summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--src/radeon.h5
-rw-r--r--src/radeon_accelfuncs.c5
-rw-r--r--src/radeon_commonfuncs.c5
-rw-r--r--src/radeon_exa_funcs.c5
-rw-r--r--src/radeon_textured_video.c108
-rw-r--r--src/radeon_textured_videofuncs.c323
-rw-r--r--src/radeon_video.c13
-rw-r--r--src/radeon_video.h12
8 files changed, 388 insertions, 88 deletions
diff --git a/src/radeon.h b/src/radeon.h
index 7e84aebc..174352d3 100644
--- a/src/radeon.h
+++ b/src/radeon.h
@@ -378,6 +378,11 @@ typedef enum {
(info->ChipFamily == CHIP_FAMILY_RS400) || \
(info->ChipFamily == CHIP_FAMILY_RS480))
+#define IS_R200_3D ((info->ChipFamily == CHIP_FAMILY_RV250) || \
+ (info->ChipFamily == CHIP_FAMILY_RV280) || \
+ (info->ChipFamily == CHIP_FAMILY_RS300) || \
+ (info->ChipFamily == CHIP_FAMILY_R200))
+
/*
* Errata workarounds
*/
diff --git a/src/radeon_accelfuncs.c b/src/radeon_accelfuncs.c
index 45eb6d55..2d6fe01d 100644
--- a/src/radeon_accelfuncs.c
+++ b/src/radeon_accelfuncs.c
@@ -1345,10 +1345,7 @@ FUNC_NAME(RADEONAccelInit)(ScreenPtr pScreen, XAAInfoRecPtr a)
xf86DrvMsg(pScrn->scrnIndex, X_INFO, "XAA Render acceleration "
"unsupported on Radeon 9500/9700 and newer. "
"Please use EXA instead.\n");
- } else if ((info->ChipFamily == CHIP_FAMILY_RV250) ||
- (info->ChipFamily == CHIP_FAMILY_RV280) ||
- (info->ChipFamily == CHIP_FAMILY_RS300) ||
- (info->ChipFamily == CHIP_FAMILY_R200)) {
+ } else if (IS_R200_3D) {
a->SetupForCPUToScreenAlphaTexture2 =
FUNC_NAME(R200SetupForCPUToScreenAlphaTexture);
a->SubsequentCPUToScreenAlphaTexture =
diff --git a/src/radeon_commonfuncs.c b/src/radeon_commonfuncs.c
index a9bc7d28..3dbe6173 100644
--- a/src/radeon_commonfuncs.c
+++ b/src/radeon_commonfuncs.c
@@ -565,10 +565,7 @@ static void FUNC_NAME(RADEONInit3DEngine)(ScrnInfoPtr pScrn)
OUT_ACCEL_REG(R300_SC_CLIP_RULE, 0xAAAA);
OUT_ACCEL_REG(R300_SC_SCREENDOOR, 0xffffff);
FINISH_ACCEL();
- } else if ((info->ChipFamily == CHIP_FAMILY_RV250) ||
- (info->ChipFamily == CHIP_FAMILY_RV280) ||
- (info->ChipFamily == CHIP_FAMILY_RS300) ||
- (info->ChipFamily == CHIP_FAMILY_R200)) {
+ } else if (IS_R200_3D) {
BEGIN_ACCEL(6);
if (info->ChipFamily == CHIP_FAMILY_RS300) {
diff --git a/src/radeon_exa_funcs.c b/src/radeon_exa_funcs.c
index 59cb46fc..6a2b25cf 100644
--- a/src/radeon_exa_funcs.c
+++ b/src/radeon_exa_funcs.c
@@ -505,10 +505,7 @@ Bool FUNC_NAME(RADEONDrawInit)(ScreenPtr pScreen)
info->accel_state->exa->DoneComposite = FUNC_NAME(RadeonDoneComposite);
} else
xf86DrvMsg(pScrn->scrnIndex, X_INFO, "EXA Composite requires CP on R5xx/IGP\n");
- } else if ((info->ChipFamily == CHIP_FAMILY_RV250) ||
- (info->ChipFamily == CHIP_FAMILY_RV280) ||
- (info->ChipFamily == CHIP_FAMILY_RS300) ||
- (info->ChipFamily == CHIP_FAMILY_R200)) {
+ } else if (IS_R200_3D) {
xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Render acceleration "
"enabled for R200 type cards.\n");
info->accel_state->exa->CheckComposite = R200CheckComposite;
diff --git a/src/radeon_textured_video.c b/src/radeon_textured_video.c
index 79671c07..bf8a2763 100644
--- a/src/radeon_textured_video.c
+++ b/src/radeon_textured_video.c
@@ -128,6 +128,21 @@ static __inline__ uint32_t float4touint(float fr, float fg, float fb, float fa)
return (ua << 24) | (ur << 16) | (ug << 8) | ub;
}
+/* Parameters for ITU-R BT.601 and ITU-R BT.709 colour spaces
+ note the difference to the parameters used in overlay are due
+ to 10bit vs. float calcs */
+static REF_TRANSFORM trans[2] =
+{
+ {1.1643, 0.0, 1.5960, -0.3918, -0.8129, 2.0172, 0.0}, /* BT.601 */
+ {1.1643, 0.0, 1.7927, -0.2132, -0.5329, 2.1124, 0.0} /* BT.709 */
+};
+
+
+#define RTFSaturation(a) (1.0 + ((a)*1.0)/1000.0)
+#define RTFBrightness(a) (((a)*1.0)/2000.0)
+#define RTFContrast(a) (1.0 + ((a)*1.0)/1000.0)
+#define RTFHue(a) (((a)*3.1416)/1000.0)
+
#define ACCEL_MMIO
#define ACCEL_PREAMBLE() unsigned char *RADEONMMIO = info->MMIO
#define BEGIN_ACCEL(n) RADEONWaitForFifo(pScrn, (n))
@@ -359,12 +374,8 @@ RADEONPutImageTextured(ScrnInfoPtr pScrn,
}
pPriv->planar_hw = pPriv->planar_state;
- if (pPriv->bicubic_enabled || !( IS_R300_3D ||
- (info->ChipFamily == CHIP_FAMILY_RV250) ||
- (info->ChipFamily == CHIP_FAMILY_RV280) ||
- (info->ChipFamily == CHIP_FAMILY_RS300) ||
- (info->ChipFamily == CHIP_FAMILY_R200) ))
- pPriv->planar_hw = 0;
+ if (pPriv->bicubic_enabled || !( IS_R300_3D || IS_R200_3D ))
+ pPriv->planar_hw = 0;
switch(id) {
case FOURCC_YV12:
@@ -636,28 +647,58 @@ static XF86VideoFormatRec Formats[NUM_FORMATS] =
{15, TrueColor}, {16, TrueColor}, {24, TrueColor}
};
-#define NUM_ATTRIBUTES 2
+#define NUM_ATTRIBUTES 1
static XF86AttributeRec Attributes[NUM_ATTRIBUTES+1] =
{
{XvSettable | XvGettable, 0, 1, "XV_VSYNC"},
+ {0, 0, 0, NULL}
+};
+
+#define NUM_ATTRIBUTES_R200 7
+
+static XF86AttributeRec Attributes_r200[NUM_ATTRIBUTES_R200+1] =
+{
+ {XvSettable | XvGettable, 0, 1, "XV_VSYNC"},
{XvSettable | XvGettable, 0, 1, "XV_HWPLANAR"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_BRIGHTNESS"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_CONTRAST"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_SATURATION"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_HUE"},
+ {XvSettable | XvGettable, 100, 10000, "XV_COLORSPACE"},
{0, 0, 0, NULL}
};
-#define NUM_ATTRIBUTES_R300 3
+#define NUM_ATTRIBUTES_R300 9
static XF86AttributeRec Attributes_r300[NUM_ATTRIBUTES_R300+1] =
{
{XvSettable | XvGettable, 0, 2, "XV_BICUBIC"},
{XvSettable | XvGettable, 0, 1, "XV_VSYNC"},
{XvSettable | XvGettable, 0, 1, "XV_HWPLANAR"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_BRIGHTNESS"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_CONTRAST"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_SATURATION"},
+ {XvSettable | XvGettable, -1000, 1000, "XV_HUE"},
+ {XvSettable | XvGettable, 100, 10000, "XV_GAMMA"},
+ {XvSettable | XvGettable, 0, 1, "XV_COLORSPACE"},
+ {0, 0, 0, NULL}
+};
+
+#define NUM_ATTRIBUTES_R500 2
+
+static XF86AttributeRec Attributes_r500[NUM_ATTRIBUTES_R500+1] =
+{
+ {XvSettable | XvGettable, 0, 2, "XV_BICUBIC"},
+ {XvSettable | XvGettable, 0, 1, "XV_VSYNC"},
{0, 0, 0, NULL}
};
static Atom xvBicubic;
static Atom xvVSync;
static Atom xvHWPlanar;
+static Atom xvBrightness, xvContrast, xvSaturation, xvHue;
+static Atom xvGamma, xvColorspace;
#define NUM_IMAGES 4
@@ -686,6 +727,18 @@ RADEONGetTexPortAttribute(ScrnInfoPtr pScrn,
*value = pPriv->vsync;
else if (attribute == xvHWPlanar)
*value = pPriv->planar_state;
+ else if (attribute == xvBrightness)
+ *value = pPriv->brightness;
+ else if (attribute == xvContrast)
+ *value = pPriv->contrast;
+ else if (attribute == xvSaturation)
+ *value = pPriv->saturation;
+ else if (attribute == xvHue)
+ *value = pPriv->hue;
+ else if (attribute == xvGamma)
+ *value = pPriv->gamma;
+ else if(attribute == xvColorspace)
+ *value = pPriv->transform_index;
else
return BadMatch;
@@ -709,6 +762,20 @@ RADEONSetTexPortAttribute(ScrnInfoPtr pScrn,
pPriv->vsync = ClipValue (value, 0, 1);
else if (attribute == xvHWPlanar)
pPriv->planar_state = ClipValue (value, 0, 1);
+ else if (attribute == xvHWPlanar)
+ pPriv->planar_state = ClipValue (value, 0, 1);
+ else if (attribute == xvBrightness)
+ pPriv->brightness = ClipValue (value, -1000, 1000);
+ else if (attribute == xvContrast)
+ pPriv->contrast = ClipValue (value, -1000, 1000);
+ else if (attribute == xvSaturation)
+ pPriv->saturation = ClipValue (value, -1000, 1000);
+ else if (attribute == xvHue)
+ pPriv->hue = ClipValue (value, -1000, 1000);
+ else if (attribute == xvGamma)
+ pPriv->gamma = ClipValue (value, 100, 10000);
+ else if(attribute == xvColorspace)
+ pPriv->transform_index = ClipValue (value, 0, 1);
else
return BadMatch;
@@ -733,6 +800,12 @@ RADEONSetupImageTexturedVideo(ScreenPtr pScreen)
xvBicubic = MAKE_ATOM("XV_BICUBIC");
xvVSync = MAKE_ATOM("XV_VSYNC");
xvHWPlanar = MAKE_ATOM("XV_HWPLANAR");
+ xvBrightness = MAKE_ATOM("XV_BRIGHTNESS");
+ xvContrast = MAKE_ATOM("XV_CONTRAST");
+ xvSaturation = MAKE_ATOM("XV_SATURATION");
+ xvHue = MAKE_ATOM("XV_HUE");
+ xvGamma = MAKE_ATOM("XV_GAMMA");
+ xvColorspace = MAKE_ATOM("XV_COLORSPACE");
adapt->type = XvWindowMask | XvInputMask | XvImageMask;
adapt->flags = 0;
@@ -752,10 +825,19 @@ RADEONSetupImageTexturedVideo(ScreenPtr pScreen)
pPortPriv =
(RADEONPortPrivPtr)(&adapt->pPortPrivates[num_texture_ports]);
- if (IS_R300_3D || IS_R500_3D) {
+ if (IS_R300_3D) {
adapt->pAttributes = Attributes_r300;
adapt->nAttributes = NUM_ATTRIBUTES_R300;
- } else {
+ }
+ else if (IS_R500_3D) {
+ adapt->pAttributes = Attributes_r500;
+ adapt->nAttributes = NUM_ATTRIBUTES_R500;
+ }
+ else if (IS_R200_3D) {
+ adapt->pAttributes = Attributes_r200;
+ adapt->nAttributes = NUM_ATTRIBUTES_R200;
+ }
+ else {
adapt->pAttributes = Attributes;
adapt->nAttributes = NUM_ATTRIBUTES;
}
@@ -783,6 +865,12 @@ RADEONSetupImageTexturedVideo(ScreenPtr pScreen)
pPriv->bicubic_state = BICUBIC_AUTO;
pPriv->vsync = TRUE;
pPriv->planar_state = 1;
+ pPriv->brightness = 0;
+ pPriv->contrast = 0;
+ pPriv->saturation = 0;
+ pPriv->hue = 0;
+ pPriv->gamma = 1000;
+ pPriv->transform_index = 0;
/* gotta uninit this someplace, XXX: shouldn't be necessary for textured */
REGION_NULL(pScreen, &pPriv->clip);
diff --git a/src/radeon_textured_videofuncs.c b/src/radeon_textured_videofuncs.c
index 6cb2870a..3c4289f0 100644
--- a/src/radeon_textured_videofuncs.c
+++ b/src/radeon_textured_videofuncs.c
@@ -743,9 +743,10 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
* DP3 might look like the straightforward solution
* but we'd need to move the texture yuv values in
* the same reg for this to work. Therefore use MADs.
- * Without changing the shader at all (only the constants)
- * could also provide hue/saturation/brightness/contrast control.
- *
+ * Brightness just adds to the off constant.
+ * Contrast is multiplication of luminance.
+ * Saturation and hue change the u and v coeffs.
+ * Default values (before adjustments - depend on colorspace):
* yco = 1.1643
* uco = 0, -0.39173, 2.017
* vco = 1.5958, -0.8129, 0
@@ -757,14 +758,46 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
* temp = MAD(uco, yuv.uuuu, temp)
* result = MAD(vco, yuv.vvvv, temp)
*/
- float yco = 1.1643;
- float uco[3] = {0.0, -0.39173, 2.018};
- float vco[3] = {1.5958, -0.8129, 0.0};
- float off[3] = {-0.0625 * yco + -0.5 * uco[0] + -0.5 * vco[0],
- -0.0625 * yco + -0.5 * uco[1] + -0.5 * vco[1],
- -0.0625 * yco + -0.5 * uco[2] + -0.5 * vco[2]};
-
- BEGIN_ACCEL(33);
+ /* TODO: don't recalc consts always */
+ const float Loff = -0.0627;
+ const float Coff = -0.502;
+ float uvcosf, uvsinf;
+ float yco;
+ float uco[3], vco[3], off[3];
+ float bright, cont, gamma;
+ int ref = pPriv->transform_index;
+ Bool needgamma = FALSE;
+
+ cont = RTFContrast(pPriv->contrast);
+ bright = RTFBrightness(pPriv->brightness);
+ gamma = (float)pPriv->gamma / 1000.0;
+ uvcosf = RTFSaturation(pPriv->saturation) * cos(RTFHue(pPriv->hue));
+ uvsinf = RTFSaturation(pPriv->saturation) * sin(RTFHue(pPriv->hue));
+ /* overlay video also does pre-gamma contrast/sat adjust, should we? */
+
+ yco = trans[ref].RefLuma * cont;
+ uco[0] = -trans[ref].RefRCr * uvsinf;
+ uco[1] = trans[ref].RefGCb * uvcosf - trans[ref].RefGCr * uvsinf;
+ uco[2] = trans[ref].RefBCb * uvcosf;
+ vco[0] = trans[ref].RefRCr * uvcosf;
+ vco[1] = trans[ref].RefGCb * uvsinf + trans[ref].RefGCr * uvcosf;
+ vco[2] = trans[ref].RefBCb * uvsinf;
+ off[0] = Loff * yco + Coff * (uco[0] + vco[0]) + bright;
+ off[1] = Loff * yco + Coff * (uco[1] + vco[1]) + bright;
+ off[2] = Loff * yco + Coff * (uco[2] + vco[2]) + bright;
+
+ if (gamma != 1.0) {
+ needgamma = TRUE;
+ /* note: gamma correction is out = in ^ gamma;
+ gpu can only do LG2/EX2 therefore we transform into
+ in ^ gamma = 2 ^ (log2(in) * gamma).
+ Lots of scalar ops, unfortunately (better solution?) -
+ without gamma that's 3 inst, with gamma it's 10...
+ could use different gamma factors per channel,
+ if that's of any use. */
+ }
+
+ BEGIN_ACCEL(needgamma ? 28 + 33 : 33);
/* 2 components: same 2 for tex0/1/2 */
OUT_ACCEL_REG(R300_RS_COUNT,
((2 << R300_RS_COUNT_IT_COUNT_SHIFT) |
@@ -779,12 +812,12 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R300_FIRST_TEX));
OUT_ACCEL_REG(R300_US_CODE_OFFSET, (R300_ALU_CODE_OFFSET(0) |
- R300_ALU_CODE_SIZE(3) |
+ R300_ALU_CODE_SIZE(needgamma ? 7 + 3 : 3) |
R300_TEX_CODE_OFFSET(0) |
R300_TEX_CODE_SIZE(3)));
OUT_ACCEL_REG(R300_US_CODE_ADDR_3, (R300_ALU_START(0) |
- R300_ALU_SIZE(2) |
+ R300_ALU_SIZE(needgamma ? 7 + 2 : 2) |
R300_TEX_START(0) |
R300_TEX_SIZE(2) |
R300_RGBA_OUT));
@@ -857,7 +890,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R300_ALU_RGB_ADDR2(0) |
R300_ALU_RGB_ADDRD(0) |
R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_RGB) |
- R300_ALU_RGB_OMASK(R300_ALU_RGB_MASK_RGB)));
+ (needgamma ? 0 : R300_ALU_RGB_OMASK(R300_ALU_RGB_MASK_RGB))));
OUT_ACCEL_REG(R300_US_ALU_RGB_INST(2), (R300_ALU_RGB_SEL_A(R300_ALU_RGB_SRC0_RGB) |
R300_ALU_RGB_MOD_A(R300_ALU_RGB_MOD_NOP) |
R300_ALU_RGB_SEL_B(R300_ALU_RGB_SRC1_RGB) |
@@ -868,14 +901,126 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE) |
R300_ALU_RGB_CLAMP));
/* write alpha 1 */
- OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(4), (R300_ALU_ALPHA_ADDRD(0) |
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(2), (R300_ALU_ALPHA_ADDRD(0) |
R300_ALU_ALPHA_OMASK(R300_ALU_ALPHA_MASK_A) |
R300_ALU_ALPHA_TARGET_A));
- OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(4), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_MAD) |
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(2), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_MAD) |
R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_0_0) |
R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_1_0)));
+ if (needgamma) {
+ /* rgb temp0.r = op_sop, set up src0 reg */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(3), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_R)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(3),
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_SOP) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE));
+ /* alpha lg2 temp0, temp0.r */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(3), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(3), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_LN2) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_SRC0_R) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+
+ /* rgb temp0.g = op_sop, set up src0 reg */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(4), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_G)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(4),
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_SOP) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE));
+ /* alpha lg2 temp0, temp0.g */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(4), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(4), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_LN2) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_SRC0_G) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+
+ /* rgb temp0.b = op_sop, set up src0 reg */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(5), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_B)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(5),
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_SOP) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE));
+ /* alpha lg2 temp0, temp0.b */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(5), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(5), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_LN2) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_SRC0_B) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+
+ /* MUL const1, temp1, temp0 */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(6), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_ADDR1(0) |
+ R300_ALU_RGB_ADDR2(0) |
+ R300_ALU_RGB_ADDRD(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_RGB)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(6), (R300_ALU_RGB_SEL_A(R300_ALU_RGB_SRC0_RGB) |
+ R300_ALU_RGB_MOD_A(R300_ALU_RGB_MOD_NOP) |
+ R300_ALU_RGB_SEL_B(R300_ALU_RGB_SRC0_AAA) |
+ R300_ALU_RGB_MOD_B(R300_ALU_RGB_MOD_NOP) |
+ R300_ALU_RGB_SEL_C(R300_ALU_RGB_0_0) |
+ R300_ALU_RGB_MOD_C(R300_ALU_RGB_MOD_NOP) |
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_MAD) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE)));
+ /* alpha nop, but set up const1 */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(6), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_ADDR0(R300_ALU_ALPHA_CONST(1)) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(6), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_MAD) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+
+ /* rgb out0.r = op_sop, set up src0 reg */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(7), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_R) |
+ R300_ALU_RGB_OMASK(R300_ALU_RGB_MASK_R)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(7),
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_SOP) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE));
+ /* alpha ex2 temp0, temp0.r */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(7), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(7), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_EX2) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_SRC0_R) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+
+ /* rgb out0.g = op_sop, set up src0 reg */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(8), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_G) |
+ R300_ALU_RGB_OMASK(R300_ALU_RGB_MASK_G)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(8),
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_SOP) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE));
+ /* alpha ex2 temp0, temp0.g */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(8), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(8), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_EX2) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_SRC0_G) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+
+ /* rgb out0.b = op_sop, set up src0 reg */
+ OUT_ACCEL_REG(R300_US_ALU_RGB_ADDR(9), (R300_ALU_RGB_ADDR0(0) |
+ R300_ALU_RGB_WMASK(R300_ALU_RGB_MASK_B) |
+ R300_ALU_RGB_OMASK(R300_ALU_RGB_MASK_B)));
+ OUT_ACCEL_REG(R300_US_ALU_RGB_INST(9),
+ R300_ALU_RGB_OP(R300_ALU_RGB_OP_SOP) |
+ R300_ALU_RGB_OMOD(R300_ALU_RGB_OMOD_NONE));
+ /* alpha ex2 temp0, temp0.b */
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_ADDR(9), (R300_ALU_ALPHA_ADDRD(0) |
+ R300_ALU_ALPHA_WMASK(R300_ALU_ALPHA_MASK_NONE)));
+ OUT_ACCEL_REG(R300_US_ALU_ALPHA_INST(9), (R300_ALU_ALPHA_OP(R300_ALU_ALPHA_OP_EX2) |
+ R300_ALU_ALPHA_SEL_A(R300_ALU_ALPHA_SRC0_B) |
+ R300_ALU_ALPHA_SEL_B(R300_ALU_ALPHA_0_0) |
+ R300_ALU_ALPHA_SEL_C(R300_ALU_ALPHA_0_0)));
+ }
+
/* Shader constants. */
/* constant 0: off, yco */
OUT_ACCEL_REG(R300_US_ALU_CONST_R(0), F_TO_24(off[0]));
@@ -886,7 +1031,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
OUT_ACCEL_REG(R300_US_ALU_CONST_R(1), F_TO_24(uco[0]));
OUT_ACCEL_REG(R300_US_ALU_CONST_G(1), F_TO_24(uco[1]));
OUT_ACCEL_REG(R300_US_ALU_CONST_B(1), F_TO_24(uco[2]));
- OUT_ACCEL_REG(R300_US_ALU_CONST_A(1), F_TO_24(0.0));
+ OUT_ACCEL_REG(R300_US_ALU_CONST_A(1), F_TO_24(gamma));
/* constant 2: vco */
OUT_ACCEL_REG(R300_US_ALU_CONST_R(2), F_TO_24(vco[0]));
OUT_ACCEL_REG(R300_US_ALU_CONST_G(2), F_TO_24(vco[1]));
@@ -1601,20 +1746,52 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
FINISH_ACCEL();
- if ((info->ChipFamily == CHIP_FAMILY_RV250) ||
- (info->ChipFamily == CHIP_FAMILY_RV280) ||
- (info->ChipFamily == CHIP_FAMILY_RS300) ||
- (info->ChipFamily == CHIP_FAMILY_R200)) {
+ if (IS_R200_3D) {
info->accel_state->texW[0] = pPriv->w;
info->accel_state->texH[0] = pPriv->h;
if (isplanar) {
/* note: in contrast to r300, use input biasing on uv components */
- float yco = 1.1643;
- float yoff = -0.0625 * yco;
- float uco[3] = {0.0, -0.39173, 2.018};
- float vco[3] = {1.5958, -0.8129, 0.0};
+ const float Loff = -0.0627;
+ float uvcosf, uvsinf;
+ float yco, yoff;
+ float uco[3], vco[3];
+ float bright, cont, sat;
+ int ref = pPriv->transform_index;
+ float ucscale = 0.25, vcscale = 0.25;
+ Bool needux8 = FALSE, needvx8 = FALSE;
+
+ /* contrast can cause constant overflow, clamp */
+ cont = RTFContrast(pPriv->contrast);
+ if (cont * trans[ref].RefLuma > 2.0)
+ cont = 2.0 / trans[ref].RefLuma;
+ /* brightness is only from -0.5 to 0.5 should be safe */
+ bright = RTFBrightness(pPriv->brightness);
+ /* saturation can also cause overflow, clamp */
+ sat = RTFSaturation(pPriv->saturation);
+ if (sat * trans[ref].RefBCb > 4.0)
+ sat = 4.0 / trans[ref].RefBCb;
+ uvcosf = sat * cos(RTFHue(pPriv->hue));
+ uvsinf = sat * sin(RTFHue(pPriv->hue));
+
+ yco = trans[ref].RefLuma * cont;
+ uco[0] = -trans[ref].RefRCr * uvsinf;
+ uco[1] = trans[ref].RefGCb * uvcosf - trans[ref].RefGCr * uvsinf;
+ uco[2] = trans[ref].RefBCb * uvcosf;
+ vco[0] = trans[ref].RefRCr * uvcosf;
+ vco[1] = trans[ref].RefGCb * uvsinf + trans[ref].RefGCr * uvcosf;
+ vco[2] = trans[ref].RefBCb * uvsinf;
+ yoff = Loff * yco + bright;
+
+ if ((uco[0] > 2.0) || (uco[2] > 2.0)) {
+ needux8 = TRUE;
+ ucscale = 0.125;
+ }
+ if ((vco[0] > 2.0) || (vco[2] > 2.0)) {
+ needvx8 = TRUE;
+ vcscale = 0.125;
+ }
/* need 2 texcoord sets (even though they are identical) due
to denormalization! hw apparently can't premultiply
@@ -1678,7 +1855,9 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
* seems the values we need seem to fit better than worst case (get about
* 6 fractional bits for this instead of 5, at least when not correcting for
* hue/saturation/contrast/brightness, which is the same as for vco - yco and
- * yoff get 8 fractional bits).
+ * yoff get 8 fractional bits). Try to preserve as much accuracy as possible
+ * even with non-default saturation/hue/contrast/brightness adjustments,
+ * it gets a little crazy and ultimately precision might still be lacking.
*
* A higher precision (8 fractional bits) version might just put uco into
* a texcoord, and calculate a new vcoconst in the shader, like so:
@@ -1709,7 +1888,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXC_ARG_A_TFACTOR_COLOR |
R200_TXC_ARG_B_R0_COLOR |
R200_TXC_ARG_C_TFACTOR_COLOR |
- R200_TXC_NEG_ARG_C |
+ (yoff < 0 ? R200_TXC_NEG_ARG_C : 0) |
R200_TXC_OP_DOT2_ADD);
OUT_ACCEL_REG(R200_PP_TXCBLEND2_0,
(0 << R200_TXC_TFACTOR_SEL_SHIFT) |
@@ -1730,7 +1909,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXC_SCALE_ARG_A |
R200_TXC_ARG_B_R1_COLOR |
R200_TXC_BIAS_ARG_B |
- R200_TXC_SCALE_ARG_B |
+ (needux8 ? R200_TXC_SCALE_ARG_B : 0) |
R200_TXC_ARG_C_R0_COLOR |
R200_TXC_OP_MADD);
OUT_ACCEL_REG(R200_PP_TXCBLEND2_1,
@@ -1751,6 +1930,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXC_SCALE_ARG_A |
R200_TXC_ARG_B_R2_COLOR |
R200_TXC_BIAS_ARG_B |
+ (needvx8 ? R200_TXC_SCALE_ARG_B : 0) |
R200_TXC_ARG_C_R0_COLOR |
R200_TXC_OP_MADD);
OUT_ACCEL_REG(R200_PP_TXCBLEND2_2,
@@ -1767,28 +1947,64 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXA_CLAMP_0_1 | R200_TXA_OUTPUT_REG_R0);
/* shader constants */
- OUT_ACCEL_REG(R200_PP_TFACTOR_0, float4touint(1.0, /* src range [1, 2] */
- yco - 1.0,
- -yoff, /* range [-1, 0] */
+ OUT_ACCEL_REG(R200_PP_TFACTOR_0, float4touint(yco > 1.0 ? 1.0 : 0.0, /* range special [0, 2] */
+ yco > 1.0 ? yco - 1.0: yco,
+ yoff < 0 ? -yoff : yoff, /* range special [-1, 1] */
0.0));
- OUT_ACCEL_REG(R200_PP_TFACTOR_1, float4touint(uco[0] * 0.125 + 0.5, /* range [-4, 4] */
- uco[1] * 0.125 + 0.5,
- uco[2] * 0.125 + 0.5,
+ OUT_ACCEL_REG(R200_PP_TFACTOR_1, float4touint(uco[0] * ucscale + 0.5, /* range [-4, 4] */
+ uco[1] * ucscale + 0.5, /* or [-2, 2] */
+ uco[2] * ucscale + 0.5,
0.0));
- OUT_ACCEL_REG(R200_PP_TFACTOR_2, float4touint(vco[0] * 0.25 + 0.5, /* range [-2, 2] */
- vco[1] * 0.25 + 0.5,
- vco[2] * 0.25 + 0.5,
+ OUT_ACCEL_REG(R200_PP_TFACTOR_2, float4touint(vco[0] * vcscale + 0.5, /* range [-2, 2] */
+ vco[1] * vcscale + 0.5, /* or [-4, 4] */
+ vco[2] * vcscale + 0.5,
0.0));
FINISH_ACCEL();
}
else if (info->ChipFamily == CHIP_FAMILY_RV250) {
/* fix up broken packed yuv - shader same as above except
- yuv compoents are all in same reg */
- float yco = 1.1643;
- float yoff = -0.0625 * yco;
- float uco[3] = {0.0, -0.39173, 2.018};
- float vco[3] = {1.5958, -0.8129, 0.0};
+ yuv components are all in same reg */
+ /* note: in contrast to r300, use input biasing on uv components */
+ const float Loff = -0.0627;
+ float uvcosf, uvsinf;
+ float yco, yoff;
+ float uco[3], vco[3];
+ float bright, cont, sat;
+ int ref = pPriv->transform_index;
+ float ucscale = 0.25, vcscale = 0.25;
+ Bool needux8 = FALSE, needvx8 = FALSE;
+
+ /* contrast can cause constant overflow, clamp */
+ cont = RTFContrast(pPriv->contrast);
+ if (cont * trans[ref].RefLuma > 2.0)
+ cont = 2.0 / trans[ref].RefLuma;
+ /* brightness is only from -0.5 to 0.5 should be safe */
+ bright = RTFBrightness(pPriv->brightness);
+ /* saturation can also cause overflow, clamp */
+ sat = RTFSaturation(pPriv->saturation);
+ if (sat * trans[ref].RefBCb > 4.0)
+ sat = 4.0 / trans[ref].RefBCb;
+ uvcosf = sat * cos(RTFHue(pPriv->hue));
+ uvsinf = sat * sin(RTFHue(pPriv->hue));
+
+ yco = trans[ref].RefLuma * cont;
+ uco[0] = -trans[ref].RefRCr * uvsinf;
+ uco[1] = trans[ref].RefGCb * uvcosf - trans[ref].RefGCr * uvsinf;
+ uco[2] = trans[ref].RefBCb * uvcosf;
+ vco[0] = trans[ref].RefRCr * uvcosf;
+ vco[1] = trans[ref].RefGCb * uvsinf + trans[ref].RefGCr * uvcosf;
+ vco[2] = trans[ref].RefBCb * uvsinf;
+ yoff = Loff * yco + bright;
+
+ if ((uco[0] > 2.0) || (uco[2] > 2.0)) {
+ needux8 = TRUE;
+ ucscale = 0.125;
+ }
+ if ((vco[0] > 2.0) || (vco[2] > 2.0)) {
+ needvx8 = TRUE;
+ vcscale = 0.125;
+ }
txformat0 = (((((pPriv->w + 1 ) >> 1) - 1) & 0x7ff) |
(((((pPriv->h + 1 ) >> 1 ) - 1) & 0x7ff) << RADEON_TEX_VSIZE_SHIFT));
@@ -1824,7 +2040,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXC_ARG_A_TFACTOR_COLOR |
R200_TXC_ARG_B_R0_COLOR |
R200_TXC_ARG_C_TFACTOR_COLOR |
- R200_TXC_NEG_ARG_C |
+ (yoff < 0 ? R200_TXC_NEG_ARG_C : 0) |
R200_TXC_OP_DOT2_ADD);
OUT_ACCEL_REG(R200_PP_TXCBLEND2_0,
(0 << R200_TXC_TFACTOR_SEL_SHIFT) |
@@ -1846,7 +2062,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXC_SCALE_ARG_A |
R200_TXC_ARG_B_R0_COLOR |
R200_TXC_BIAS_ARG_B |
- R200_TXC_SCALE_ARG_B |
+ (needux8 ? R200_TXC_SCALE_ARG_B : 0) |
R200_TXC_ARG_C_R1_COLOR |
R200_TXC_OP_MADD);
OUT_ACCEL_REG(R200_PP_TXCBLEND2_1,
@@ -1868,6 +2084,7 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXC_SCALE_ARG_A |
R200_TXC_ARG_B_R0_COLOR |
R200_TXC_BIAS_ARG_B |
+ (needvx8 ? R200_TXC_SCALE_ARG_B : 0) |
R200_TXC_ARG_C_R1_COLOR |
R200_TXC_OP_MADD);
OUT_ACCEL_REG(R200_PP_TXCBLEND2_2,
@@ -1885,17 +2102,17 @@ FUNC_NAME(RADEONDisplayTexturedVideo)(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv
R200_TXA_CLAMP_0_1 | R200_TXA_OUTPUT_REG_R0);
/* shader constants */
- OUT_ACCEL_REG(R200_PP_TFACTOR_0, float4touint(1.0, /* src range [1, 2] */
- yco - 1.0,
- -yoff, /* range [-1, 0] */
+ OUT_ACCEL_REG(R200_PP_TFACTOR_0, float4touint(yco > 1.0 ? 1.0 : 0.0, /* range special [0, 2] */
+ yco > 1.0 ? yco - 1.0: yco,
+ yoff < 0 ? -yoff : yoff, /* range special [-1, 1] */
0.0));
- OUT_ACCEL_REG(R200_PP_TFACTOR_1, float4touint(uco[0] * 0.125 + 0.5, /* range [-4, 4] */
- uco[1] * 0.125 + 0.5,
- uco[2] * 0.125 + 0.5,
+ OUT_ACCEL_REG(R200_PP_TFACTOR_1, float4touint(uco[0] * ucscale + 0.5, /* range [-4, 4] */
+ uco[1] * ucscale + 0.5, /* or [-2, 2] */
+ uco[2] * ucscale + 0.5,
0.0));
- OUT_ACCEL_REG(R200_PP_TFACTOR_2, float4touint(vco[0] * 0.25 + 0.5, /* range [-2, 2] */
- vco[1] * 0.25 + 0.5,
- vco[2] * 0.25 + 0.5,
+ OUT_ACCEL_REG(R200_PP_TFACTOR_2, float4touint(vco[0] * vcscale + 0.5, /* range [-2, 2] */
+ vco[1] * vcscale + 0.5, /* or [-4, 4] */
+ vco[2] * vcscale + 0.5,
0.0));
FINISH_ACCEL();
diff --git a/src/radeon_video.c b/src/radeon_video.c
index 6314eb18..84791608 100644
--- a/src/radeon_video.c
+++ b/src/radeon_video.c
@@ -541,18 +541,6 @@ static XF86ImageRec Images[NUM_IMAGES] =
#endif
-/* Reference color space transform data */
-typedef struct tagREF_TRANSFORM
-{
- float RefLuma;
- float RefRCb;
- float RefRCr;
- float RefGCb;
- float RefGCr;
- float RefBCb;
- float RefBCr;
-} REF_TRANSFORM;
-
/* Parameters for ITU-R BT.601 and ITU-R BT.709 colour spaces */
static REF_TRANSFORM trans[2] =
{
@@ -560,7 +548,6 @@ static REF_TRANSFORM trans[2] =
{1.1678, 0.0, 1.7980, -0.2139, -0.5345, 2.1186, 0.0} /* BT.709 */
};
-
/* Gamma curve definition for preset gammas */
typedef struct tagGAMMA_CURVE_R100
{
diff --git a/src/radeon_video.h b/src/radeon_video.h
index 44980025..be338715 100644
--- a/src/radeon_video.h
+++ b/src/radeon_video.h
@@ -123,6 +123,18 @@ typedef struct {
int vsync;
} RADEONPortPrivRec, *RADEONPortPrivPtr;
+/* Reference color space transform data */
+typedef struct tagREF_TRANSFORM
+{
+ float RefLuma;
+ float RefRCb;
+ float RefRCr;
+ float RefGCb;
+ float RefGCr;
+ float RefBCb;
+ float RefBCr;
+} REF_TRANSFORM;
+
xf86CrtcPtr
radeon_xv_pick_best_crtc(ScrnInfoPtr pScrn,
int x1, int x2, int y1, int y2);
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-22 17:24:41 +0000