer); + + debug_assert(psurf->u.tex.first_layer == psurf->u.tex.last_layer); + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_FLAG_DST_LO, 4); + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_LO */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_HI */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_PITCH */ + OUT_RING(ring, 0x00000000); /* RB_BLIT_FLAG_DST_ARRAY_PITCH */ + + OUT_PKT4(ring, REG_A5XX_RB_RESOLVE_CNTL_3, 5); + OUT_RING(ring, 0x00000004); /* XXX RB_RESOLVE_CNTL_3 */ + OUT_RELOCW(ring, rsc->bo, offset, 0, 0); /* RB_BLIT_DST_LO/HI */ + OUT_RING(ring, A5XX_RB_BLIT_DST_PITCH(slice->pitch * rsc->cpp)); + OUT_RING(ring, A5XX_RB_BLIT_DST_ARRAY_PITCH(slice->size0)); + + OUT_PKT4(ring, REG_A5XX_RB_BLIT_CNTL, 1); + OUT_RING(ring, A5XX_RB_BLIT_CNTL_BUF(buf)); + + fd5_emit_blit(batch->ctx, ring); +} + +static void +fd5_emit_tile_gmem2mem(struct fd_batch *batch, struct fd_tile *tile) +{ + struct fd_context *ctx = batch->ctx; + struct fd_gmem_stateobj *gmem = &ctx->gmem; + struct pipe_framebuffer_state *pfb = &batch->framebuffer; + + if (batch->resolve & (FD_BUFFER_DEPTH | FD_BUFFER_STENCIL)) { + struct fd_resource *rsc = fd_resource(pfb->zsbuf->texture); + // XXX BLIT_ZS vs BLIT_Z32 .. need some more cmdstream traces + // with z32_x24s8.. + if (!rsc->stencil || (batch->resolve & FD_BUFFER_DEPTH)) + emit_gmem2mem_surf(batch, gmem->zsbuf_base[0], pfb->zsbuf, BLIT_ZS); + if (rsc->stencil && (batch->resolve & FD_BUFFER_STENCIL)) + emit_gmem2mem_surf(batch, gmem->zsbuf_base[1], pfb->zsbuf, BLIT_ZS); + } + + if (batch->resolve & FD_BUFFER_COLOR) { + unsigned i; + for (i = 0; i < pfb->nr_cbufs; i++) { + if (!pfb->cbufs[i]) + continue; + if (!(batch->resolve & (PIPE_CLEAR_COLOR0 << i))) + continue; + emit_gmem2mem_surf(batch, gmem->cbuf_base[i], + pfb->cbufs[i], BLIT_MRT0 + i); + } + } +} + +static void +fd5_emit_tile_fini(struct fd_batch *batch) +{ + struct fd_ringbuffer *ring = batch->gmem; + + OUT_PKT7(ring, CP_SKIP_IB2_ENABLE_GLOBAL, 1); + OUT_RING(ring, 0x0); + + fd5_emit_lrz_flush(ring); + + fd5_cache_flush(batch, ring); + fd5_set_render_mode(batch->ctx, ring, BYPASS); +} + +static void +fd5_emit_sysmem_prep(struct fd_batch *batch) +{ + struct pipe_framebuffer_state *pfb = &batch->framebuffer; + struct fd_ringbuffer *ring = batch->gmem; + + fd5_emit_restore(batch, ring); + + fd5_emit_lrz_flush(ring); + + OUT_PKT7(ring, CP_SKIP_IB2_ENABLE_GLOBAL, 1); + OUT_RING(ring, 0x0); + + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, UNK_19); + + OUT_PKT4(ring, REG_A5XX_PC_POWER_CNTL, 1); + OUT_RING(ring, 0x00000003); /* PC_POWER_CNTL */ + + OUT_PKT4(ring, REG_A5XX_VFD_POWER_CNTL, 1); + OUT_RING(ring, 0x00000003); /* VFD_POWER_CNTL */ + + /* 0x10000000 for BYPASS.. 0x7c13c080 for GMEM: */ + fd_wfi(batch, ring); + OUT_PKT4(ring, REG_A5XX_RB_CCU_CNTL, 1); + OUT_RING(ring, 0x10000000); /* RB_CCU_CNTL */ + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_WINDOW_SCISSOR_TL, 2); + OUT_RING(ring, A5XX_GRAS_SC_WINDOW_SCISSOR_TL_X(0) | + A5XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(0)); + OUT_RING(ring, A5XX_GRAS_SC_WINDOW_SCISSOR_BR_X(pfb->width - 1) | + A5XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(pfb->height - 1)); + + OUT_PKT4(ring, REG_A5XX_RB_RESOLVE_CNTL_1, 2); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_1_X(0) | + A5XX_RB_RESOLVE_CNTL_1_Y(0)); + OUT_RING(ring, A5XX_RB_RESOLVE_CNTL_2_X(pfb->width - 1) | + A5XX_RB_RESOLVE_CNTL_2_Y(pfb->height - 1)); + + OUT_PKT4(ring, REG_A5XX_RB_WINDOW_OFFSET, 1); + OUT_RING(ring, A5XX_RB_WINDOW_OFFSET_X(0) | + A5XX_RB_WINDOW_OFFSET_Y(0)); + + OUT_PKT7(ring, CP_SET_VISIBILITY_OVERRIDE, 1); + OUT_RING(ring, 0x1); + + OUT_PKT4(ring, REG_A5XX_RB_CNTL, 1); + OUT_RING(ring, A5XX_RB_CNTL_WIDTH(0) | + A5XX_RB_CNTL_HEIGHT(0) | + A5XX_RB_CNTL_BYPASS); + + patch_draws(batch, IGNORE_VISIBILITY); + + emit_zs(ring, pfb->zsbuf, NULL); + emit_mrt(ring, pfb->nr_cbufs, pfb->cbufs, NULL); + + // TODO MSAA + OUT_PKT4(ring, REG_A5XX_TPL1_TP_RAS_MSAA_CNTL, 2); + OUT_RING(ring, A5XX_TPL1_TP_RAS_MSAA_CNTL_SAMPLES(MSAA_ONE)); + OUT_RING(ring, A5XX_TPL1_TP_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE) | + A5XX_TPL1_TP_DEST_MSAA_CNTL_MSAA_DISABLE); + + OUT_PKT4(ring, REG_A5XX_RB_RAS_MSAA_CNTL, 2); + OUT_RING(ring, A5XX_RB_RAS_MSAA_CNTL_SAMPLES(MSAA_ONE)); + OUT_RING(ring, A5XX_RB_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE) | + A5XX_RB_DEST_MSAA_CNTL_MSAA_DISABLE); + + OUT_PKT4(ring, REG_A5XX_GRAS_SC_RAS_MSAA_CNTL, 2); + OUT_RING(ring, A5XX_GRAS_SC_RAS_MSAA_CNTL_SAMPLES(MSAA_ONE)); + OUT_RING(ring, A5XX_GRAS_SC_DEST_MSAA_CNTL_SAMPLES(MSAA_ONE) | + A5XX_GRAS_SC_DEST_MSAA_CNTL_MSAA_DISABLE); +} + +static void +fd5_emit_sysmem_fini(struct fd_batch *batch) +{ + struct fd5_context *fd5_ctx = fd5_context(batch->ctx); + struct fd_ringbuffer *ring = batch->gmem; + + OUT_PKT7(ring, CP_SKIP_IB2_ENABLE_GLOBAL, 1); + OUT_RING(ring, 0x0); + + fd5_emit_lrz_flush(ring); + + OUT_PKT7(ring, CP_EVENT_WRITE, 4); + OUT_RING(ring, UNK_1D); + OUT_RELOCW(ring, fd5_ctx->blit_mem, 0, 0, 0); /* ADDR_LO/HI */ + OUT_RING(ring, 0x00000000); +} + +void +fd5_gmem_init(struct pipe_context *pctx) +{ + struct fd_context *ctx = fd_context(pctx); + + ctx->emit_tile_init = fd5_emit_tile_init; + ctx->emit_tile_prep = fd5_emit_tile_prep; + ctx->emit_tile_mem2gmem = fd5_emit_tile_mem2gmem; + ctx->emit_tile_renderprep = fd5_emit_tile_renderprep; + ctx->emit_tile_gmem2mem = fd5_emit_tile_gmem2mem; + ctx->emit_tile_fini = fd5_emit_tile_fini; + ctx->emit_sysmem_prep = fd5_emit_sysmem_prep; + ctx->emit_sysmem_fini = fd5_emit_sysmem_fini; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_gmem.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_gmem.h new file mode 100644 index 000000000..7794bfb33 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_gmem.h @@ -0,0 +1,34 @@ +/* + * Copyright (C) 2015 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_GMEM_H_ +#define FD5_GMEM_H_ + +#include "pipe/p_context.h" + +void fd5_gmem_init(struct pipe_context *pctx); + +#endif /* FD5_GMEM_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_program.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_program.c new file mode 100644 index 000000000..aa4babdf5 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_program.c @@ -0,0 +1,732 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" +#include "util/u_inlines.h" +#include "util/u_format.h" +#include "util/bitset.h" + +#include "freedreno_program.h" + +#include "fd5_program.h" +#include "fd5_emit.h" +#include "fd5_texture.h" +#include "fd5_format.h" + +static void +delete_shader_stateobj(struct fd5_shader_stateobj *so) +{ + ir3_shader_destroy(so->shader); + free(so); +} + +static struct fd5_shader_stateobj * +create_shader_stateobj(struct pipe_context *pctx, const struct pipe_shader_state *cso, + enum shader_t type) +{ + struct fd_context *ctx = fd_context(pctx); + struct ir3_compiler *compiler = ctx->screen->compiler; + struct fd5_shader_stateobj *so = CALLOC_STRUCT(fd5_shader_stateobj); + so->shader = ir3_shader_create(compiler, cso, type, &ctx->debug); + return so; +} + +static void * +fd5_fp_state_create(struct pipe_context *pctx, + const struct pipe_shader_state *cso) +{ + return create_shader_stateobj(pctx, cso, SHADER_FRAGMENT); +} + +static void +fd5_fp_state_delete(struct pipe_context *pctx, void *hwcso) +{ + struct fd5_shader_stateobj *so = hwcso; + delete_shader_stateobj(so); +} + +static void * +fd5_vp_state_create(struct pipe_context *pctx, + const struct pipe_shader_state *cso) +{ + return create_shader_stateobj(pctx, cso, SHADER_VERTEX); +} + +static void +fd5_vp_state_delete(struct pipe_context *pctx, void *hwcso) +{ + struct fd5_shader_stateobj *so = hwcso; + delete_shader_stateobj(so); +} + +void +fd5_emit_shader(struct fd_ringbuffer *ring, const struct ir3_shader_variant *so) +{ + const struct ir3_info *si = &so->info; + enum a4xx_state_block sb = fd4_stage2shadersb(so->type); + enum a4xx_state_src src; + uint32_t i, sz, *bin; + + if (fd_mesa_debug & FD_DBG_DIRECT) { + sz = si->sizedwords; + src = SS4_DIRECT; + bin = fd_bo_map(so->bo); + } else { + sz = 0; + src = SS4_INDIRECT; + bin = NULL; + } + + OUT_PKT7(ring, CP_LOAD_STATE4, 3 + sz); + OUT_RING(ring, CP_LOAD_STATE4_0_DST_OFF(0) | + CP_LOAD_STATE4_0_STATE_SRC(src) | + CP_LOAD_STATE4_0_STATE_BLOCK(sb) | + CP_LOAD_STATE4_0_NUM_UNIT(so->instrlen)); + if (bin) { + OUT_RING(ring, CP_LOAD_STATE4_1_EXT_SRC_ADDR(0) | + CP_LOAD_STATE4_1_STATE_TYPE(ST4_SHADER)); + OUT_RING(ring, CP_LOAD_STATE4_2_EXT_SRC_ADDR_HI(0)); + } else { + OUT_RELOC(ring, so->bo, 0, + CP_LOAD_STATE4_1_STATE_TYPE(ST4_SHADER), 0); + } + + /* for how clever coverity is, it is sometimes rather dull, and + * doesn't realize that the only case where bin==NULL, sz==0: + */ + assume(bin || (sz == 0)); + + for (i = 0; i < sz; i++) { + OUT_RING(ring, bin[i]); + } +} + +/* Add any missing varyings needed for stream-out. Otherwise varyings not + * used by fragment shader will be stripped out. + */ +static void +link_stream_out(struct ir3_shader_linkage *l, const struct ir3_shader_variant *v) +{ + const struct pipe_stream_output_info *strmout = &v->shader->stream_output; + + /* + * First, any stream-out varyings not already in linkage map (ie. also + * consumed by frag shader) need to be added: + */ + for (unsigned i = 0; i < strmout->num_outputs; i++) { + const struct pipe_stream_output *out = &strmout->output[i]; + unsigned k = out->register_index; + unsigned compmask = + (1 << (out->num_components + out->start_component)) - 1; + unsigned idx, nextloc = 0; + + /* psize/pos need to be the last entries in linkage map, and will + * get added link_stream_out, so skip over them: + */ + if ((v->outputs[k].slot == VARYING_SLOT_PSIZ) || + (v->outputs[k].slot == VARYING_SLOT_POS)) + continue; + + for (idx = 0; idx < l->cnt; idx++) { + if (l->var[idx].regid == v->outputs[k].regid) + break; + nextloc = MAX2(nextloc, l->var[idx].loc + 4); + } + + /* add if not already in linkage map: */ + if (idx == l->cnt) + ir3_link_add(l, v->outputs[k].regid, compmask, nextloc); + + /* expand component-mask if needed, ie streaming out all components + * but frag shader doesn't consume all components: + */ + if (compmask & ~l->var[idx].compmask) { + l->var[idx].compmask |= compmask; + l->max_loc = MAX2(l->max_loc, + l->var[idx].loc + util_last_bit(l->var[idx].compmask)); + } + } +} + +/* TODO maybe some of this we could pre-compute once rather than having + * so much draw-time logic? + */ +static void +emit_stream_out(struct fd_ringbuffer *ring, const struct ir3_shader_variant *v, + struct ir3_shader_linkage *l) +{ + const struct pipe_stream_output_info *strmout = &v->shader->stream_output; + unsigned ncomp[PIPE_MAX_SO_BUFFERS] = {0}; + unsigned prog[align(l->max_loc, 2) / 2]; + + memset(prog, 0, sizeof(prog)); + + for (unsigned i = 0; i < strmout->num_outputs; i++) { + const struct pipe_stream_output *out = &strmout->output[i]; + unsigned k = out->register_index; + unsigned idx; + + ncomp[out->output_buffer] += out->num_components; + + /* linkage map sorted by order frag shader wants things, so + * a bit less ideal here.. + */ + for (idx = 0; idx < l->cnt; idx++) + if (l->var[idx].regid == v->outputs[k].regid) + break; + + debug_assert(idx < l->cnt); + + for (unsigned j = 0; j < out->num_components; j++) { + unsigned c = j + out->start_component; + unsigned loc = l->var[idx].loc + c; + unsigned off = j + out->dst_offset; /* in dwords */ + + if (loc & 1) { + prog[loc/2] |= A5XX_VPC_SO_PROG_B_EN | + A5XX_VPC_SO_PROG_B_BUF(out->output_buffer) | + A5XX_VPC_SO_PROG_B_OFF(off * 4); + } else { + prog[loc/2] |= A5XX_VPC_SO_PROG_A_EN | + A5XX_VPC_SO_PROG_A_BUF(out->output_buffer) | + A5XX_VPC_SO_PROG_A_OFF(off * 4); + } + } + } + + OUT_PKT7(ring, CP_CONTEXT_REG_BUNCH, 12 + (2 * ARRAY_SIZE(prog))); + OUT_RING(ring, REG_A5XX_VPC_SO_BUF_CNTL); + OUT_RING(ring, A5XX_VPC_SO_BUF_CNTL_ENABLE | + COND(ncomp[0] > 0, A5XX_VPC_SO_BUF_CNTL_BUF0) | + COND(ncomp[1] > 0, A5XX_VPC_SO_BUF_CNTL_BUF1) | + COND(ncomp[2] > 0, A5XX_VPC_SO_BUF_CNTL_BUF2) | + COND(ncomp[3] > 0, A5XX_VPC_SO_BUF_CNTL_BUF3)); + OUT_RING(ring, REG_A5XX_VPC_SO_NCOMP(0)); + OUT_RING(ring, ncomp[0]); + OUT_RING(ring, REG_A5XX_VPC_SO_NCOMP(1)); + OUT_RING(ring, ncomp[1]); + OUT_RING(ring, REG_A5XX_VPC_SO_NCOMP(2)); + OUT_RING(ring, ncomp[2]); + OUT_RING(ring, REG_A5XX_VPC_SO_NCOMP(3)); + OUT_RING(ring, ncomp[3]); + OUT_RING(ring, REG_A5XX_VPC_SO_CNTL); + OUT_RING(ring, A5XX_VPC_SO_CNTL_ENABLE); + for (unsigned i = 0; i < ARRAY_SIZE(prog); i++) { + OUT_RING(ring, REG_A5XX_VPC_SO_PROG); + OUT_RING(ring, prog[i]); + } +} + +struct stage { + const struct ir3_shader_variant *v; + const struct ir3_info *i; + /* const sizes are in units of 4 * vec4 */ + uint8_t constoff; + uint8_t constlen; + /* instr sizes are in units of 16 instructions */ + uint8_t instroff; + uint8_t instrlen; +}; + +enum { + VS = 0, + FS = 1, + HS = 2, + DS = 3, + GS = 4, + MAX_STAGES +}; + +static void +setup_stages(struct fd5_emit *emit, struct stage *s) +{ + unsigned i; + + s[VS].v = fd5_emit_get_vp(emit); + s[FS].v = fd5_emit_get_fp(emit); + + s[HS].v = s[DS].v = s[GS].v = NULL; /* for now */ + + for (i = 0; i < MAX_STAGES; i++) { + if (s[i].v) { + s[i].i = &s[i].v->info; + /* constlen is in units of 4 * vec4: */ + s[i].constlen = align(s[i].v->constlen, 4) / 4; + /* instrlen is already in units of 16 instr.. although + * probably we should ditch that and not make the compiler + * care about instruction group size of a3xx vs a5xx + */ + s[i].instrlen = s[i].v->instrlen; + } else { + s[i].i = NULL; + s[i].constlen = 0; + s[i].instrlen = 0; + } + } + + /* NOTE: at least for gles2, blob partitions VS at bottom of const + * space and FS taking entire remaining space. We probably don't + * need to do that the same way, but for now mimic what the blob + * does to make it easier to diff against register values from blob + * + * NOTE: if VS.instrlen + FS.instrlen > 64, then one or both shaders + * is run from external memory. + */ + if ((s[VS].instrlen + s[FS].instrlen) > 64) { + /* prioritize FS for internal memory: */ + if (s[FS].instrlen < 64) { + /* if FS can fit, kick VS out to external memory: */ + s[VS].instrlen = 0; + } else if (s[VS].instrlen < 64) { + /* otherwise if VS can fit, kick out FS: */ + s[FS].instrlen = 0; + } else { + /* neither can fit, run both from external memory: */ + s[VS].instrlen = 0; + s[FS].instrlen = 0; + } + } + + unsigned constoff = 0; + for (i = 0; i < MAX_STAGES; i++) { + s[i].constoff = constoff; + constoff += s[i].constlen; + } + + s[VS].instroff = 0; + s[FS].instroff = 64 - s[FS].instrlen; + s[HS].instroff = s[DS].instroff = s[GS].instroff = s[FS].instroff; +} + +void +fd5_program_emit(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct fd5_emit *emit) +{ + struct stage s[MAX_STAGES]; + uint32_t pos_regid, psize_regid, color_regid[8]; + uint32_t face_regid, coord_regid, zwcoord_regid; + uint32_t vcoord_regid, vertex_regid, instance_regid; + enum a3xx_threadsize fssz; + uint8_t psize_loc = ~0; + int i, j; + + setup_stages(emit, s); + + fssz = (s[FS].i->max_reg >= 24) ? TWO_QUADS : FOUR_QUADS; + + pos_regid = ir3_find_output_regid(s[VS].v, VARYING_SLOT_POS); + psize_regid = ir3_find_output_regid(s[VS].v, VARYING_SLOT_PSIZ); + vertex_regid = ir3_find_sysval_regid(s[VS].v, SYSTEM_VALUE_VERTEX_ID); + instance_regid = ir3_find_sysval_regid(s[VS].v, SYSTEM_VALUE_INSTANCE_ID); + + if (s[FS].v->color0_mrt) { + color_regid[0] = color_regid[1] = color_regid[2] = color_regid[3] = + color_regid[4] = color_regid[5] = color_regid[6] = color_regid[7] = + ir3_find_output_regid(s[FS].v, FRAG_RESULT_COLOR); + } else { + color_regid[0] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA0); + color_regid[1] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA1); + color_regid[2] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA2); + color_regid[3] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA3); + color_regid[4] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA4); + color_regid[5] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA5); + color_regid[6] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA6); + color_regid[7] = ir3_find_output_regid(s[FS].v, FRAG_RESULT_DATA7); + } + + /* TODO get these dynamically: */ + face_regid = s[FS].v->frag_face ? regid(0,0) : regid(63,0); + coord_regid = s[FS].v->frag_coord ? regid(0,0) : regid(63,0); + zwcoord_regid = s[FS].v->frag_coord ? regid(0,2) : regid(63,0); + vcoord_regid = (s[FS].v->total_in > 0) ? s[FS].v->pos_regid : regid(63,0); + + /* we could probably divide this up into things that need to be + * emitted if frag-prog is dirty vs if vert-prog is dirty.. + */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_VS_CONFIG, 5); + OUT_RING(ring, A5XX_HLSQ_VS_CONFIG_CONSTOBJECTOFFSET(s[VS].constoff) | + A5XX_HLSQ_VS_CONFIG_SHADEROBJOFFSET(s[VS].instroff) | + COND(s[VS].v, A5XX_HLSQ_VS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_HLSQ_FS_CONFIG_CONSTOBJECTOFFSET(s[FS].constoff) | + A5XX_HLSQ_FS_CONFIG_SHADEROBJOFFSET(s[FS].instroff) | + COND(s[FS].v, A5XX_HLSQ_FS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_HLSQ_HS_CONFIG_CONSTOBJECTOFFSET(s[HS].constoff) | + A5XX_HLSQ_HS_CONFIG_SHADEROBJOFFSET(s[HS].instroff) | + COND(s[HS].v, A5XX_HLSQ_HS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_HLSQ_DS_CONFIG_CONSTOBJECTOFFSET(s[DS].constoff) | + A5XX_HLSQ_DS_CONFIG_SHADEROBJOFFSET(s[DS].instroff) | + COND(s[DS].v, A5XX_HLSQ_DS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_HLSQ_GS_CONFIG_CONSTOBJECTOFFSET(s[GS].constoff) | + A5XX_HLSQ_GS_CONFIG_SHADEROBJOFFSET(s[GS].instroff) | + COND(s[GS].v, A5XX_HLSQ_GS_CONFIG_ENABLED)); + + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_CONFIG, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_HLSQ_VS_CNTL, 5); + OUT_RING(ring, A5XX_HLSQ_VS_CNTL_INSTRLEN(s[VS].instrlen) | + COND(s[VS].v && s[VS].v->has_ssbo, A5XX_HLSQ_VS_CNTL_SSBO_ENABLE)); + OUT_RING(ring, A5XX_HLSQ_FS_CNTL_INSTRLEN(s[FS].instrlen) | + COND(s[FS].v && s[FS].v->has_ssbo, A5XX_HLSQ_FS_CNTL_SSBO_ENABLE)); + OUT_RING(ring, A5XX_HLSQ_HS_CNTL_INSTRLEN(s[HS].instrlen) | + COND(s[HS].v && s[HS].v->has_ssbo, A5XX_HLSQ_HS_CNTL_SSBO_ENABLE)); + OUT_RING(ring, A5XX_HLSQ_DS_CNTL_INSTRLEN(s[DS].instrlen) | + COND(s[DS].v && s[DS].v->has_ssbo, A5XX_HLSQ_DS_CNTL_SSBO_ENABLE)); + OUT_RING(ring, A5XX_HLSQ_GS_CNTL_INSTRLEN(s[GS].instrlen) | + COND(s[GS].v && s[GS].v->has_ssbo, A5XX_HLSQ_GS_CNTL_SSBO_ENABLE)); + + OUT_PKT4(ring, REG_A5XX_SP_VS_CONFIG, 5); + OUT_RING(ring, A5XX_SP_VS_CONFIG_CONSTOBJECTOFFSET(s[VS].constoff) | + A5XX_SP_VS_CONFIG_SHADEROBJOFFSET(s[VS].instroff) | + COND(s[VS].v, A5XX_SP_VS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_SP_FS_CONFIG_CONSTOBJECTOFFSET(s[FS].constoff) | + A5XX_SP_FS_CONFIG_SHADEROBJOFFSET(s[FS].instroff) | + COND(s[FS].v, A5XX_SP_FS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_SP_HS_CONFIG_CONSTOBJECTOFFSET(s[HS].constoff) | + A5XX_SP_HS_CONFIG_SHADEROBJOFFSET(s[HS].instroff) | + COND(s[HS].v, A5XX_SP_HS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_SP_DS_CONFIG_CONSTOBJECTOFFSET(s[DS].constoff) | + A5XX_SP_DS_CONFIG_SHADEROBJOFFSET(s[DS].instroff) | + COND(s[DS].v, A5XX_SP_DS_CONFIG_ENABLED)); + OUT_RING(ring, A5XX_SP_GS_CONFIG_CONSTOBJECTOFFSET(s[GS].constoff) | + A5XX_SP_GS_CONFIG_SHADEROBJOFFSET(s[GS].instroff) | + COND(s[GS].v, A5XX_SP_GS_CONFIG_ENABLED)); + + OUT_PKT4(ring, REG_A5XX_SP_CS_CONFIG, 1); + OUT_RING(ring, 0x00000000); + + OUT_PKT4(ring, REG_A5XX_HLSQ_VS_CONSTLEN, 2); + OUT_RING(ring, s[VS].constlen); /* HLSQ_VS_CONSTLEN */ + OUT_RING(ring, s[VS].instrlen); /* HLSQ_VS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_FS_CONSTLEN, 2); + OUT_RING(ring, s[FS].constlen); /* HLSQ_FS_CONSTLEN */ + OUT_RING(ring, s[FS].instrlen); /* HLSQ_FS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_HS_CONSTLEN, 2); + OUT_RING(ring, s[HS].constlen); /* HLSQ_HS_CONSTLEN */ + OUT_RING(ring, s[HS].instrlen); /* HLSQ_HS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_DS_CONSTLEN, 2); + OUT_RING(ring, s[DS].constlen); /* HLSQ_DS_CONSTLEN */ + OUT_RING(ring, s[DS].instrlen); /* HLSQ_DS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_GS_CONSTLEN, 2); + OUT_RING(ring, s[GS].constlen); /* HLSQ_GS_CONSTLEN */ + OUT_RING(ring, s[GS].instrlen); /* HLSQ_GS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_HLSQ_CS_CONSTLEN, 2); + OUT_RING(ring, 0x00000000); /* HLSQ_CS_CONSTLEN */ + OUT_RING(ring, 0x00000000); /* HLSQ_CS_INSTRLEN */ + + OUT_PKT4(ring, REG_A5XX_SP_VS_CTRL_REG0, 1); + OUT_RING(ring, A5XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(s[VS].i->max_half_reg + 1) | + A5XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(s[VS].i->max_reg + 1) | + 0x6 | /* XXX seems to be always set? */ + A5XX_SP_VS_CTRL_REG0_BRANCHSTACK(0x3) | // XXX need to figure this out somehow.. + COND(s[VS].v->has_samp, A5XX_SP_VS_CTRL_REG0_PIXLODENABLE)); + + struct ir3_shader_linkage l = {0}; + ir3_link_shaders(&l, s[VS].v, s[FS].v); + + if ((s[VS].v->shader->stream_output.num_outputs > 0) && + !emit->key.binning_pass) + link_stream_out(&l, s[VS].v); + + BITSET_DECLARE(varbs, 128) = {0}; + uint32_t *varmask = (uint32_t *)varbs; + + for (i = 0; i < l.cnt; i++) + for (j = 0; j < util_last_bit(l.var[i].compmask); j++) + BITSET_SET(varbs, l.var[i].loc + j); + + OUT_PKT4(ring, REG_A5XX_VPC_VAR_DISABLE(0), 4); + OUT_RING(ring, ~varmask[0]); /* VPC_VAR[0].DISABLE */ + OUT_RING(ring, ~varmask[1]); /* VPC_VAR[1].DISABLE */ + OUT_RING(ring, ~varmask[2]); /* VPC_VAR[2].DISABLE */ + OUT_RING(ring, ~varmask[3]); /* VPC_VAR[3].DISABLE */ + + /* a5xx appends pos/psize to end of the linkage map: */ + if (pos_regid != regid(63,0)) + ir3_link_add(&l, pos_regid, 0xf, l.max_loc); + + if (psize_regid != regid(63,0)) { + psize_loc = l.max_loc; + ir3_link_add(&l, psize_regid, 0x1, l.max_loc); + } + + if ((s[VS].v->shader->stream_output.num_outputs > 0) && + !emit->key.binning_pass) { + emit_stream_out(ring, s[VS].v, &l); + + OUT_PKT4(ring, REG_A5XX_VPC_SO_OVERRIDE, 1); + OUT_RING(ring, 0x00000000); + } else { + OUT_PKT4(ring, REG_A5XX_VPC_SO_OVERRIDE, 1); + OUT_RING(ring, A5XX_VPC_SO_OVERRIDE_SO_DISABLE); + } + + for (i = 0, j = 0; (i < 16) && (j < l.cnt); i++) { + uint32_t reg = 0; + + OUT_PKT4(ring, REG_A5XX_SP_VS_OUT_REG(i), 1); + + reg |= A5XX_SP_VS_OUT_REG_A_REGID(l.var[j].regid); + reg |= A5XX_SP_VS_OUT_REG_A_COMPMASK(l.var[j].compmask); + j++; + + reg |= A5XX_SP_VS_OUT_REG_B_REGID(l.var[j].regid); + reg |= A5XX_SP_VS_OUT_REG_B_COMPMASK(l.var[j].compmask); + j++; + + OUT_RING(ring, reg); + } + + for (i = 0, j = 0; (i < 8) && (j < l.cnt); i++) { + uint32_t reg = 0; + + OUT_PKT4(ring, REG_A5XX_SP_VS_VPC_DST_REG(i), 1); + + reg |= A5XX_SP_VS_VPC_DST_REG_OUTLOC0(l.var[j++].loc); + reg |= A5XX_SP_VS_VPC_DST_REG_OUTLOC1(l.var[j++].loc); + reg |= A5XX_SP_VS_VPC_DST_REG_OUTLOC2(l.var[j++].loc); + reg |= A5XX_SP_VS_VPC_DST_REG_OUTLOC3(l.var[j++].loc); + + OUT_RING(ring, reg); + } + + OUT_PKT4(ring, REG_A5XX_SP_VS_OBJ_START_LO, 2); + OUT_RELOC(ring, s[VS].v->bo, 0, 0, 0); /* SP_VS_OBJ_START_LO/HI */ + + if (s[VS].instrlen) + fd5_emit_shader(ring, s[VS].v); + + // TODO depending on other bits in this reg (if any) set somewhere else? + OUT_PKT4(ring, REG_A5XX_PC_PRIM_VTX_CNTL, 1); + OUT_RING(ring, COND(s[VS].v->writes_psize, A5XX_PC_PRIM_VTX_CNTL_PSIZE)); + + OUT_PKT4(ring, REG_A5XX_SP_PRIMITIVE_CNTL, 1); + OUT_RING(ring, A5XX_SP_PRIMITIVE_CNTL_VSOUT(l.cnt)); + + OUT_PKT4(ring, REG_A5XX_VPC_CNTL_0, 1); + OUT_RING(ring, A5XX_VPC_CNTL_0_STRIDE_IN_VPC(l.max_loc) | + COND(s[FS].v->total_in > 0, A5XX_VPC_CNTL_0_VARYING) | + COND(s[FS].v->frag_coord, A5XX_VPC_CNTL_0_VARYING) | + 0x10000); // XXX + + fd5_context(ctx)->max_loc = l.max_loc; + + if (emit->key.binning_pass) { + OUT_PKT4(ring, REG_A5XX_SP_FS_OBJ_START_LO, 2); + OUT_RING(ring, 0x00000000); /* SP_FS_OBJ_START_LO */ + OUT_RING(ring, 0x00000000); /* SP_FS_OBJ_START_HI */ + } else { + OUT_PKT4(ring, REG_A5XX_SP_FS_OBJ_START_LO, 2); + OUT_RELOC(ring, s[FS].v->bo, 0, 0, 0); /* SP_FS_OBJ_START_LO/HI */ + } + + OUT_PKT4(ring, REG_A5XX_HLSQ_CONTROL_0_REG, 5); + OUT_RING(ring, A5XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(fssz) | + A5XX_HLSQ_CONTROL_0_REG_CSTHREADSIZE(TWO_QUADS) | + 0x00000880); /* XXX HLSQ_CONTROL_0 */ + OUT_RING(ring, A5XX_HLSQ_CONTROL_1_REG_PRIMALLOCTHRESHOLD(63)); + OUT_RING(ring, A5XX_HLSQ_CONTROL_2_REG_FACEREGID(face_regid) | + 0xfcfcfc00); /* XXX */ + OUT_RING(ring, A5XX_HLSQ_CONTROL_3_REG_FRAGCOORDXYREGID(vcoord_regid) | + 0xfcfcfc00); /* XXX */ + OUT_RING(ring, A5XX_HLSQ_CONTROL_4_REG_XYCOORDREGID(coord_regid) | + A5XX_HLSQ_CONTROL_4_REG_ZWCOORDREGID(zwcoord_regid) | + 0x0000fcfc); /* XXX */ + + OUT_PKT4(ring, REG_A5XX_SP_FS_CTRL_REG0, 1); + OUT_RING(ring, COND(s[FS].v->total_in > 0, A5XX_SP_FS_CTRL_REG0_VARYING) | + COND(s[FS].v->frag_coord, A5XX_SP_FS_CTRL_REG0_VARYING) | + 0x40006 | /* XXX set pretty much everywhere */ + A5XX_SP_FS_CTRL_REG0_THREADSIZE(fssz) | + A5XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(s[FS].i->max_half_reg + 1) | + A5XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(s[FS].i->max_reg + 1) | + A5XX_SP_FS_CTRL_REG0_BRANCHSTACK(0x3) | // XXX need to figure this out somehow.. + COND(s[FS].v->has_samp, A5XX_SP_FS_CTRL_REG0_PIXLODENABLE)); + + OUT_PKT4(ring, REG_A5XX_HLSQ_UPDATE_CNTL, 1); + OUT_RING(ring, 0x020fffff); /* XXX */ + + OUT_PKT4(ring, REG_A5XX_VPC_GS_SIV_CNTL, 1); + OUT_RING(ring, 0x0000ffff); /* XXX */ + + OUT_PKT4(ring, REG_A5XX_SP_SP_CNTL, 1); + OUT_RING(ring, 0x00000010); /* XXX */ + + OUT_PKT4(ring, REG_A5XX_GRAS_CNTL, 1); + OUT_RING(ring, COND(s[FS].v->total_in > 0, A5XX_GRAS_CNTL_VARYING) | + COND(s[FS].v->frag_coord, A5XX_GRAS_CNTL_XCOORD | + A5XX_GRAS_CNTL_YCOORD | + A5XX_GRAS_CNTL_ZCOORD | + A5XX_GRAS_CNTL_WCOORD | + A5XX_GRAS_CNTL_UNK3) | + COND(s[FS].v->frag_face, A5XX_GRAS_CNTL_UNK3)); + + OUT_PKT4(ring, REG_A5XX_RB_RENDER_CONTROL0, 2); + OUT_RING(ring, COND(s[FS].v->total_in > 0, A5XX_RB_RENDER_CONTROL0_VARYING) | + COND(s[FS].v->frag_coord, A5XX_RB_RENDER_CONTROL0_XCOORD | + A5XX_RB_RENDER_CONTROL0_YCOORD | + A5XX_RB_RENDER_CONTROL0_ZCOORD | + A5XX_RB_RENDER_CONTROL0_WCOORD | + A5XX_RB_RENDER_CONTROL0_UNK3) | + COND(s[FS].v->frag_face, A5XX_RB_RENDER_CONTROL0_UNK3)); + OUT_RING(ring, COND(s[FS].v->frag_face, A5XX_RB_RENDER_CONTROL1_FACENESS)); + + OUT_PKT4(ring, REG_A5XX_SP_FS_OUTPUT_REG(0), 8); + for (i = 0; i < 8; i++) { + OUT_RING(ring, A5XX_SP_FS_OUTPUT_REG_REGID(color_regid[i]) | + COND(emit->key.half_precision, + A5XX_SP_FS_OUTPUT_REG_HALF_PRECISION)); + } + + + OUT_PKT4(ring, REG_A5XX_VPC_PACK, 1); + OUT_RING(ring, A5XX_VPC_PACK_NUMNONPOSVAR(s[FS].v->total_in) | + A5XX_VPC_PACK_PSIZELOC(psize_loc)); + + if (!emit->key.binning_pass) { + uint32_t vinterp[8], vpsrepl[8]; + + memset(vinterp, 0, sizeof(vinterp)); + memset(vpsrepl, 0, sizeof(vpsrepl)); + + /* looks like we need to do int varyings in the frag + * shader on a5xx (no flatshad reg? or a420.0 bug?): + * + * (sy)(ss)nop + * (sy)ldlv.u32 r0.x,l[r0.x], 1 + * ldlv.u32 r0.y,l[r0.x+1], 1 + * (ss)bary.f (ei)r63.x, 0, r0.x + * (ss)(rpt1)cov.s32f16 hr0.x, (r)r0.x + * (rpt5)nop + * sam (f16)(xyzw)hr0.x, hr0.x, s#0, t#0 + * + * Possibly on later a5xx variants we'll be able to use + * something like the code below instead of workaround + * in the shader: + */ + /* figure out VARYING_INTERP / VARYING_PS_REPL register values: */ + for (j = -1; (j = ir3_next_varying(s[FS].v, j)) < (int)s[FS].v->inputs_count; ) { + /* NOTE: varyings are packed, so if compmask is 0xb + * then first, third, and fourth component occupy + * three consecutive varying slots: + */ + unsigned compmask = s[FS].v->inputs[j].compmask; + + uint32_t inloc = s[FS].v->inputs[j].inloc; + + if ((s[FS].v->inputs[j].interpolate == INTERP_MODE_FLAT) || + (s[FS].v->inputs[j].rasterflat && emit->rasterflat)) { + uint32_t loc = inloc; + + for (i = 0; i < 4; i++) { + if (compmask & (1 << i)) { + vinterp[loc / 16] |= 1 << ((loc % 16) * 2); + //flatshade[loc / 32] |= 1 << (loc % 32); + loc++; + } + } + } + + gl_varying_slot slot = s[FS].v->inputs[j].slot; + + /* since we don't enable PIPE_CAP_TGSI_TEXCOORD: */ + if (slot >= VARYING_SLOT_VAR0) { + unsigned texmask = 1 << (slot - VARYING_SLOT_VAR0); + /* Replace the .xy coordinates with S/T from the point sprite. Set + * interpolation bits for .zw such that they become .01 + */ + if (emit->sprite_coord_enable & texmask) { + /* mask is two 2-bit fields, where: + * '01' -> S + * '10' -> T + * '11' -> 1 - T (flip mode) + */ + unsigned mask = emit->sprite_coord_mode ? 0b1101 : 0b1001; + uint32_t loc = inloc; + if (compmask & 0x1) { + vpsrepl[loc / 16] |= ((mask >> 0) & 0x3) << ((loc % 16) * 2); + loc++; + } + if (compmask & 0x2) { + vpsrepl[loc / 16] |= ((mask >> 2) & 0x3) << ((loc % 16) * 2); + loc++; + } + if (compmask & 0x4) { + /* .z <- 0.0f */ + vinterp[loc / 16] |= 0b10 << ((loc % 16) * 2); + loc++; + } + if (compmask & 0x8) { + /* .w <- 1.0f */ + vinterp[loc / 16] |= 0b11 << ((loc % 16) * 2); + loc++; + } + } + } + } + + OUT_PKT4(ring, REG_A5XX_VPC_VARYING_INTERP_MODE(0), 8); + for (i = 0; i < 8; i++) + OUT_RING(ring, vinterp[i]); /* VPC_VARYING_INTERP[i].MODE */ + + OUT_PKT4(ring, REG_A5XX_VPC_VARYING_PS_REPL_MODE(0), 8); + for (i = 0; i < 8; i++) + OUT_RING(ring, vpsrepl[i]); /* VPC_VARYING_PS_REPL[i] */ + } + + if (!emit->key.binning_pass) + if (s[FS].instrlen) + fd5_emit_shader(ring, s[FS].v); + + OUT_PKT4(ring, REG_A5XX_VFD_CONTROL_1, 5); + OUT_RING(ring, A5XX_VFD_CONTROL_1_REGID4VTX(vertex_regid) | + A5XX_VFD_CONTROL_1_REGID4INST(instance_regid) | + 0xfc0000); + OUT_RING(ring, 0x0000fcfc); /* VFD_CONTROL_2 */ + OUT_RING(ring, 0x0000fcfc); /* VFD_CONTROL_3 */ + OUT_RING(ring, 0x000000fc); /* VFD_CONTROL_4 */ + OUT_RING(ring, 0x00000000); /* VFD_CONTROL_5 */ +} + +void +fd5_prog_init(struct pipe_context *pctx) +{ + pctx->create_fs_state = fd5_fp_state_create; + pctx->delete_fs_state = fd5_fp_state_delete; + + pctx->create_vs_state = fd5_vp_state_create; + pctx->delete_vs_state = fd5_vp_state_delete; + + fd_prog_init(pctx); +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_program.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_program.h new file mode 100644 index 000000000..585263e0a --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_program.h @@ -0,0 +1,47 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_PROGRAM_H_ +#define FD5_PROGRAM_H_ + +#include "pipe/p_context.h" +#include "freedreno_context.h" +#include "ir3_shader.h" + +struct fd5_shader_stateobj { + struct ir3_shader *shader; +}; + +struct fd5_emit; + +void fd5_emit_shader(struct fd_ringbuffer *ring, const struct ir3_shader_variant *so); + +void fd5_program_emit(struct fd_context *ctx, struct fd_ringbuffer *ring, + struct fd5_emit *emit); + +void fd5_prog_init(struct pipe_context *pctx); + +#endif /* FD5_PROGRAM_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_query.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_query.c new file mode 100644 index 000000000..87417f1f9 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_query.c @@ -0,0 +1,263 @@ +/* + * Copyright (C) 2017 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +/* NOTE: see https://github.com/freedreno/freedreno/wiki/A5xx-Queries */ + +#include "freedreno_query_acc.h" +#include "freedreno_resource.h" + +#include "fd5_context.h" +#include "fd5_format.h" +#include "fd5_query.h" + +struct PACKED fd5_query_sample { + uint64_t start; + uint64_t result; + uint64_t stop; +}; + +#define query_sample(aq, field) \ + fd_resource((aq)->prsc)->bo, \ + offsetof(struct fd5_query_sample, field), \ + 0, 0 + +/* + * Occlusion Query: + * + * OCCLUSION_COUNTER and OCCLUSION_PREDICATE differ only in how they + * interpret results + */ + +static void +occlusion_resume(struct fd_acc_query *aq, struct fd_batch *batch) +{ + struct fd_ringbuffer *ring = batch->draw; + + OUT_PKT4(ring, REG_A5XX_RB_SAMPLE_COUNT_CONTROL, 1); + OUT_RING(ring, A5XX_RB_SAMPLE_COUNT_CONTROL_COPY); + + OUT_PKT4(ring, REG_A5XX_RB_SAMPLE_COUNT_ADDR_LO, 2); + OUT_RELOCW(ring, query_sample(aq, start)); + + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, ZPASS_DONE); + fd_reset_wfi(batch); + + fd5_context(batch->ctx)->samples_passed_queries++; +} + +static void +occlusion_pause(struct fd_acc_query *aq, struct fd_batch *batch) +{ + struct fd_ringbuffer *ring = batch->draw; + + OUT_PKT7(ring, CP_MEM_WRITE, 4); + OUT_RELOCW(ring, query_sample(aq, stop)); + OUT_RING(ring, 0xffffffff); + OUT_RING(ring, 0xffffffff); + + OUT_PKT7(ring, CP_WAIT_MEM_WRITES, 0); + + OUT_PKT4(ring, REG_A5XX_RB_SAMPLE_COUNT_CONTROL, 1); + OUT_RING(ring, A5XX_RB_SAMPLE_COUNT_CONTROL_COPY); + + OUT_PKT4(ring, REG_A5XX_RB_SAMPLE_COUNT_ADDR_LO, 2); + OUT_RELOCW(ring, query_sample(aq, stop)); + + OUT_PKT7(ring, CP_EVENT_WRITE, 1); + OUT_RING(ring, ZPASS_DONE); + fd_reset_wfi(batch); + + OUT_PKT7(ring, CP_WAIT_REG_MEM, 6); + OUT_RING(ring, 0x00000014); // XXX + OUT_RELOC(ring, query_sample(aq, stop)); + OUT_RING(ring, 0xffffffff); + OUT_RING(ring, 0xffffffff); + OUT_RING(ring, 0x00000010); // XXX + + /* result += stop - start: */ + OUT_PKT7(ring, CP_MEM_TO_MEM, 9); + OUT_RING(ring, CP_MEM_TO_MEM_0_DOUBLE | + CP_MEM_TO_MEM_0_NEG_C); + OUT_RELOCW(ring, query_sample(aq, result)); /* dst */ + OUT_RELOC(ring, query_sample(aq, result)); /* srcA */ + OUT_RELOC(ring, query_sample(aq, stop)); /* srcB */ + OUT_RELOC(ring, query_sample(aq, start)); /* srcC */ + + fd5_context(batch->ctx)->samples_passed_queries--; +} + +static void +occlusion_counter_result(struct fd_context *ctx, void *buf, + union pipe_query_result *result) +{ + struct fd5_query_sample *sp = buf; + result->u64 = sp->result; +} + +static void +occlusion_predicate_result(struct fd_context *ctx, void *buf, + union pipe_query_result *result) +{ + struct fd5_query_sample *sp = buf; + result->b = !!sp->result; +} + +static const struct fd_acc_sample_provider occlusion_counter = { + .query_type = PIPE_QUERY_OCCLUSION_COUNTER, + .active = FD_STAGE_DRAW, + .size = sizeof(struct fd5_query_sample), + .resume = occlusion_resume, + .pause = occlusion_pause, + .result = occlusion_counter_result, +}; + +static const struct fd_acc_sample_provider occlusion_predicate = { + .query_type = PIPE_QUERY_OCCLUSION_PREDICATE, + .active = FD_STAGE_DRAW, + .size = sizeof(struct fd5_query_sample), + .resume = occlusion_resume, + .pause = occlusion_pause, + .result = occlusion_predicate_result, +}; + +static const struct fd_acc_sample_provider occlusion_predicate_conservative = { + .query_type = PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE, + .active = FD_STAGE_DRAW, + .size = sizeof(struct fd5_query_sample), + .resume = occlusion_resume, + .pause = occlusion_pause, + .result = occlusion_predicate_result, +}; + +/* + * Timestamp Queries: + */ + +static void +timestamp_resume(struct fd_acc_query *aq, struct fd_batch *batch) +{ + struct fd_ringbuffer *ring = batch->draw; + + OUT_PKT7(ring, CP_EVENT_WRITE, 4); + OUT_RING(ring, CP_EVENT_WRITE_0_EVENT(CACHE_FLUSH_AND_INV_EVENT) | + CP_EVENT_WRITE_0_TIMESTAMP); + OUT_RELOCW(ring, query_sample(aq, start)); + OUT_RING(ring, 0x00000000); + + fd_reset_wfi(batch); +} + +static void +timestamp_pause(struct fd_acc_query *aq, struct fd_batch *batch) +{ + struct fd_ringbuffer *ring = batch->draw; + + OUT_PKT7(ring, CP_EVENT_WRITE, 4); + OUT_RING(ring, CP_EVENT_WRITE_0_EVENT(CACHE_FLUSH_AND_INV_EVENT) | + CP_EVENT_WRITE_0_TIMESTAMP); + OUT_RELOCW(ring, query_sample(aq, stop)); + OUT_RING(ring, 0x00000000); + + fd_reset_wfi(batch); + fd_wfi(batch, ring); + + /* result += stop - start: */ + OUT_PKT7(ring, CP_MEM_TO_MEM, 9); + OUT_RING(ring, CP_MEM_TO_MEM_0_DOUBLE | + CP_MEM_TO_MEM_0_NEG_C); + OUT_RELOCW(ring, query_sample(aq, result)); /* dst */ + OUT_RELOC(ring, query_sample(aq, result)); /* srcA */ + OUT_RELOC(ring, query_sample(aq, stop)); /* srcB */ + OUT_RELOC(ring, query_sample(aq, start)); /* srcC */ +} + +static uint64_t +ticks_to_ns(struct fd_context *ctx, uint32_t ts) +{ + /* This is based on the 19.2MHz always-on rbbm timer. + * + * TODO we should probably query this value from kernel.. + */ + return ts * (1000000000 / 19200000); +} + +static void +time_elapsed_accumulate_result(struct fd_context *ctx, void *buf, + union pipe_query_result *result) +{ + struct fd5_query_sample *sp = buf; + result->u64 = ticks_to_ns(ctx, sp->result); +} + +static void +timestamp_accumulate_result(struct fd_context *ctx, void *buf, + union pipe_query_result *result) +{ + struct fd5_query_sample *sp = buf; + result->u64 = ticks_to_ns(ctx, sp->result); +} + +static const struct fd_acc_sample_provider time_elapsed = { + .query_type = PIPE_QUERY_TIME_ELAPSED, + .active = FD_STAGE_DRAW | FD_STAGE_CLEAR, + .size = sizeof(struct fd5_query_sample), + .resume = timestamp_resume, + .pause = timestamp_pause, + .result = time_elapsed_accumulate_result, +}; + +/* NOTE: timestamp query isn't going to give terribly sensible results + * on a tiler. But it is needed by qapitrace profile heatmap. If you + * add in a binning pass, the results get even more non-sensical. So + * we just return the timestamp on the first tile and hope that is + * kind of good enough. + */ + +static const struct fd_acc_sample_provider timestamp = { + .query_type = PIPE_QUERY_TIMESTAMP, + .active = FD_STAGE_ALL, + .size = sizeof(struct fd5_query_sample), + .resume = timestamp_resume, + .pause = timestamp_pause, + .result = timestamp_accumulate_result, +}; + +void +fd5_query_context_init(struct pipe_context *pctx) +{ + struct fd_context *ctx = fd_context(pctx); + + ctx->create_query = fd_acc_create_query; + ctx->query_set_stage = fd_acc_query_set_stage; + + fd_acc_query_register_provider(pctx, &occlusion_counter); + fd_acc_query_register_provider(pctx, &occlusion_predicate); + fd_acc_query_register_provider(pctx, &occlusion_predicate_conservative); + + fd_acc_query_register_provider(pctx, &time_elapsed); + fd_acc_query_register_provider(pctx, ×tamp); +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_query.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_query.h new file mode 100644 index 000000000..2e563b0d5 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_query.h @@ -0,0 +1,34 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_QUERY_H_ +#define FD5_QUERY_H_ + +#include "pipe/p_context.h" + +void fd5_query_context_init(struct pipe_context *pctx); + +#endif /* FD5_QUERY_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_rasterizer.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_rasterizer.c new file mode 100644 index 000000000..2bbcbf2d6 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_rasterizer.c @@ -0,0 +1,98 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" + +#include "fd5_rasterizer.h" +#include "fd5_context.h" +#include "fd5_format.h" + +void * +fd5_rasterizer_state_create(struct pipe_context *pctx, + const struct pipe_rasterizer_state *cso) +{ + struct fd5_rasterizer_stateobj *so; + float psize_min, psize_max; + + so = CALLOC_STRUCT(fd5_rasterizer_stateobj); + if (!so) + return NULL; + + so->base = *cso; + + if (cso->point_size_per_vertex) { + psize_min = util_get_min_point_size(cso); + psize_max = 4092; + } else { + /* Force the point size to be as if the vertex output was disabled. */ + psize_min = cso->point_size; + psize_max = cso->point_size; + } + + so->gras_su_point_minmax = + A5XX_GRAS_SU_POINT_MINMAX_MIN(psize_min) | + A5XX_GRAS_SU_POINT_MINMAX_MAX(psize_max); + so->gras_su_point_size = A5XX_GRAS_SU_POINT_SIZE(cso->point_size); + so->gras_su_poly_offset_scale = + A5XX_GRAS_SU_POLY_OFFSET_SCALE(cso->offset_scale); + so->gras_su_poly_offset_offset = + A5XX_GRAS_SU_POLY_OFFSET_OFFSET(cso->offset_units); + so->gras_su_poly_offset_clamp = + A5XX_GRAS_SU_POLY_OFFSET_OFFSET_CLAMP(cso->offset_clamp); + + so->gras_su_cntl = + A5XX_GRAS_SU_CNTL_LINEHALFWIDTH(cso->line_width/2.0); + so->pc_raster_cntl = + A5XX_PC_RASTER_CNTL_POLYMODE_FRONT_PTYPE(fd_polygon_mode(cso->fill_front)) | + A5XX_PC_RASTER_CNTL_POLYMODE_BACK_PTYPE(fd_polygon_mode(cso->fill_back)); + + if (cso->fill_front != PIPE_POLYGON_MODE_FILL || + cso->fill_back != PIPE_POLYGON_MODE_FILL) + so->pc_raster_cntl |= A5XX_PC_RASTER_CNTL_POLYMODE_ENABLE; + + if (cso->cull_face & PIPE_FACE_FRONT) + so->gras_su_cntl |= A5XX_GRAS_SU_CNTL_CULL_FRONT; + if (cso->cull_face & PIPE_FACE_BACK) + so->gras_su_cntl |= A5XX_GRAS_SU_CNTL_CULL_BACK; + if (!cso->front_ccw) + so->gras_su_cntl |= A5XX_GRAS_SU_CNTL_FRONT_CW; + if (cso->offset_tri) + so->gras_su_cntl |= A5XX_GRAS_SU_CNTL_POLY_OFFSET; + + if (!cso->flatshade_first) + so->pc_primitive_cntl |= A5XX_PC_PRIMITIVE_CNTL_PROVOKING_VTX_LAST; + +// if (!cso->depth_clip) +// so->gras_cl_clip_cntl |= A5XX_GRAS_CL_CLIP_CNTL_ZNEAR_CLIP_DISABLE | +// A5XX_GRAS_CL_CLIP_CNTL_ZFAR_CLIP_DISABLE; + if (cso->clip_halfz) + so->gras_cl_clip_cntl |= A5XX_GRAS_CL_CNTL_ZERO_GB_SCALE_Z; + + return so; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_rasterizer.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_rasterizer.h new file mode 100644 index 000000000..b59758153 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_rasterizer.h @@ -0,0 +1,57 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_RASTERIZER_H_ +#define FD5_RASTERIZER_H_ + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + +struct fd5_rasterizer_stateobj { + struct pipe_rasterizer_state base; + + uint32_t gras_su_point_minmax; + uint32_t gras_su_point_size; + uint32_t gras_su_poly_offset_scale; + uint32_t gras_su_poly_offset_offset; + uint32_t gras_su_poly_offset_clamp; + + uint32_t gras_su_cntl; + uint32_t gras_cl_clip_cntl; + uint32_t pc_primitive_cntl; + uint32_t pc_raster_cntl; +}; + +static inline struct fd5_rasterizer_stateobj * +fd5_rasterizer_stateobj(struct pipe_rasterizer_state *rast) +{ + return (struct fd5_rasterizer_stateobj *)rast; +} + +void * fd5_rasterizer_state_create(struct pipe_context *pctx, + const struct pipe_rasterizer_state *cso); + +#endif /* FD5_RASTERIZER_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_screen.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_screen.c new file mode 100644 index 000000000..96f83ed33 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_screen.c @@ -0,0 +1,104 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_screen.h" +#include "util/u_format.h" + +#include "fd5_screen.h" +#include "fd5_context.h" +#include "fd5_format.h" +#include "ir3_compiler.h" + +static boolean +fd5_screen_is_format_supported(struct pipe_screen *pscreen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned sample_count, + unsigned usage) +{ + unsigned retval = 0; + + if ((target >= PIPE_MAX_TEXTURE_TYPES) || + (sample_count > 1) || /* TODO add MSAA */ + !util_format_is_supported(format, usage)) { + DBG("not supported: format=%s, target=%d, sample_count=%d, usage=%x", + util_format_name(format), target, sample_count, usage); + return FALSE; + } + + if ((usage & PIPE_BIND_VERTEX_BUFFER) && + (fd5_pipe2vtx(format) != (enum a5xx_vtx_fmt)~0)) { + retval |= PIPE_BIND_VERTEX_BUFFER; + } + + if ((usage & PIPE_BIND_SAMPLER_VIEW) && + (target == PIPE_BUFFER || + util_format_get_blocksize(format) != 12) && + (fd5_pipe2tex(format) != (enum a5xx_tex_fmt)~0)) { + retval |= PIPE_BIND_SAMPLER_VIEW; + } + + if ((usage & (PIPE_BIND_RENDER_TARGET | + PIPE_BIND_DISPLAY_TARGET | + PIPE_BIND_SCANOUT | + PIPE_BIND_SHARED)) && + (fd5_pipe2color(format) != (enum a5xx_color_fmt)~0) && + (fd5_pipe2tex(format) != (enum a5xx_tex_fmt)~0)) { + retval |= usage & (PIPE_BIND_RENDER_TARGET | + PIPE_BIND_DISPLAY_TARGET | + PIPE_BIND_SCANOUT | + PIPE_BIND_SHARED); + } + + if ((usage & PIPE_BIND_DEPTH_STENCIL) && + (fd5_pipe2depth(format) != (enum a5xx_depth_format)~0) && + (fd5_pipe2tex(format) != (enum a5xx_tex_fmt)~0)) { + retval |= PIPE_BIND_DEPTH_STENCIL; + } + + if ((usage & PIPE_BIND_INDEX_BUFFER) && + (fd_pipe2index(format) != (enum pc_di_index_size)~0)) { + retval |= PIPE_BIND_INDEX_BUFFER; + } + + if (retval != usage) { + DBG("not supported: format=%s, target=%d, sample_count=%d, " + "usage=%x, retval=%x", util_format_name(format), + target, sample_count, usage, retval); + } + + return retval == usage; +} + +void +fd5_screen_init(struct pipe_screen *pscreen) +{ + struct fd_screen *screen = fd_screen(pscreen); + screen->max_rts = A5XX_MAX_RENDER_TARGETS; + screen->compiler = ir3_compiler_create(screen->dev, screen->gpu_id); + pscreen->context_create = fd5_context_create; + pscreen->is_format_supported = fd5_screen_is_format_supported; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_screen.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_screen.h new file mode 100644 index 000000000..ba0c7f15f --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_screen.h @@ -0,0 +1,34 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_SCREEN_H_ +#define FD5_SCREEN_H_ + +#include "pipe/p_screen.h" + +void fd5_screen_init(struct pipe_screen *pscreen); + +#endif /* FD5_SCREEN_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_texture.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_texture.c new file mode 100644 index 000000000..87b69ea1c --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_texture.c @@ -0,0 +1,348 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" +#include "util/u_inlines.h" +#include "util/u_format.h" + +#include "fd5_texture.h" +#include "fd5_format.h" + +static enum a5xx_tex_clamp +tex_clamp(unsigned wrap, bool clamp_to_edge, bool *needs_border) +{ + /* Hardware does not support _CLAMP, but we emulate it: */ + if (wrap == PIPE_TEX_WRAP_CLAMP) { + wrap = (clamp_to_edge) ? + PIPE_TEX_WRAP_CLAMP_TO_EDGE : PIPE_TEX_WRAP_CLAMP_TO_BORDER; + } + + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + return A5XX_TEX_REPEAT; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return A5XX_TEX_CLAMP_TO_EDGE; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + *needs_border = true; + return A5XX_TEX_CLAMP_TO_BORDER; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + /* only works for PoT.. need to emulate otherwise! */ + return A5XX_TEX_MIRROR_CLAMP; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return A5XX_TEX_MIRROR_REPEAT; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + /* these two we could perhaps emulate, but we currently + * just don't advertise PIPE_CAP_TEXTURE_MIRROR_CLAMP + */ + default: + DBG("invalid wrap: %u", wrap); + return 0; + } +} + +static enum a5xx_tex_filter +tex_filter(unsigned filter, bool aniso) +{ + switch (filter) { + case PIPE_TEX_FILTER_NEAREST: + return A5XX_TEX_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return aniso ? A5XX_TEX_ANISO : A5XX_TEX_LINEAR; + default: + DBG("invalid filter: %u", filter); + return 0; + } +} + +static void * +fd5_sampler_state_create(struct pipe_context *pctx, + const struct pipe_sampler_state *cso) +{ + struct fd5_sampler_stateobj *so = CALLOC_STRUCT(fd5_sampler_stateobj); + unsigned aniso = util_last_bit(MIN2(cso->max_anisotropy >> 1, 8)); + bool miplinear = false; + bool clamp_to_edge; + + if (!so) + return NULL; + + so->base = *cso; + + if (cso->min_mip_filter == PIPE_TEX_MIPFILTER_LINEAR) + miplinear = true; + + /* + * For nearest filtering, _CLAMP means _CLAMP_TO_EDGE; for linear + * filtering, _CLAMP means _CLAMP_TO_BORDER while additionally + * clamping the texture coordinates to [0.0, 1.0]. + * + * The clamping will be taken care of in the shaders. There are two + * filters here, but let the minification one has a say. + */ + clamp_to_edge = (cso->min_img_filter == PIPE_TEX_FILTER_NEAREST); + if (!clamp_to_edge) { + so->saturate_s = (cso->wrap_s == PIPE_TEX_WRAP_CLAMP); + so->saturate_t = (cso->wrap_t == PIPE_TEX_WRAP_CLAMP); + so->saturate_r = (cso->wrap_r == PIPE_TEX_WRAP_CLAMP); + } + + so->needs_border = false; + so->texsamp0 = + COND(miplinear, A5XX_TEX_SAMP_0_MIPFILTER_LINEAR_NEAR) | + A5XX_TEX_SAMP_0_XY_MAG(tex_filter(cso->mag_img_filter, aniso)) | + A5XX_TEX_SAMP_0_XY_MIN(tex_filter(cso->min_img_filter, aniso)) | + A5XX_TEX_SAMP_0_ANISO(aniso) | + A5XX_TEX_SAMP_0_WRAP_S(tex_clamp(cso->wrap_s, clamp_to_edge, &so->needs_border)) | + A5XX_TEX_SAMP_0_WRAP_T(tex_clamp(cso->wrap_t, clamp_to_edge, &so->needs_border)) | + A5XX_TEX_SAMP_0_WRAP_R(tex_clamp(cso->wrap_r, clamp_to_edge, &so->needs_border)); + + so->texsamp1 = + COND(!cso->seamless_cube_map, A5XX_TEX_SAMP_1_CUBEMAPSEAMLESSFILTOFF) | + COND(!cso->normalized_coords, A5XX_TEX_SAMP_1_UNNORM_COORDS); + + if (cso->min_mip_filter != PIPE_TEX_MIPFILTER_NONE) { + so->texsamp0 |= A5XX_TEX_SAMP_0_LOD_BIAS(cso->lod_bias); + so->texsamp1 |= + A5XX_TEX_SAMP_1_MIN_LOD(cso->min_lod) | + A5XX_TEX_SAMP_1_MAX_LOD(cso->max_lod); + } + + if (cso->compare_mode) + so->texsamp1 |= A5XX_TEX_SAMP_1_COMPARE_FUNC(cso->compare_func); /* maps 1:1 */ + + return so; +} + +static void +fd5_sampler_states_bind(struct pipe_context *pctx, + enum pipe_shader_type shader, unsigned start, + unsigned nr, void **hwcso) +{ + struct fd_context *ctx = fd_context(pctx); + struct fd5_context *fd5_ctx = fd5_context(ctx); + uint16_t saturate_s = 0, saturate_t = 0, saturate_r = 0; + unsigned i; + + if (!hwcso) + nr = 0; + + for (i = 0; i < nr; i++) { + if (hwcso[i]) { + struct fd5_sampler_stateobj *sampler = + fd5_sampler_stateobj(hwcso[i]); + if (sampler->saturate_s) + saturate_s |= (1 << i); + if (sampler->saturate_t) + saturate_t |= (1 << i); + if (sampler->saturate_r) + saturate_r |= (1 << i); + } + } + + fd_sampler_states_bind(pctx, shader, start, nr, hwcso); + + if (shader == PIPE_SHADER_FRAGMENT) { + fd5_ctx->fsaturate = + (saturate_s != 0) || + (saturate_t != 0) || + (saturate_r != 0); + fd5_ctx->fsaturate_s = saturate_s; + fd5_ctx->fsaturate_t = saturate_t; + fd5_ctx->fsaturate_r = saturate_r; + } else if (shader == PIPE_SHADER_VERTEX) { + fd5_ctx->vsaturate = + (saturate_s != 0) || + (saturate_t != 0) || + (saturate_r != 0); + fd5_ctx->vsaturate_s = saturate_s; + fd5_ctx->vsaturate_t = saturate_t; + fd5_ctx->vsaturate_r = saturate_r; + } +} + +static enum a5xx_tex_type +tex_type(unsigned target) +{ + switch (target) { + default: + assert(0); + case PIPE_BUFFER: + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_1D_ARRAY: + return A5XX_TEX_1D; + case PIPE_TEXTURE_RECT: + case PIPE_TEXTURE_2D: + case PIPE_TEXTURE_2D_ARRAY: + return A5XX_TEX_2D; + case PIPE_TEXTURE_3D: + return A5XX_TEX_3D; + case PIPE_TEXTURE_CUBE: + case PIPE_TEXTURE_CUBE_ARRAY: + return A5XX_TEX_CUBE; + } +} + +static bool +use_astc_srgb_workaround(struct pipe_context *pctx, enum pipe_format format) +{ + return false; // TODO check if this is still needed on a5xx +} + +static struct pipe_sampler_view * +fd5_sampler_view_create(struct pipe_context *pctx, struct pipe_resource *prsc, + const struct pipe_sampler_view *cso) +{ + struct fd5_pipe_sampler_view *so = CALLOC_STRUCT(fd5_pipe_sampler_view); + struct fd_resource *rsc = fd_resource(prsc); + unsigned lvl, layers; + + if (!so) + return NULL; + + so->base = *cso; + pipe_reference(NULL, &prsc->reference); + so->base.texture = prsc; + so->base.reference.count = 1; + so->base.context = pctx; + + so->texconst0 = + A5XX_TEX_CONST_0_FMT(fd5_pipe2tex(cso->format)) | + fd5_tex_swiz(cso->format, cso->swizzle_r, cso->swizzle_g, + cso->swizzle_b, cso->swizzle_a); + + if (util_format_is_srgb(cso->format)) { + if (use_astc_srgb_workaround(pctx, cso->format)) + so->astc_srgb = true; + so->texconst0 |= A5XX_TEX_CONST_0_SRGB; + } + + if (cso->target == PIPE_BUFFER) { + unsigned elements = cso->u.buf.size / util_format_get_blocksize(cso->format); + + lvl = 0; + so->texconst1 = + A5XX_TEX_CONST_1_WIDTH(elements) | + A5XX_TEX_CONST_1_HEIGHT(1); + so->texconst2 = + A5XX_TEX_CONST_2_FETCHSIZE(fd5_pipe2fetchsize(cso->format)) | + A5XX_TEX_CONST_2_PITCH(elements * rsc->cpp); + so->offset = cso->u.buf.offset; + } else { + unsigned miplevels; + + lvl = fd_sampler_first_level(cso); + miplevels = fd_sampler_last_level(cso) - lvl; + layers = cso->u.tex.last_layer - cso->u.tex.first_layer + 1; + + so->texconst0 |= A5XX_TEX_CONST_0_MIPLVLS(miplevels); + so->texconst1 = + A5XX_TEX_CONST_1_WIDTH(u_minify(prsc->width0, lvl)) | + A5XX_TEX_CONST_1_HEIGHT(u_minify(prsc->height0, lvl)); + so->texconst2 = + A5XX_TEX_CONST_2_FETCHSIZE(fd5_pipe2fetchsize(cso->format)) | + A5XX_TEX_CONST_2_PITCH( + util_format_get_nblocksx( + cso->format, rsc->slices[lvl].pitch) * rsc->cpp); + so->offset = fd_resource_offset(rsc, lvl, cso->u.tex.first_layer); + } + + so->texconst2 |= A5XX_TEX_CONST_2_TYPE(tex_type(cso->target)); + + switch (cso->target) { + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: + so->texconst3 = + A5XX_TEX_CONST_3_ARRAY_PITCH(rsc->layer_size); + so->texconst5 = + A5XX_TEX_CONST_5_DEPTH(1); + break; + case PIPE_TEXTURE_1D_ARRAY: + case PIPE_TEXTURE_2D_ARRAY: + so->texconst3 = + A5XX_TEX_CONST_3_ARRAY_PITCH(rsc->layer_size); + so->texconst5 = + A5XX_TEX_CONST_5_DEPTH(layers); + break; + case PIPE_TEXTURE_CUBE: + case PIPE_TEXTURE_CUBE_ARRAY: + so->texconst3 = + A5XX_TEX_CONST_3_ARRAY_PITCH(rsc->layer_size); + so->texconst5 = + A5XX_TEX_CONST_5_DEPTH(layers / 6); + break; + case PIPE_TEXTURE_3D: + so->texconst3 = + A5XX_TEX_CONST_3_ARRAY_PITCH(rsc->slices[lvl].size0); + so->texconst5 = + A5XX_TEX_CONST_5_DEPTH(u_minify(prsc->depth0, lvl)); + break; + default: + so->texconst3 = 0x00000000; + break; + } + + return &so->base; +} + +static void +fd5_set_sampler_views(struct pipe_context *pctx, enum pipe_shader_type shader, + unsigned start, unsigned nr, + struct pipe_sampler_view **views) +{ + struct fd_context *ctx = fd_context(pctx); + struct fd5_context *fd5_ctx = fd5_context(ctx); + uint16_t astc_srgb = 0; + unsigned i; + + for (i = 0; i < nr; i++) { + if (views[i]) { + struct fd5_pipe_sampler_view *view = + fd5_pipe_sampler_view(views[i]); + if (view->astc_srgb) + astc_srgb |= (1 << i); + } + } + + fd_set_sampler_views(pctx, shader, start, nr, views); + + if (shader == PIPE_SHADER_FRAGMENT) { + fd5_ctx->fastc_srgb = astc_srgb; + } else if (shader == PIPE_SHADER_VERTEX) { + fd5_ctx->vastc_srgb = astc_srgb; + } +} + +void +fd5_texture_init(struct pipe_context *pctx) +{ + pctx->create_sampler_state = fd5_sampler_state_create; + pctx->bind_sampler_states = fd5_sampler_states_bind; + pctx->create_sampler_view = fd5_sampler_view_create; + pctx->set_sampler_views = fd5_set_sampler_views; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_texture.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_texture.h new file mode 100644 index 000000000..c4d109376 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_texture.h @@ -0,0 +1,70 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_TEXTURE_H_ +#define FD5_TEXTURE_H_ + +#include "pipe/p_context.h" + +#include "freedreno_texture.h" +#include "freedreno_resource.h" + +#include "fd5_context.h" +#include "fd5_format.h" + +struct fd5_sampler_stateobj { + struct pipe_sampler_state base; + uint32_t texsamp0, texsamp1, texsamp2, texsamp3; + bool saturate_s, saturate_t, saturate_r; + bool needs_border; +}; + +static inline struct fd5_sampler_stateobj * +fd5_sampler_stateobj(struct pipe_sampler_state *samp) +{ + return (struct fd5_sampler_stateobj *)samp; +} + +struct fd5_pipe_sampler_view { + struct pipe_sampler_view base; + uint32_t texconst0, texconst1, texconst2, texconst3, texconst5; + uint32_t texconst6, texconst7, texconst8, texconst9, texconst10, texconst11; + uint32_t offset; + bool astc_srgb; +}; + +static inline struct fd5_pipe_sampler_view * +fd5_pipe_sampler_view(struct pipe_sampler_view *pview) +{ + return (struct fd5_pipe_sampler_view *)pview; +} + +unsigned fd5_get_const_idx(struct fd_context *ctx, + struct fd_texture_stateobj *tex, unsigned samp_id); + +void fd5_texture_init(struct pipe_context *pctx); + +#endif /* FD5_TEXTURE_H_ */ diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_zsa.c b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_zsa.c new file mode 100644 index 000000000..495a4cc8a --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_zsa.c @@ -0,0 +1,119 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + + +#include "pipe/p_state.h" +#include "util/u_string.h" +#include "util/u_memory.h" + +#include "fd5_zsa.h" +#include "fd5_context.h" +#include "fd5_format.h" + +void * +fd5_zsa_state_create(struct pipe_context *pctx, + const struct pipe_depth_stencil_alpha_state *cso) +{ + struct fd5_zsa_stateobj *so; + + so = CALLOC_STRUCT(fd5_zsa_stateobj); + if (!so) + return NULL; + + so->base = *cso; + + switch (cso->depth.func) { + case PIPE_FUNC_LESS: + case PIPE_FUNC_LEQUAL: + so->gras_lrz_cntl = A5XX_GRAS_LRZ_CNTL_ENABLE; + break; + + case PIPE_FUNC_GREATER: + case PIPE_FUNC_GEQUAL: + so->gras_lrz_cntl = A5XX_GRAS_LRZ_CNTL_ENABLE | A5XX_GRAS_LRZ_CNTL_GREATER; + break; + + default: + /* LRZ not enabled */ + so->gras_lrz_cntl = 0; + break; + } + + if (!(cso->stencil->enabled || cso->alpha.enabled || !cso->depth.writemask)) + so->lrz_write = true; + + so->rb_depth_cntl |= + A5XX_RB_DEPTH_CNTL_ZFUNC(cso->depth.func); /* maps 1:1 */ + + if (cso->depth.enabled) + so->rb_depth_cntl |= + A5XX_RB_DEPTH_CNTL_Z_ENABLE | + A5XX_RB_DEPTH_CNTL_Z_TEST_ENABLE; + + if (cso->depth.writemask) + so->rb_depth_cntl |= A5XX_RB_DEPTH_CNTL_Z_WRITE_ENABLE; + + if (cso->stencil[0].enabled) { + const struct pipe_stencil_state *s = &cso->stencil[0]; + + so->rb_stencil_control |= + A5XX_RB_STENCIL_CONTROL_STENCIL_READ | + A5XX_RB_STENCIL_CONTROL_STENCIL_ENABLE | + A5XX_RB_STENCIL_CONTROL_FUNC(s->func) | /* maps 1:1 */ + A5XX_RB_STENCIL_CONTROL_FAIL(fd_stencil_op(s->fail_op)) | + A5XX_RB_STENCIL_CONTROL_ZPASS(fd_stencil_op(s->zpass_op)) | + A5XX_RB_STENCIL_CONTROL_ZFAIL(fd_stencil_op(s->zfail_op)); + so->rb_stencilrefmask |= + A5XX_RB_STENCILREFMASK_STENCILWRITEMASK(s->writemask) | + A5XX_RB_STENCILREFMASK_STENCILMASK(s->valuemask); + + if (cso->stencil[1].enabled) { + const struct pipe_stencil_state *bs = &cso->stencil[1]; + + so->rb_stencil_control |= + A5XX_RB_STENCIL_CONTROL_STENCIL_ENABLE_BF | + A5XX_RB_STENCIL_CONTROL_FUNC_BF(bs->func) | /* maps 1:1 */ + A5XX_RB_STENCIL_CONTROL_FAIL_BF(fd_stencil_op(bs->fail_op)) | + A5XX_RB_STENCIL_CONTROL_ZPASS_BF(fd_stencil_op(bs->zpass_op)) | + A5XX_RB_STENCIL_CONTROL_ZFAIL_BF(fd_stencil_op(bs->zfail_op)); + so->rb_stencilrefmask_bf |= + A5XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK(bs->writemask) | + A5XX_RB_STENCILREFMASK_BF_STENCILMASK(bs->valuemask); + } + } + + if (cso->alpha.enabled) { + uint32_t ref = cso->alpha.ref_value * 255.0; + so->rb_alpha_control = + A5XX_RB_ALPHA_CONTROL_ALPHA_TEST | + A5XX_RB_ALPHA_CONTROL_ALPHA_REF(ref) | + A5XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC(cso->alpha.func); +// so->rb_depth_control |= +// A5XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE; + } + + return so; +} diff --git a/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_zsa.h b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_zsa.h new file mode 100644 index 000000000..c15ba1aa8 --- /dev/null +++ b/lib/mesa/src/gallium/drivers/freedreno/a5xx/fd5_zsa.h @@ -0,0 +1,57 @@ +/* + * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Authors: + * Rob Clark <robclark@freedesktop.org> + */ + +#ifndef FD5_ZSA_H_ +#define FD5_ZSA_H_ + + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + +#include "freedreno_util.h" + +struct fd5_zsa_stateobj { + struct pipe_depth_stencil_alpha_state base; + + uint32_t rb_alpha_control; + uint32_t rb_depth_cntl; + uint32_t rb_stencil_control; + uint32_t rb_stencilrefmask; + uint32_t rb_stencilrefmask_bf; + uint32_t gras_lrz_cntl; + bool lrz_write; +}; + +static inline struct fd5_zsa_stateobj * +fd5_zsa_stateobj(struct pipe_depth_stencil_alpha_state *zsa) +{ + return (struct fd5_zsa_stateobj *)zsa; +} + +void * fd5_zsa_state_create(struct pipe_context *pctx, + const struct pipe_depth_stencil_alpha_state *cso); + +#endif /* FD5_ZSA_H_ */ |