diff options
author | Eric Anholt <eric@anholt.net> | 2006-09-27 16:38:01 -0700 |
---|---|---|
committer | Eric Anholt <eric@anholt.net> | 2006-09-27 16:38:01 -0700 |
commit | 6ea16bf6b06c8b3aed4d2c98679ab28304d1b56c (patch) | |
tree | 1f711e8fb7338d6bee527d084a6ea12421087d79 /src/i830_video.c | |
parent | fdb6de663579d3b9f31bf9e8a93430b8505ca73f (diff) | |
parent | b970166eab95ac024ff481b0f5fd9aaf3644aabf (diff) |
Merge branch 'master' into modesetting
This reverts most of the mergedfb code. This will instead be done in device-
independent RandR code.
Conflicts:
src/Makefile.am
src/i810_driver.c
src/i810_reg.h
src/i830.h
src/i830_cursor.c
src/i830_driver.c
src/i830_modes.c
src/i830_video.c
Diffstat (limited to 'src/i830_video.c')
-rw-r--r-- | src/i830_video.c | 1417 |
1 files changed, 1241 insertions, 176 deletions
diff --git a/src/i830_video.c b/src/i830_video.c index 05f7f462..83c032bc 100644 --- a/src/i830_video.c +++ b/src/i830_video.c @@ -66,6 +66,7 @@ THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include <math.h> #include <string.h> +#include <assert.h> #include "xf86.h" #include "xf86_OSproc.h" @@ -77,12 +78,15 @@ THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "regionstr.h" #include "randrstr.h" #include "i830.h" +#include "i830_video.h" #include "xf86xv.h" #include <X11/extensions/Xv.h> #include "xaa.h" #include "xaalocal.h" #include "dixstruct.h" #include "fourcc.h" +#include "brw_defines.h" +#include "brw_structs.h" #ifndef USE_USLEEP_FOR_VIDEO #define USE_USLEEP_FOR_VIDEO 0 @@ -99,7 +103,8 @@ THE USE OR OTHER DEALINGS IN THE SOFTWARE. static void I830InitOffscreenImages(ScreenPtr); -static XF86VideoAdaptorPtr I830SetupImageVideo(ScreenPtr); +static XF86VideoAdaptorPtr I830SetupImageVideoOverlay(ScreenPtr); +static XF86VideoAdaptorPtr I830SetupImageVideoTextured(ScreenPtr); static void I830StopVideo(ScrnInfoPtr, pointer, Bool); static int I830SetPortAttribute(ScrnInfoPtr, Atom, INT32, pointer); static int I830GetPortAttribute(ScrnInfoPtr, Atom, INT32 *, pointer); @@ -109,20 +114,30 @@ static void I830QueryBestSize(ScrnInfoPtr, Bool, static int I830PutImage(ScrnInfoPtr, short, short, short, short, short, short, short, short, int, unsigned char *, short, short, Bool, RegionPtr, pointer, DrawablePtr); -static int I830QueryImageAttributes(ScrnInfoPtr, int, unsigned short *, - unsigned short *, int *, int *); +static int I830QueryImageAttributesOverlay(ScrnInfoPtr, int, unsigned short *, + unsigned short *, int *, int *); +static int I830QueryImageAttributesTextured(ScrnInfoPtr, int, unsigned short *, + unsigned short *, int *, int *); static void I830BlockHandler(int, pointer, pointer, pointer); +static FBLinearPtr +I830AllocateMemory(ScrnInfoPtr pScrn, FBLinearPtr linear, int size); + #define MAKE_ATOM(a) MakeAtom(a, sizeof(a) - 1, TRUE) static Atom xvBrightness, xvContrast, xvSaturation, xvColorKey, xvPipe, xvDoubleBuffer; static Atom xvGamma0, xvGamma1, xvGamma2, xvGamma3, xvGamma4, xvGamma5; #define IMAGE_MAX_WIDTH 1920 -#define IMAGE_MAX_HEIGHT 1088 +#define IMAGE_MAX_HEIGHT 1080 #define IMAGE_MAX_WIDTH_LEGACY 1024 -#define IMAGE_MAX_HEIGHT_LEGACY 1088 +#define IMAGE_MAX_HEIGHT_LEGACY 1080 + +/* + * Broadwater requires a bit of extra video memory for state information + */ +#define BRW_LINEAR_EXTRA (32*1024) #if !VIDEO_DEBUG #define ErrorF Edummy @@ -157,7 +172,10 @@ Edummy(const char *dummy, ...) OUT_RING(MI_NOOP); \ OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_FLIP_CONTINUE); \ } \ - OUT_RING(pI830->OverlayMem->Physical | OFC_UPDATE); \ + if (IS_I965G(pI830)) \ + OUT_RING(pI830->OverlayMem->Start | OFC_UPDATE); \ + else \ + OUT_RING(pI830->OverlayMem->Physical | OFC_UPDATE); \ ADVANCE_LP_RING(); \ ErrorF("OVERLAY_UPDATE\n"); \ } while(0) @@ -172,11 +190,17 @@ Edummy(const char *dummy, ...) OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP); \ OUT_RING(MI_NOOP); \ OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_FLIP_CONTINUE); \ - OUT_RING(pI830->OverlayMem->Physical | OFC_UPDATE); \ + if (IS_I965G(pI830)) \ + OUT_RING(pI830->OverlayMem->Start | OFC_UPDATE); \ + else \ + OUT_RING(pI830->OverlayMem->Physical | OFC_UPDATE); \ OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP); \ OUT_RING(MI_NOOP); \ OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_FLIP_OFF); \ - OUT_RING(pI830->OverlayMem->Physical | OFC_UPDATE); \ + if (IS_I965G(pI830)) \ + OUT_RING(pI830->OverlayMem->Start | OFC_UPDATE); \ + else \ + OUT_RING(pI830->OverlayMem->Physical | OFC_UPDATE); \ OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP); \ OUT_RING(MI_NOOP); \ ADVANCE_LP_RING(); \ @@ -287,6 +311,12 @@ static XF86AttributeRec Attributes[NUM_ATTRIBUTES] = { {XvSettable | XvGettable, 0, 1, "XV_DOUBLE_BUFFER"} }; +#define NUM_TEXTURED_ATTRIBUTES 2 +static XF86AttributeRec TexturedAttributes[NUM_ATTRIBUTES] = { + {XvSettable | XvGettable, -128, 127, "XV_BRIGHTNESS"}, + {XvSettable | XvGettable, 0, 255, "XV_CONTRAST"}, +}; + #define GAMMA_ATTRIBUTES 6 static XF86AttributeRec GammaAttributes[GAMMA_ATTRIBUTES] = { {XvSettable | XvGettable, 0, 0xffffff, "XV_GAMMA0"}, @@ -335,18 +365,18 @@ typedef struct { CARD32 OCONFIG; CARD32 OCMD; CARD32 RESERVED1; /* 0x6C */ - CARD32 AWINPOS; - CARD32 AWINSZ; - CARD32 RESERVED2; /* 0x78 */ - CARD32 RESERVED3; /* 0x7C */ - CARD32 RESERVED4; /* 0x80 */ - CARD32 RESERVED5; /* 0x84 */ - CARD32 RESERVED6; /* 0x88 */ - CARD32 RESERVED7; /* 0x8C */ - CARD32 RESERVED8; /* 0x90 */ - CARD32 RESERVED9; /* 0x94 */ - CARD32 RESERVEDA; /* 0x98 */ - CARD32 RESERVEDB; /* 0x9C */ + CARD32 OSTART_0Y; /* for i965 */ + CARD32 OSTART_1Y; /* for i965 */ + CARD32 OSTART_0U; + CARD32 OSTART_0V; + CARD32 OSTART_1U; + CARD32 OSTART_1V; + CARD32 OTILEOFF_0Y; + CARD32 OTILEOFF_1Y; + CARD32 OTILEOFF_0U; + CARD32 OTILEOFF_0V; + CARD32 OTILEOFF_1U; + CARD32 OTILEOFF_1V; CARD32 FASTHSCALE; /* 0xA0 */ CARD32 UVSCALEV; /* 0xA4 */ @@ -361,46 +391,6 @@ typedef struct { CARD16 RESERVEDG[0x100 / 2 - N_HORIZ_UV_TAPS * N_PHASES]; } I830OverlayRegRec, *I830OverlayRegPtr; -typedef struct { - CARD32 YBuf0offset; - CARD32 UBuf0offset; - CARD32 VBuf0offset; - - CARD32 YBuf1offset; - CARD32 UBuf1offset; - CARD32 VBuf1offset; - - unsigned char currentBuf; - - int brightness; - int contrast; - int saturation; - int pipe; - int doubleBuffer; - - RegionRec clip; - CARD32 colorKey; - - CARD32 gamma0; - CARD32 gamma1; - CARD32 gamma2; - CARD32 gamma3; - CARD32 gamma4; - CARD32 gamma5; - - CARD32 videoStatus; - Time offTime; - Time freeTime; - FBLinearPtr linear; - - Bool overlayOK; - int oneLineMode; - int scaleRatio; -} I830PortPrivRec, *I830PortPrivPtr; - -#define GET_PORT_PRIVATE(pScrn) \ - (I830PortPrivPtr)((I830PTR(pScrn))->adaptor->pPortPrivates[0].ptr) - #if VIDEO_DEBUG static void CompareOverlay(I830Ptr pI830, CARD32 * overlay, int size) @@ -426,8 +416,9 @@ void I830InitVideo(ScreenPtr pScreen) { ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + I830Ptr pI830 = I830PTR(pScrn); XF86VideoAdaptorPtr *adaptors, *newAdaptors = NULL; - XF86VideoAdaptorPtr newAdaptor = NULL; + XF86VideoAdaptorPtr overlayAdaptor = NULL, texturedAdaptor = NULL; int num_adaptors; DPRINTF(PFX, "I830InitVideo\n"); @@ -446,35 +437,54 @@ I830InitVideo(ScreenPtr pScreen) } #endif - if (pScrn->bitsPerPixel != 8) { - newAdaptor = I830SetupImageVideo(pScreen); + num_adaptors = xf86XVListGenericAdaptors(pScrn, &adaptors); + /* Give our adaptor list enough space for the overlay and/or texture video + * adaptors. + */ + newAdaptors = xalloc((num_adaptors + 2) * sizeof(XF86VideoAdaptorPtr *)); + if (newAdaptors == NULL) + return; + + memcpy(newAdaptors, adaptors, num_adaptors * sizeof(XF86VideoAdaptorPtr)); + adaptors = newAdaptors; + + /* Add the adaptors supported by our hardware. First, set up the atoms + * that will be used by both output adaptors. + */ + xvBrightness = MAKE_ATOM("XV_BRIGHTNESS"); + xvContrast = MAKE_ATOM("XV_CONTRAST"); + + /* Set up overlay video if we can do it at this depth. */ + if (!IS_I965G(pI830) && pScrn->bitsPerPixel != 8) { + overlayAdaptor = I830SetupImageVideoOverlay(pScreen); + if (overlayAdaptor != NULL) { + adaptors[num_adaptors++] = overlayAdaptor; + xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Set up overlay video\n"); + } else { + xf86DrvMsg(pScrn->scrnIndex, X_ERROR, + "Failed to set up overlay video\n"); + } I830InitOffscreenImages(pScreen); } - num_adaptors = xf86XVListGenericAdaptors(pScrn, &adaptors); - - if (newAdaptor) { - if (!num_adaptors) { - num_adaptors = 1; - adaptors = &newAdaptor; + /* Set up textured video if we can do it at this depth and we are on + * supported hardware. + */ + if (pScrn->bitsPerPixel >= 16 && (IS_I9XX(pI830) || IS_I965G(pI830))) { + texturedAdaptor = I830SetupImageVideoTextured(pScreen); + if (texturedAdaptor != NULL) { + adaptors[num_adaptors++] = texturedAdaptor; + xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Set up textured video\n"); } else { - newAdaptors = /* need to free this someplace */ - xalloc((num_adaptors + 1) * sizeof(XF86VideoAdaptorPtr *)); - if (newAdaptors) { - memcpy(newAdaptors, adaptors, num_adaptors * - sizeof(XF86VideoAdaptorPtr)); - newAdaptors[num_adaptors] = newAdaptor; - adaptors = newAdaptors; - num_adaptors++; - } + xf86DrvMsg(pScrn->scrnIndex, X_ERROR, + "Failed to set up textured video\n"); } } if (num_adaptors) xf86XVScreenInit(pScreen, adaptors, num_adaptors); - if (newAdaptors) - xfree(newAdaptors); + xfree(adaptors); } static void @@ -503,8 +513,10 @@ I830ResetVideo(ScrnInfoPtr pScrn) overlay->SHEIGHT = 0; overlay->OCLRC0 = (pPriv->contrast << 18) | (pPriv->brightness & 0xff); overlay->OCLRC1 = pPriv->saturation; +#if 0 overlay->AWINPOS = 0; overlay->AWINSZ = 0; +#endif overlay->FASTHSCALE = 0; /* @@ -644,7 +656,7 @@ I830UpdateGamma(ScrnInfoPtr pScrn) } static XF86VideoAdaptorPtr -I830SetupImageVideo(ScreenPtr pScreen) +I830SetupImageVideoOverlay(ScreenPtr pScreen) { ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; I830Ptr pI830 = I830PTR(pScrn); @@ -652,7 +664,7 @@ I830SetupImageVideo(ScreenPtr pScreen) I830PortPrivPtr pPriv; XF86AttributePtr att; - DPRINTF(PFX, "I830SetupImageVideo\n"); + DPRINTF(PFX, "I830SetupImageVideoOverlay\n"); if (!(adapt = xcalloc(1, sizeof(XF86VideoAdaptorRec) + sizeof(I830PortPrivRec) + sizeof(DevUnion)))) @@ -705,8 +717,9 @@ I830SetupImageVideo(ScreenPtr pScreen) adapt->GetPortAttribute = I830GetPortAttribute; adapt->QueryBestSize = I830QueryBestSize; adapt->PutImage = I830PutImage; - adapt->QueryImageAttributes = I830QueryImageAttributes; + adapt->QueryImageAttributes = I830QueryImageAttributesOverlay; + pPriv->textured = FALSE; pPriv->colorKey = pI830->colorKey & ((1 << pScrn->depth) - 1); pPriv->videoStatus = 0; pPriv->brightness = 0; @@ -770,6 +783,86 @@ I830SetupImageVideo(ScreenPtr pScreen) return adapt; } +static XF86VideoAdaptorPtr +I830SetupImageVideoTextured(ScreenPtr pScreen) +{ + XF86VideoAdaptorPtr adapt; + XF86VideoEncodingPtr encoding; + XF86AttributePtr attrs; + I830PortPrivPtr portPrivs; + DevUnion *devUnions; + int nports = 16, i; + int nAttributes; + + DPRINTF(PFX, "I830SetupImageVideoOverlay\n"); + + nAttributes = NUM_TEXTURED_ATTRIBUTES; + + adapt = xcalloc(1, sizeof(XF86VideoAdaptorRec)); + portPrivs = xcalloc(nports, sizeof(I830PortPrivRec)); + devUnions = xcalloc(nports, sizeof(DevUnion)); + encoding = xcalloc(1, sizeof(XF86VideoEncodingRec)); + attrs = xcalloc(nAttributes, sizeof(XF86AttributeRec)); + if (adapt == NULL || portPrivs == NULL || devUnions == NULL || + encoding == NULL || attrs == NULL) + { + xfree(adapt); + xfree(portPrivs); + xfree(devUnions); + xfree(encoding); + xfree(attrs); + return NULL; + } + + adapt->type = XvWindowMask | XvInputMask | XvImageMask; + adapt->flags = 0; + adapt->name = "Intel(R) Textured Video"; + adapt->nEncodings = 1; + adapt->pEncodings = encoding; + adapt->pEncodings[0].id = 0; + adapt->pEncodings[0].name = "XV_IMAGE"; + adapt->pEncodings[0].width = 2048; + adapt->pEncodings[0].height = 2048; + adapt->pEncodings[0].rate.numerator = 1; + adapt->pEncodings[0].rate.denominator = 1; + adapt->nFormats = NUM_FORMATS; + adapt->pFormats = Formats; + adapt->nPorts = nports; + adapt->pPortPrivates = devUnions; + adapt->nAttributes = nAttributes; + adapt->pAttributes = attrs; + memcpy(attrs, TexturedAttributes, nAttributes * sizeof(XF86AttributeRec)); + adapt->nImages = NUM_IMAGES; + adapt->pImages = Images; + adapt->PutVideo = NULL; + adapt->PutStill = NULL; + adapt->GetVideo = NULL; + adapt->GetStill = NULL; + adapt->StopVideo = I830StopVideo; + adapt->SetPortAttribute = I830SetPortAttribute; + adapt->GetPortAttribute = I830GetPortAttribute; + adapt->QueryBestSize = I830QueryBestSize; + adapt->PutImage = I830PutImage; + adapt->QueryImageAttributes = I830QueryImageAttributesTextured; + + for (i = 0; i < nports; i++) { + I830PortPrivPtr pPriv = &portPrivs[i]; + + pPriv->textured = TRUE; + pPriv->videoStatus = 0; + pPriv->linear = NULL; + pPriv->currentBuf = 0; + pPriv->doubleBuffer = 0; + + /* gotta uninit this someplace, XXX: shouldn't be necessary for textured */ + REGION_NULL(pScreen, &pPriv->clip); + + adapt->pPortPrivates[i].ptr = (pointer) (pPriv); + } + + return adapt; +} + static Bool RegionsEqual(RegionPtr A, RegionPtr B) { @@ -807,6 +900,9 @@ I830StopVideo(ScrnInfoPtr pScrn, pointer data, Bool shutdown) I830OverlayRegPtr overlay = (I830OverlayRegPtr) (pI830->FbBase + pI830->OverlayMem->Start); + if (pPriv->textured) + return; + DPRINTF(PFX, "I830StopVideo\n"); REGION_EMPTY(pScrn->pScreen, &pPriv->clip); @@ -846,6 +942,14 @@ I830SetPortAttribute(ScrnInfoPtr pScrn, I830OverlayRegPtr overlay = (I830OverlayRegPtr) (pI830->FbBase + pI830->OverlayMem->Start); + if (pPriv->textured) { + /* XXX: Currently the brightness/saturation attributes aren't hooked up. + * However, apps expect them to be there, and the spec seems to let us + * sneak out of actually implementing them for now. + */ + return Success; + } + if (attribute == xvBrightness) { if ((value < -128) || (value > 127)) return BadValue; @@ -1010,13 +1114,12 @@ I830QueryBestSize(ScrnInfoPtr pScrn, } static void -I830CopyPackedData(ScrnInfoPtr pScrn, +I830CopyPackedData(ScrnInfoPtr pScrn, I830PortPrivPtr pPriv, unsigned char *buf, int srcPitch, int dstPitch, int top, int left, int h, int w) { I830Ptr pI830 = I830PTR(pScrn); - I830PortPrivPtr pPriv = pI830->adaptor->pPortPrivates[0].ptr; unsigned char *src, *dst; int i,j; unsigned char *s; @@ -1104,13 +1207,72 @@ I830CopyPackedData(ScrnInfoPtr pScrn, } } +/* Copies planar data in *buf to UYVY-packed data in the screen atYBufXOffset. + */ +static void +I830CopyPlanarToPackedData(ScrnInfoPtr pScrn, I830PortPrivPtr pPriv, + unsigned char *buf, int srcPitch, + int srcPitch2, int dstPitch, int srcH, + int top, int left, int h, int w, int id) +{ + I830Ptr pI830 = I830PTR(pScrn); + CARD8 *dst1, *srcy, *srcu, *srcv; + int y; + + if (pPriv->currentBuf == 0) + dst1 = pI830->FbBase + pPriv->YBuf0offset; + else + dst1 = pI830->FbBase + pPriv->YBuf1offset; + + srcy = buf + (top * srcPitch) + left; + if (id == FOURCC_YV12) { + srcu = buf + (srcH * srcPitch) + ((top / 2) * srcPitch2) + (left / 2); + srcv = buf + (srcH * srcPitch) + ((srcH / 2) * srcPitch2) + + ((top / 2) * srcPitch2) + (left / 2); + } else { + srcv = buf + (srcH * srcPitch) + ((top / 2) * srcPitch2) + (left / 2); + srcu = buf + (srcH * srcPitch) + ((srcH / 2) * srcPitch2) + + ((top / 2) * srcPitch2) + (left / 2); + } + + for (y = 0; y < h; y++) { + CARD32 *dst = (CARD32 *)dst1; + CARD8 *sy = srcy; + CARD8 *su = srcu; + CARD8 *sv = srcv; + int i; + + i = w / 2; + while(i > 4) { + dst[0] = sy[0] | (sy[1] << 16) | (sv[0] << 8) | (su[0] << 24); + dst[1] = sy[2] | (sy[3] << 16) | (sv[1] << 8) | (su[1] << 24); + dst[2] = sy[4] | (sy[5] << 16) | (sv[2] << 8) | (su[2] << 24); + dst[3] = sy[6] | (sy[7] << 16) | (sv[3] << 8) | (su[3] << 24); + dst += 4; su += 4; sv += 4; sy += 8; + i -= 4; + } + while(i--) { + dst[0] = sy[0] | (sy[1] << 16) | (sv[0] << 8) | (su[0] << 24); + dst++; su++; sv++; + sy += 2; + } + + dst1 += dstPitch; + srcy += srcPitch; + if (y & 1) { + srcu += srcPitch2; + srcv += srcPitch2; + } + } +} + static void -I830CopyPlanarData(ScrnInfoPtr pScrn, unsigned char *buf, int srcPitch, +I830CopyPlanarData(ScrnInfoPtr pScrn, I830PortPrivPtr pPriv, + unsigned char *buf, int srcPitch, int srcPitch2, int dstPitch, int srcH, int top, int left, int h, int w, int id) { I830Ptr pI830 = I830PTR(pScrn); - I830PortPrivPtr pPriv = pI830->adaptor->pPortPrivates[0].ptr; int i, j = 0; unsigned char *src1, *src2, *src3, *dst1, *dst2, *dst3; unsigned char *s; @@ -1427,15 +1589,6 @@ I830DisplayVideo(ScrnInfoPtr pScrn, int id, short width, short height, CompareOverlay(pI830, (CARD32 *) overlay, 0x100); #endif - /* When in dual head with different bpp setups we need to refresh the - * color key, so let's reset the video parameters and refresh here */ - if (pI830->entityPrivate) - I830ResetVideo(pScrn); - - /* Ensure overlay is turned on with OVERLAY_ENABLE at 0 */ - if (!*pI830->overlayOn) - OVERLAY_UPDATE; - switch (pI830->rotation) { case RR_Rotate_0: dstBox->x1 -= pScrn->frameX0; @@ -1475,6 +1628,16 @@ I830DisplayVideo(ScrnInfoPtr pScrn, int id, short width, short height, break; } + /* When in dual head with different bpp setups we need to refresh the + * color key, so let's reset the video parameters and refresh here. + * In MergedFB mode, we may need to flip pipes too. */ + if (pI830->entityPrivate) + I830ResetVideo(pScrn); + + /* Ensure overlay is turned on with OVERLAY_ENABLE at 0 */ + if (!*pI830->overlayOn) + OVERLAY_UPDATE; + /* Fix up the dstBox if outside the visible screen */ { int offset_x = (dstBox->x1 < 0) ? -dstBox->x1 : 0; @@ -1538,7 +1701,7 @@ I830DisplayVideo(ScrnInfoPtr pScrn, int id, short width, short height, /* Keep the engine happy and clip to the real vertical size just * in case an LFP is in use and it's not at it's native resolution. */ - int vactive = pI830->pipe ? (INREG(VTOTAL_B) & 0x7FF) : (INREG(VTOTAL_A) & 0x7FF); + int vactive = pPriv->pipe ? (INREG(VTOTAL_B) & 0x7FF) : (INREG(VTOTAL_A) & 0x7FF); vactive += 1; @@ -1663,14 +1826,31 @@ I830DisplayVideo(ScrnInfoPtr pScrn, int id, short width, short height, dstBox->x2, dstBox->y2); /* buffer locations */ - overlay->OBUF_0Y = pPriv->YBuf0offset; - overlay->OBUF_0U = pPriv->UBuf0offset; - overlay->OBUF_0V = pPriv->VBuf0offset; - - if(pPriv->doubleBuffer) { - overlay->OBUF_1Y = pPriv->YBuf1offset; - overlay->OBUF_1U = pPriv->UBuf1offset; - overlay->OBUF_1V = pPriv->VBuf1offset; + if (IS_I965G(pI830)) + { + overlay->OBUF_0Y = 0; + overlay->OBUF_0U = 0; + overlay->OBUF_0V = 0; + overlay->OSTART_0Y = pPriv->YBuf0offset; + overlay->OSTART_0U = pPriv->UBuf0offset; + overlay->OSTART_0V = pPriv->VBuf0offset; + if(pPriv->doubleBuffer) { + overlay->OBUF_1Y = 0; + overlay->OBUF_1U = 0; + overlay->OBUF_1V = 0; + overlay->OSTART_1Y = pPriv->YBuf1offset; + overlay->OSTART_1U = pPriv->UBuf1offset; + overlay->OSTART_1V = pPriv->VBuf1offset; + } + } else { + overlay->OBUF_0Y = pPriv->YBuf0offset; + overlay->OBUF_0U = pPriv->UBuf0offset; + overlay->OBUF_0V = pPriv->VBuf0offset; + if(pPriv->doubleBuffer) { + overlay->OBUF_1Y = pPriv->YBuf1offset; + overlay->OBUF_1U = pPriv->UBuf1offset; + overlay->OBUF_1V = pPriv->VBuf1offset; + } } ErrorF("Buffers: Y0: 0x%lx, U0: 0x%lx, V0: 0x%lx\n", overlay->OBUF_0Y, @@ -1863,6 +2043,794 @@ I830DisplayVideo(ScrnInfoPtr pScrn, int id, short width, short height, OVERLAY_UPDATE; } +static const CARD32 sip_kernel_static[][4] = { +/* wait (1) a0<1>UW a145<0,1,0>UW { align1 + } */ + { 0x00000030, 0x20000108, 0x00001220, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +/* nop (4) g0<1>UD { align1 + } */ + { 0x0040007e, 0x20000c21, 0x00690000, 0x00000000 }, +}; + +/* + * this program computes dA/dx and dA/dy for the texture coordinates along + * with the base texture coordinate. It was extracted from the Mesa driver. + * It uses about 10 GRF registers. + */ + +#define SF_KERNEL_NUM_GRF 16 +#define SF_MAX_THREADS 1 + +static const CARD32 sf_kernel_static[][4] = { +#include "sf_prog.h" +}; + +/* + * Ok, this kernel picks up the required data flow values in g0 and g1 + * and passes those along in m0 and m1. In m2-m9, it sticks constant + * values (bright pink). + */ + +/* Our PS kernel uses less than 32 GRF registers (about 20) */ +#define PS_KERNEL_NUM_GRF 32 +#define PS_MAX_THREADS 32 + +#define BRW_GRF_BLOCKS(nreg) ((nreg + 15) / 16 - 1) + +static const CARD32 ps_kernel_static[][4] = { +#include "wm_prog.h" +}; + +#define ALIGN(i,m) (((i) + (m) - 1) & ~((m) - 1)) +#define MIN(a,b) ((a) < (b) ? (a) : (b)) + +#define WM_BINDING_TABLE_ENTRIES 2 + +static CARD32 float_to_uint (float f) { + union {CARD32 i; float f;} x; + x.f = f; + return x.i; +} + +#if 0 +static struct { + CARD32 svg_ctl; + char *name; +} svg_ctl_bits[] = { + { BRW_SVG_CTL_GS_BA, "General State Base Address" }, + { BRW_SVG_CTL_SS_BA, "Surface State Base Address" }, + { BRW_SVG_CTL_IO_BA, "Indirect Object Base Address" }, + { BRW_SVG_CTL_GS_AUB, "Generate State Access Upper Bound" }, + { BRW_SVG_CTL_IO_AUB, "Indirect Object Access Upper Bound" }, + { BRW_SVG_CTL_SIP, "System Instruction Pointer" }, + { 0, 0 }, +}; + +static void +brw_debug (ScrnInfoPtr pScrn, char *when) +{ + I830Ptr pI830 = I830PTR(pScrn); + int i; + CARD32 v; + + I830Sync (pScrn); + ErrorF("brw_debug: %s\n", when); + for (i = 0; svg_ctl_bits[i].name; i++) { + OUTREG(BRW_SVG_CTL, svg_ctl_bits[i].svg_ctl); + v = INREG(BRW_SVG_RDATA); + ErrorF("\t%34.34s: 0x%08x\n", svg_ctl_bits[i].name, v); + } +} +#endif + +#define WATCH_SF 0 +#define WATCH_WIZ 0 +#define WATCH_STATS 0 + +static void +BroadwaterDisplayVideoTextured(ScrnInfoPtr pScrn, I830PortPrivPtr pPriv, int id, + RegionPtr dstRegion, + short width, short height, int video_pitch, + int x1, int y1, int x2, int y2, + short src_w, short src_h, + short drw_w, short drw_h, + DrawablePtr pDraw) +{ + I830Ptr pI830 = I830PTR(pScrn); + BoxPtr pbox; + int nbox, dxo, dyo; + int urb_vs_start, urb_vs_size; + int urb_gs_start, urb_gs_size; + int urb_clip_start, urb_clip_size; + int urb_sf_start, urb_sf_size; + int urb_cs_start, urb_cs_size; + struct brw_surface_state *dest_surf_state; + struct brw_surface_state *src_surf_state; + struct brw_sampler_state *src_sampler_state; + struct brw_vs_unit_state *vs_state; + struct brw_sf_unit_state *sf_state; + struct brw_wm_unit_state *wm_state; + struct brw_cc_unit_state *cc_state; + struct brw_cc_viewport *cc_viewport; + struct brw_instruction *sf_kernel; + struct brw_instruction *ps_kernel; + struct brw_instruction *sip_kernel; + float *vb; + CARD32 *binding_table; + Bool first_output = TRUE; + int dest_surf_offset, src_surf_offset, src_sampler_offset, vs_offset; + int sf_offset, wm_offset, cc_offset, vb_offset, cc_viewport_offset; + int wm_scratch_offset; + int sf_kernel_offset, ps_kernel_offset, sip_kernel_offset; + int binding_table_offset; + int next_offset, total_state_size; + int vb_size = (4 * 4) * 4; /* 4 DWORDS per vertex */ + char *state_base; + int state_base_offset; + +#if 0 + ErrorF("BroadwaterDisplayVideoTextured: %dx%d (pitch %d)\n", width, height, + video_pitch); +#endif + + /* enable debug */ + OUTREG (INST_PM, + (1 << (16 + 4)) | + (1 << 4)); +#if 0 + ErrorF ("INST_PM 0x%08x\n", INREG(INST_PM)); +#endif + + assert((id == FOURCC_UYVY) || (id == FOURCC_YUY2)); + + /* Tell the rotation code that we have stomped its invariant state by + * setting a high bit. We don't use any invariant 3D state for video, so we + * don't have to worry about it ourselves. + */ + *pI830->used3D |= 1 << 30; + +#ifdef XF86DRI + /* Tell the DRI that we're smashing its state. */ + if (pI830->directRenderingEnabled) { + drmI830Sarea *pSAREAPriv = DRIGetSAREAPrivate(pScrn->pScreen); + + pSAREAPriv->ctxOwner = DRIGetContext(pScrn->pScreen); + } +#endif /* XF86DRI */ + + next_offset = 0; + + /* Set up our layout of state in framebuffer. First the general state: */ + vs_offset = ALIGN(next_offset, 64); + next_offset = vs_offset + sizeof(*vs_state); + sf_offset = ALIGN(next_offset, 32); + next_offset = sf_offset + sizeof(*sf_state); + wm_offset = ALIGN(next_offset, 32); + next_offset = wm_offset + sizeof(*wm_state); + wm_scratch_offset = ALIGN(next_offset, 1024); + next_offset = wm_scratch_offset + 1024 * PS_MAX_THREADS; + cc_offset = ALIGN(next_offset, 32); + next_offset = cc_offset + sizeof(*cc_state); + + sf_kernel_offset = ALIGN(next_offset, 64); + next_offset = sf_kernel_offset + sizeof (sf_kernel_static); + ps_kernel_offset = ALIGN(next_offset, 64); + next_offset = ps_kernel_offset + sizeof (ps_kernel_static); + sip_kernel_offset = ALIGN(next_offset, 64); + next_offset = sip_kernel_offset + sizeof (sip_kernel_static); + cc_viewport_offset = ALIGN(next_offset, 32); + next_offset = cc_viewport_offset + sizeof(*cc_viewport); + + src_sampler_offset = ALIGN(next_offset, 32); + next_offset = src_sampler_offset + sizeof(*src_sampler_state); + + /* Align VB to native size of elements, for safety */ + vb_offset = ALIGN(next_offset, 8); + next_offset = vb_offset + vb_size; + + /* And then the general state: */ + dest_surf_offset = ALIGN(next_offset, 32); + next_offset = dest_surf_offset + sizeof(*dest_surf_state); + src_surf_offset = ALIGN(next_offset, 32); + next_offset = src_surf_offset + sizeof(*src_surf_state); + binding_table_offset = ALIGN(next_offset, 32); + next_offset = binding_table_offset + (WM_BINDING_TABLE_ENTRIES * 4); + + /* Allocate an area in framebuffer for our state layout we just set up */ + total_state_size = next_offset; + assert (total_state_size < BRW_LINEAR_EXTRA); + + /* + * Use the extra space allocated at the end of the Xv buffer + */ + state_base_offset = (pPriv->YBuf0offset + + pPriv->linear->size * pI830->cpp - + BRW_LINEAR_EXTRA); + state_base_offset = ALIGN(state_base_offset, 64); + + state_base = (char *)(pI830->FbBase + state_base_offset); + /* Set up our pointers to state structures in framebuffer. It would probably + * be a good idea to fill these structures out in system memory and then dump + * them there, instead. + */ + vs_state = (void *)(state_base + vs_offset); + sf_state = (void *)(state_base + sf_offset); + wm_state = (void *)(state_base + wm_offset); + cc_state = (void *)(state_base + cc_offset); + sf_kernel = (void *)(state_base + sf_kernel_offset); + ps_kernel = (void *)(state_base + ps_kernel_offset); + sip_kernel = (void *)(state_base + sip_kernel_offset); + + cc_viewport = (void *)(state_base + cc_viewport_offset); + dest_surf_state = (void *)(state_base + dest_surf_offset); + src_surf_state = (void *)(state_base + src_surf_offset); + src_sampler_state = (void *)(state_base + src_sampler_offset); + binding_table = (void *)(state_base + binding_table_offset); + vb = (void *)(state_base + vb_offset); + + /* For 3D, the VS must have 8, 12, 16, 24, or 32 VUEs allocated to it. + * A VUE consists of a 256-bit vertex header followed by the vertex data, + * which in our case is 4 floats (128 bits), thus a single 512-bit URB + * entry. + */ +#define URB_VS_ENTRIES 8 +#define URB_VS_ENTRY_SIZE 1 + +#define URB_GS_ENTRIES 0 +#define URB_GS_ENTRY_SIZE 0 + +#define URB_CLIP_ENTRIES 0 +#define URB_CLIP_ENTRY_SIZE 0 + + /* The SF kernel we use outputs only 4 256-bit registers, leading to an + * entry size of 2 512-bit URBs. We don't need to have many entries to + * output as we're generally working on large rectangles and don't care + * about having WM threads running on different rectangles simultaneously. + */ +#define URB_SF_ENTRIES 1 +#define URB_SF_ENTRY_SIZE 2 + +#define URB_CS_ENTRIES 0 +#define URB_CS_ENTRY_SIZE 0 + + urb_vs_start = 0; + urb_vs_size = URB_VS_ENTRIES * URB_VS_ENTRY_SIZE; + urb_gs_start = urb_vs_start + urb_vs_size; + urb_gs_size = URB_GS_ENTRIES * URB_GS_ENTRY_SIZE; + urb_clip_start = urb_gs_start + urb_gs_size; + urb_clip_size = URB_CLIP_ENTRIES * URB_CLIP_ENTRY_SIZE; + urb_sf_start = urb_clip_start + urb_clip_size; + urb_sf_size = URB_SF_ENTRIES * URB_SF_ENTRY_SIZE; + urb_cs_start = urb_sf_start + urb_sf_size; + urb_cs_size = URB_CS_ENTRIES * URB_CS_ENTRY_SIZE; + + /* We'll be poking the state buffers that could be in use by the 3d hardware + * here, but we should have synced the 3D engine already in I830PutImage. + */ + + memset (cc_viewport, 0, sizeof (*cc_viewport)); + cc_viewport->min_depth = -1.e35; + cc_viewport->max_depth = 1.e35; + + /* Color calculator state */ + memset(cc_state, 0, sizeof(*cc_state)); + cc_state->cc0.stencil_enable = 0; /* disable stencil */ + cc_state->cc2.depth_test = 0; /* disable depth test */ + cc_state->cc2.logicop_enable = 1; /* enable logic op */ + cc_state->cc3.ia_blend_enable = 1; /* blend alpha just like colors */ + cc_state->cc3.blend_enable = 0; /* disable color blend */ + cc_state->cc3.alpha_test = 0; /* disable alpha test */ + cc_state->cc4.cc_viewport_state_offset = (state_base_offset + cc_viewport_offset) >> 5; + cc_state->cc5.dither_enable = 0; /* disable dither */ + cc_state->cc5.logicop_func = 0xc; /* WHITE */ + cc_state->cc5.statistics_enable = 1; + cc_state->cc5.ia_blend_function = BRW_BLENDFUNCTION_ADD; + cc_state->cc5.ia_src_blend_factor = BRW_BLENDFACTOR_ONE; + cc_state->cc5.ia_dest_blend_factor = BRW_BLENDFACTOR_ONE; + + /* Upload system kernel */ + memcpy (sip_kernel, sip_kernel_static, sizeof (sip_kernel_static)); + + /* Set up the state buffer for the destination surface */ + memset(dest_surf_state, 0, sizeof(*dest_surf_state)); + dest_surf_state->ss0.surface_type = BRW_SURFACE_2D; + dest_surf_state->ss0.data_return_format = BRW_SURFACERETURNFORMAT_FLOAT32; + if (pI830->cpp == 2) { + dest_surf_state->ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; + } else { + dest_surf_state->ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + } + dest_surf_state->ss0.writedisable_alpha = 0; + dest_surf_state->ss0.writedisable_red = 0; + dest_surf_state->ss0.writedisable_green = 0; + dest_surf_state->ss0.writedisable_blue = 0; + dest_surf_state->ss0.color_blend = 1; + dest_surf_state->ss0.vert_line_stride = 0; + dest_surf_state->ss0.vert_line_stride_ofs = 0; + dest_surf_state->ss0.mipmap_layout_mode = 0; + dest_surf_state->ss0.render_cache_read_mode = 0; + + dest_surf_state->ss1.base_addr = pI830->FrontBuffer.Start; + dest_surf_state->ss2.height = pScrn->virtualY - 1; + dest_surf_state->ss2.width = pScrn->virtualX - 1; + dest_surf_state->ss2.mip_count = 0; + dest_surf_state->ss2.render_target_rotation = 0; + dest_surf_state->ss3.pitch = (pI830->displayWidth * pI830->cpp) - 1; + + /* Set up the source surface state buffer */ + memset(src_surf_state, 0, sizeof(*src_surf_state)); + src_surf_state->ss0.surface_type = BRW_SURFACE_2D; +/* src_surf_state->ss0.data_return_format = BRW_SURFACERETURNFORMAT_FLOAT32; */ + switch (id) { + case FOURCC_YUY2: + src_surf_state->ss0.surface_format = BRW_SURFACEFORMAT_YCRCB_NORMAL; + break; + case FOURCC_UYVY: + src_surf_state->ss0.surface_format = BRW_SURFACEFORMAT_YCRCB_SWAPY; + break; + } + src_surf_state->ss0.writedisable_alpha = 0; + src_surf_state->ss0.writedisable_red = 0; + src_surf_state->ss0.writedisable_green = 0; + src_surf_state->ss0.writedisable_blue = 0; + src_surf_state->ss0.color_blend = 1; + src_surf_state->ss0.vert_line_stride = 0; + src_surf_state->ss0.vert_line_stride_ofs = 0; + src_surf_state->ss0.mipmap_layout_mode = 0; + src_surf_state->ss0.render_cache_read_mode = 0; + + src_surf_state->ss1.base_addr = pPriv->YBuf0offset; + src_surf_state->ss2.width = width - 1; + src_surf_state->ss2.height = height - 1; + src_surf_state->ss2.mip_count = 0; + src_surf_state->ss2.render_target_rotation = 0; + src_surf_state->ss3.pitch = video_pitch - 1; + + /* Set up a binding table for our two surfaces. Only the PS will use it */ + /* XXX: are these offset from the right place? */ + binding_table[0] = state_base_offset + dest_surf_offset; + binding_table[1] = state_base_offset + src_surf_offset; + + /* Set up the packed YUV source sampler. Doesn't do colorspace conversion. + */ + memset(src_sampler_state, 0, sizeof(*src_sampler_state)); + src_sampler_state->ss0.min_filter = BRW_MAPFILTER_LINEAR; + src_sampler_state->ss0.mag_filter = BRW_MAPFILTER_LINEAR; + src_sampler_state->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CLAMP; + src_sampler_state->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CLAMP; + src_sampler_state->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CLAMP; + + /* Set up the vertex shader to be disabled (passthrough) */ + memset(vs_state, 0, sizeof(*vs_state)); + vs_state->thread4.nr_urb_entries = URB_VS_ENTRIES; + vs_state->thread4.urb_entry_allocation_size = URB_VS_ENTRY_SIZE - 1; + vs_state->vs6.vs_enable = 0; + vs_state->vs6.vert_cache_disable = 1; + + /* Set up the SF kernel to do coord interp: for each attribute, + * calculate dA/dx and dA/dy. Hand these interpolation coefficients + * back to SF which then hands pixels off to WM. + */ + + memcpy (sf_kernel, sf_kernel_static, sizeof (sf_kernel_static)); + memset(sf_state, 0, sizeof(*sf_state)); +#if 0 + ErrorF ("sf kernel: 0x%08x\n", state_base_offset + sf_kernel_offset); +#endif + sf_state->thread0.kernel_start_pointer = + (state_base_offset + sf_kernel_offset) >> 6; + sf_state->thread0.grf_reg_count = BRW_GRF_BLOCKS(SF_KERNEL_NUM_GRF); + sf_state->sf1.single_program_flow = 1; /* XXX */ + sf_state->sf1.binding_table_entry_count = 0; + sf_state->sf1.thread_priority = 0; + sf_state->sf1.floating_point_mode = 0; /* Mesa does this */ + sf_state->sf1.illegal_op_exception_enable = 1; + sf_state->sf1.mask_stack_exception_enable = 1; + sf_state->sf1.sw_exception_enable = 1; + sf_state->thread2.per_thread_scratch_space = 0; + sf_state->thread2.scratch_space_base_pointer = 0; /* not used in our kernel */ + sf_state->thread3.const_urb_entry_read_length = 0; /* no const URBs */ + sf_state->thread3.const_urb_entry_read_offset = 0; /* no const URBs */ + sf_state->thread3.urb_entry_read_length = 1; /* 1 URB per vertex */ + sf_state->thread3.urb_entry_read_offset = 0; + sf_state->thread3.dispatch_grf_start_reg = 3; + sf_state->thread4.max_threads = SF_MAX_THREADS - 1; + sf_state->thread4.urb_entry_allocation_size = URB_SF_ENTRY_SIZE - 1; + sf_state->thread4.nr_urb_entries = URB_SF_ENTRIES; + sf_state->thread4.stats_enable = 1; + sf_state->sf5.viewport_transform = FALSE; /* skip viewport */ + sf_state->sf6.cull_mode = BRW_CULLMODE_NONE; + sf_state->sf6.scissor = 0; + sf_state->sf7.trifan_pv = 2; + sf_state->sf6.dest_org_vbias = 0x8; + sf_state->sf6.dest_org_hbias = 0x8; + + memcpy (ps_kernel, ps_kernel_static, sizeof (ps_kernel_static)); +#if 0 + ErrorF ("ps kernel: 0x%08x\n", state_base_offset + ps_kernel_offset); +#endif + memset (wm_state, 0, sizeof (*wm_state)); + wm_state->thread0.kernel_start_pointer = + (state_base_offset + ps_kernel_offset) >> 6; + wm_state->thread0.grf_reg_count = BRW_GRF_BLOCKS(PS_KERNEL_NUM_GRF); + wm_state->thread1.single_program_flow = 1; /* XXX */ + wm_state->thread1.binding_table_entry_count = 2; + /* Though we never use the scratch space in our WM kernel, it has to be + * set, and the minimum allocation is 1024 bytes. + */ + wm_state->thread2.scratch_space_base_pointer = (state_base_offset + + wm_scratch_offset) >> 10; + wm_state->thread2.per_thread_scratch_space = 0; /* 1024 bytes */ + wm_state->thread3.dispatch_grf_start_reg = 3; /* XXX */ + wm_state->thread3.const_urb_entry_read_length = 0; + wm_state->thread3.const_urb_entry_read_offset = 0; + wm_state->thread3.urb_entry_read_length = 1; /* XXX */ + wm_state->thread3.urb_entry_read_offset = 0; /* XXX */ + wm_state->wm4.stats_enable = 1; + wm_state->wm4.sampler_state_pointer = (state_base_offset + src_sampler_offset) >> 5; + wm_state->wm4.sampler_count = 1; /* 1-4 samplers used */ + wm_state->wm5.max_threads = PS_MAX_THREADS - 1; + wm_state->wm5.thread_dispatch_enable = 1; + wm_state->wm5.enable_16_pix = 1; + wm_state->wm5.enable_8_pix = 0; + wm_state->wm5.early_depth_test = 1; + + { + BEGIN_LP_RING(2); + OUT_RING(MI_FLUSH | + MI_STATE_INSTRUCTION_CACHE_FLUSH | + BRW_MI_GLOBAL_SNAPSHOT_RESET); + OUT_RING(MI_NOOP); + ADVANCE_LP_RING(); + } + +/* brw_debug (pScrn, "before base address modify"); */ + { BEGIN_LP_RING(12); + /* Match Mesa driver setup */ + OUT_RING(BRW_PIPELINE_SELECT | PIPELINE_SELECT_3D); + + /* Mesa does this. Who knows... */ + OUT_RING(BRW_CS_URB_STATE | 0); + OUT_RING((0 << 4) | /* URB Entry Allocation Size */ + (0 << 0)); /* Number of URB Entries */ + + /* Zero out the two base address registers so all offsets are absolute */ + OUT_RING(BRW_STATE_BASE_ADDRESS | 4); + OUT_RING(0 | BASE_ADDRESS_MODIFY); /* Generate state base address */ + OUT_RING(0 | BASE_ADDRESS_MODIFY); /* Surface state base address */ + OUT_RING(0 | BASE_ADDRESS_MODIFY); /* media base addr, don't care */ + OUT_RING(0x10000000 | BASE_ADDRESS_MODIFY); /* general state max addr, disabled */ + OUT_RING(0x10000000 | BASE_ADDRESS_MODIFY); /* media object state max addr, disabled */ + + /* Set system instruction pointer */ + OUT_RING(BRW_STATE_SIP | 0); + OUT_RING(state_base_offset + sip_kernel_offset); /* system instruction pointer */ + + OUT_RING(MI_NOOP); + ADVANCE_LP_RING(); } + +/* brw_debug (pScrn, "after base address modify"); */ + + { BEGIN_LP_RING(42); + /* Enable VF statistics */ + OUT_RING(BRW_3DSTATE_VF_STATISTICS | 1); + + /* Pipe control */ + OUT_RING(BRW_PIPE_CONTROL | + BRW_PIPE_CONTROL_NOWRITE | + BRW_PIPE_CONTROL_IS_FLUSH | + 2); + OUT_RING(0); /* Destination address */ + OUT_RING(0); /* Immediate data low DW */ + OUT_RING(0); /* Immediate data high DW */ + + /* Binding table pointers */ + OUT_RING(BRW_3DSTATE_BINDING_TABLE_POINTERS | 4); + OUT_RING(0); /* vs */ + OUT_RING(0); /* gs */ + OUT_RING(0); /* clip */ + OUT_RING(0); /* sf */ + /* Only the PS uses the binding table */ + OUT_RING(state_base_offset + binding_table_offset); /* ps */ + + /* Blend constant color (magenta is fun) */ + OUT_RING(BRW_3DSTATE_CONSTANT_COLOR | 3); + OUT_RING(float_to_uint (1.0)); + OUT_RING(float_to_uint (0.0)); + OUT_RING(float_to_uint (1.0)); + OUT_RING(float_to_uint (1.0)); + + /* The drawing rectangle clipping is always on. Set it to values that + * shouldn't do any clipping. + */ + OUT_RING(BRW_3DSTATE_DRAWING_RECTANGLE | 2); /* XXX 3 for BLC or CTG */ + OUT_RING(0x00000000); /* ymin, xmin */ + OUT_RING((pScrn->virtualX - 1) | + (pScrn->virtualY - 1) << 16); /* ymax, xmax */ + OUT_RING(0x00000000); /* yorigin, xorigin */ + + /* skip the depth buffer */ + /* skip the polygon stipple */ + /* skip the polygon stipple offset */ + /* skip the line stipple */ + + /* Set the pointers to the 3d pipeline state */ + OUT_RING(BRW_3DSTATE_PIPELINED_POINTERS | 5); + OUT_RING(state_base_offset + vs_offset); /* 32 byte aligned */ + OUT_RING(BRW_GS_DISABLE); /* disable GS, resulting in passthrough */ + OUT_RING(BRW_CLIP_DISABLE); /* disable CLIP, resulting in passthrough */ + OUT_RING(state_base_offset + sf_offset); /* 32 byte aligned */ + OUT_RING(state_base_offset + wm_offset); /* 32 byte aligned */ + OUT_RING(state_base_offset + cc_offset); /* 64 byte aligned */ + + /* URB fence */ + OUT_RING(BRW_URB_FENCE | + UF0_CS_REALLOC | + UF0_SF_REALLOC | + UF0_CLIP_REALLOC | + UF0_GS_REALLOC | + UF0_VS_REALLOC | + 1); + OUT_RING(((urb_clip_start + urb_clip_size) << UF1_CLIP_FENCE_SHIFT) | + ((urb_gs_start + urb_gs_size) << UF1_GS_FENCE_SHIFT) | + ((urb_vs_start + urb_vs_size) << UF1_VS_FENCE_SHIFT)); + OUT_RING(((urb_cs_start + urb_cs_size) << UF2_CS_FENCE_SHIFT) | + ((urb_sf_start + urb_sf_size) << UF2_SF_FENCE_SHIFT)); + + /* Constant buffer state */ + OUT_RING(BRW_CS_URB_STATE | 0); + OUT_RING(((URB_CS_ENTRY_SIZE - 1) << 4) | /* URB Entry Allocation Size */ + (URB_CS_ENTRIES << 0)); /* Number of URB Entries */ + + /* Set up the pointer to our vertex buffer */ + OUT_RING(BRW_3DSTATE_VERTEX_BUFFERS | 2); + OUT_RING((0 << VB0_BUFFER_INDEX_SHIFT) | + VB0_VERTEXDATA | + ((4 * 4) << VB0_BUFFER_PITCH_SHIFT)); /* four 32-bit floats per vertex */ + OUT_RING(state_base_offset + vb_offset); + OUT_RING(3); /* four corners to our rectangle */ + + /* Set up our vertex elements, sourced from the single vertex buffer. */ + OUT_RING(BRW_3DSTATE_VERTEX_ELEMENTS | 3); + /* offset 0: X,Y -> {X, Y, 1.0, 1.0} */ + OUT_RING((0 << VE0_VERTEX_BUFFER_INDEX_SHIFT) | + VE0_VALID | + (BRW_SURFACEFORMAT_R32G32_FLOAT << VE0_FORMAT_SHIFT) | + (0 << VE0_OFFSET_SHIFT)); + OUT_RING((BRW_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT) | + (BRW_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT) | + (BRW_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT) | + (BRW_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT) | + (0 << VE1_DESTINATION_ELEMENT_OFFSET_SHIFT)); + /* offset 8: S0, T0 -> {S0, T0, 1.0, 1.0} */ + OUT_RING((0 << VE0_VERTEX_BUFFER_INDEX_SHIFT) | + VE0_VALID | + (BRW_SURFACEFORMAT_R32G32_FLOAT << VE0_FORMAT_SHIFT) | + (8 << VE0_OFFSET_SHIFT)); + OUT_RING((BRW_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_0_SHIFT) | + (BRW_VFCOMPONENT_STORE_SRC << VE1_VFCOMPONENT_1_SHIFT) | + (BRW_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_2_SHIFT) | + (BRW_VFCOMPONENT_STORE_1_FLT << VE1_VFCOMPONENT_3_SHIFT) | + (4 << VE1_DESTINATION_ELEMENT_OFFSET_SHIFT)); + + OUT_RING(MI_NOOP); /* pad to quadword */ + ADVANCE_LP_RING(); } + + dxo = dstRegion->extents.x1; + dyo = dstRegion->extents.y1; + + pbox = REGION_RECTS(dstRegion); + nbox = REGION_NUM_RECTS(dstRegion); + while (nbox--) + { + int box_x1 = pbox->x1; + int box_y1 = pbox->y1; + int box_x2 = pbox->x2; + int box_y2 = pbox->y2; + int i; + float src_scale_x, src_scale_y; + + if (!first_output) { + /* Since we use the same little vertex buffer over and over, sync for + * subsequent rectangles. + */ + if (pI830->AccelInfoRec && pI830->AccelInfoRec->NeedToSync) { + (*pI830->AccelInfoRec->Sync)(pScrn); + pI830->AccelInfoRec->NeedToSync = FALSE; + } + } + + pbox++; + + /* Use normalized texture coordinates */ + src_scale_x = (float)1.0 / (float)drw_w; + src_scale_y = (float)1.0 / (float)drw_h; + + i = 0; + vb[i++] = (box_x2 - dxo) * src_scale_x; + vb[i++] = (box_y2 - dyo) * src_scale_y; + vb[i++] = (float) box_x2; + vb[i++] = (float) box_y2; + + vb[i++] = (box_x1 - dxo) * src_scale_x; + vb[i++] = (box_y2 - dyo) * src_scale_y; + vb[i++] = (float) box_x1; + vb[i++] = (float) box_y2; + + vb[i++] = (box_x1 - dxo) * src_scale_x; + vb[i++] = (box_y1 - dyo) * src_scale_y; + vb[i++] = (float) box_x1; + vb[i++] = (float) box_y1; + +#if 0 + ErrorF ("before EU_ATT 0x%08x%08x EU_ATT_DATA 0x%08x%08x\n", + INREG(BRW_EU_ATT_1), INREG(BRW_EU_ATT_0), + INREG(BRW_EU_ATT_DATA_1), INREG(BRW_EU_ATT_DATA_0)); + + OUTREG(BRW_VF_CTL, + BRW_VF_CTL_SNAPSHOT_MUX_SELECT_THREADID | + BRW_VF_CTL_SNAPSHOT_TYPE_VERTEX_INDEX | + BRW_VF_CTL_SNAPSHOT_ENABLE); + OUTREG(BRW_VF_STRG_VAL, 0); +#endif + +#if 0 + OUTREG(BRW_VS_CTL, + BRW_VS_CTL_SNAPSHOT_ALL_THREADS | + BRW_VS_CTL_SNAPSHOT_MUX_VALID_COUNT | + BRW_VS_CTL_THREAD_SNAPSHOT_ENABLE); + + OUTREG(BRW_VS_STRG_VAL, 0); +#endif + +#if WATCH_SF + OUTREG(BRW_SF_CTL, + BRW_SF_CTL_SNAPSHOT_MUX_VERTEX_COUNT | + BRW_SF_CTL_SNAPSHOT_ALL_THREADS | + BRW_SF_CTL_THREAD_SNAPSHOT_ENABLE); + OUTREG(BRW_SF_STRG_VAL, 0); +#endif + +#if WATCH_WIZ + OUTREG(BRW_WIZ_CTL, + BRW_WIZ_CTL_SNAPSHOT_MUX_SUBSPAN_INSTANCE | + BRW_WIZ_CTL_SNAPSHOT_ALL_THREADS | + BRW_WIZ_CTL_SNAPSHOT_ENABLE); + OUTREG(BRW_WIZ_STRG_VAL, + (box_x1) | (box_y1 << 16)); +#endif + +#if 0 + OUTREG(BRW_TS_CTL, + BRW_TS_CTL_SNAPSHOT_MESSAGE_ERROR | + BRW_TS_CTL_SNAPSHOT_ALL_CHILD_THREADS | + BRW_TS_CTL_SNAPSHOT_ALL_ROOT_THREADS | + BRW_TS_CTL_SNAPSHOT_ENABLE); +#endif + + BEGIN_LP_RING(6); + OUT_RING(BRW_3DPRIMITIVE | + BRW_3DPRIMITIVE_VERTEX_SEQUENTIAL | + (_3DPRIM_RECTLIST << BRW_3DPRIMITIVE_TOPOLOGY_SHIFT) | + (0 << 9) | /* CTG - indirect vertex count */ + 4); + OUT_RING(3); /* vertex count per instance */ + OUT_RING(0); /* start vertex offset */ + OUT_RING(1); /* single instance */ + OUT_RING(0); /* start instance location */ + OUT_RING(0); /* index buffer offset, ignored */ + ADVANCE_LP_RING(); + +#if 0 + for (j = 0; j < 100000; j++) { + ctl = INREG(BRW_VF_CTL); + if (ctl & BRW_VF_CTL_SNAPSHOT_COMPLETE) + break; + } + + rdata = INREG(BRW_VF_RDATA); + OUTREG(BRW_VF_CTL, 0); + ErrorF ("VF_CTL: 0x%08x VF_RDATA: 0x%08x\n", ctl, rdata); +#endif + +#if 0 + for (j = 0; j < 1000000; j++) { + ctl = INREG(BRW_VS_CTL); + if (ctl & BRW_VS_CTL_SNAPSHOT_COMPLETE) + break; + } + + rdata = INREG(BRW_VS_RDATA); + for (k = 0; k <= 3; k++) { + OUTREG(BRW_VS_CTL, + BRW_VS_CTL_SNAPSHOT_COMPLETE | + (k << 8)); + rdata = INREG(BRW_VS_RDATA); + ErrorF ("VS_CTL: 0x%08x VS_RDATA(%d): 0x%08x\n", ctl, k, rdata); + } + + OUTREG(BRW_VS_CTL, 0); +#endif + +#if WATCH_SF + for (j = 0; j < 1000000; j++) { + ctl = INREG(BRW_SF_CTL); + if (ctl & BRW_SF_CTL_SNAPSHOT_COMPLETE) + break; + } + + for (k = 0; k <= 7; k++) { + OUTREG(BRW_SF_CTL, + BRW_SF_CTL_SNAPSHOT_COMPLETE | + (k << 8)); + rdata = INREG(BRW_SF_RDATA); + ErrorF ("SF_CTL: 0x%08x SF_RDATA(%d): 0x%08x\n", ctl, k, rdata); + } + + OUTREG(BRW_SF_CTL, 0); +#endif + +#if WATCH_WIZ + for (j = 0; j < 100000; j++) { + ctl = INREG(BRW_WIZ_CTL); + if (ctl & BRW_WIZ_CTL_SNAPSHOT_COMPLETE) + break; + } + + rdata = INREG(BRW_WIZ_RDATA); + OUTREG(BRW_WIZ_CTL, 0); + ErrorF ("WIZ_CTL: 0x%08x WIZ_RDATA: 0x%08x\n", ctl, rdata); +#endif + +#if 0 + for (j = 0; j < 100000; j++) { + ctl = INREG(BRW_TS_CTL); + if (ctl & BRW_TS_CTL_SNAPSHOT_COMPLETE) + break; + } + + rdata = INREG(BRW_TS_RDATA); + OUTREG(BRW_TS_CTL, 0); + ErrorF ("TS_CTL: 0x%08x TS_RDATA: 0x%08x\n", ctl, rdata); + + ErrorF ("after EU_ATT 0x%08x%08x EU_ATT_DATA 0x%08x%08x\n", + INREG(BRW_EU_ATT_1), INREG(BRW_EU_ATT_0), + INREG(BRW_EU_ATT_DATA_1), INREG(BRW_EU_ATT_DATA_0)); +#endif + +#if 0 + for (j = 0; j < 256; j++) { + OUTREG(BRW_TD_CTL, j << BRW_TD_CTL_MUX_SHIFT); + rdata = INREG(BRW_TD_RDATA); + ErrorF ("TD_RDATA(%d): 0x%08x\n", j, rdata); + } +#endif + first_output = FALSE; + if (pI830->AccelInfoRec) + pI830->AccelInfoRec->NeedToSync = TRUE; + } + + if (pI830->AccelInfoRec) + (*pI830->AccelInfoRec->Sync)(pScrn); +#if WATCH_STATS + I830PrintErrorState (pScrn); +#endif +} + static FBLinearPtr I830AllocateMemory(ScrnInfoPtr pScrn, FBLinearPtr linear, int size) { @@ -1905,6 +2873,19 @@ I830AllocateMemory(ScrnInfoPtr pScrn, FBLinearPtr linear, int size) return new_linear; } +/* + * The source rectangle of the video is defined by (src_x, src_y, src_w, src_h). + * The dest rectangle of the video is defined by (drw_x, drw_y, drw_w, drw_h). + * id is a fourcc code for the format of the video. + * buf is the pointer to the source data in system memory. + * width and height are the w/h of the source data. + * If "sync" is TRUE, then we must be finished with *buf at the point of return + * (which we always are). + * clipBoxes is the clipping region in screen space. + * data is a pointer to our port private. + * pDraw is a Drawable, which might not be the screen in the case of + * compositing. It's a new argument to the function in the 1.1 server. + */ static int I830PutImage(ScrnInfoPtr pScrn, short src_x, short src_y, @@ -1922,9 +2903,11 @@ I830PutImage(ScrnInfoPtr pScrn, I830OverlayRegPtr overlay = (I830OverlayRegPtr) (pI830->FbBase + pI830->OverlayMem->Start); INT32 x1, x2, y1, y2; - int srcPitch, srcPitch2 = 0, dstPitch; + int srcPitch, srcPitch2 = 0, dstPitch, destId; int top, left, npixels, nlines, size, loops; BoxRec dstBox; + int pitchAlignMask; + int extraLinear; DPRINTF(PFX, "I830PutImage: src: (%d,%d)(%d,%d), dst: (%d,%d)(%d,%d)\n" "width %d, height %d\n", src_x, src_y, src_w, src_h, drw_x, drw_y, @@ -1968,63 +2951,85 @@ I830PutImage(ScrnInfoPtr pScrn, width, height)) return Success; + destId = id; switch (id) { case FOURCC_YV12: case FOURCC_I420: srcPitch = (width + 3) & ~3; srcPitch2 = ((width >> 1) + 3) & ~3; -#if 1 - if (pI830->rotation & (RR_Rotate_90 | RR_Rotate_270)) { - dstPitch = ((height / 2) + 63) & ~63; - size = dstPitch * width * 3; - } else { - dstPitch = ((width / 2) + 63) & ~63; /* of chroma */ - size = dstPitch * height * 3; + if (pPriv->textured) { + destId = FOURCC_YUY2; } -#else + break; + case FOURCC_UYVY: + case FOURCC_YUY2: + default: + srcPitch = width << 1; + break; + } + + /* Only needs to be DWORD-aligned for textured on i915, but overlay has + * stricter requirements. + */ + if (pPriv->textured) { + pitchAlignMask = 3; + } else { + if (IS_I965G(pI830)) + pitchAlignMask = 255; + else + pitchAlignMask = 63; + } + + /* Determine the desired destination pitch (representing the chroma's pitch, + * in the planar case. + */ + switch (destId) { + case FOURCC_YV12: + case FOURCC_I420: if (pI830->rotation & (RR_Rotate_90 | RR_Rotate_270)) { - dstPitch = ((height / 2) + 511) & ~511; + dstPitch = ((height / 2) + pitchAlignMask) & ~pitchAlignMask; size = dstPitch * width * 3; } else { - dstPitch = ((width / 2) + 511) & ~511; /* of chroma */ + dstPitch = ((width / 2) + pitchAlignMask) & ~pitchAlignMask; size = dstPitch * height * 3; } -#endif break; case FOURCC_UYVY: case FOURCC_YUY2: default: - srcPitch = width << 1; -#if 1 if (pI830->rotation & (RR_Rotate_90 | RR_Rotate_270)) { - dstPitch = ((height << 1) + 63) & ~63; + dstPitch = ((height << 1) + pitchAlignMask) & ~pitchAlignMask; size = dstPitch * width; } else { - dstPitch = ((width << 1) + 63) & ~63; /* of chroma */ + dstPitch = ((width << 1) + pitchAlignMask) & ~pitchAlignMask; size = dstPitch * height; } -#else - if (pI830->rotation & (RR_Rotate_90 | RR_Rotate_270)) { - dstPitch = ((height << 1) + 511) & ~511; - size = dstPitch * width; - } else { - dstPitch = ((width << 1) + 511) & ~511; /* of chroma */ - size = dstPitch * height; - } -#endif break; } +#if 0 ErrorF("srcPitch: %d, dstPitch: %d, size: %d\n", srcPitch, dstPitch, size); +#endif + + if (IS_I965G(pI830)) + extraLinear = BRW_LINEAR_EXTRA; + else + extraLinear = 0; /* size is multiplied by 2 because we have two buffers that are flipping */ pPriv->linear = I830AllocateMemory(pScrn, pPriv->linear, - (pPriv->doubleBuffer ? size * 2 : size) / pI830->cpp); + (extraLinear + + (pPriv->doubleBuffer ? size * 2 : size)) / + pI830->cpp); if(!pPriv->linear || pPriv->linear->offset < (pScrn->virtualX * pScrn->virtualY)) return BadAlloc; /* fixup pointers */ +#if 0 + pPriv->YBuf0offset = pScrn->fbOffset + pPriv->linear->offset * pI830->cpp; +#else pPriv->YBuf0offset = pI830->FrontBuffer.Start + pPriv->linear->offset * pI830->cpp; +#endif if (pI830->rotation & (RR_Rotate_90 | RR_Rotate_270)) { pPriv->UBuf0offset = pPriv->YBuf0offset + (dstPitch * 2 * width); pPriv->VBuf0offset = pPriv->UBuf0offset + (dstPitch * width / 2); @@ -2045,7 +3050,9 @@ I830PutImage(ScrnInfoPtr pScrn, /* Make sure this buffer isn't in use */ loops = 0; - if (*pI830->overlayOn && pPriv->doubleBuffer && (overlay->OCMD & OVERLAY_ENABLE)) { + if (!pPriv->textured && *pI830->overlayOn && pPriv->doubleBuffer && + (overlay->OCMD & OVERLAY_ENABLE)) + { while (loops < 1000000) { #if USE_USLEEP_FOR_VIDEO usleep(10); @@ -2074,32 +3081,58 @@ I830PutImage(ScrnInfoPtr pScrn, left = (x1 >> 16) & ~1; npixels = ((((x2 + 0xffff) >> 16) + 1) & ~1) - left; + if (pPriv->textured) { + /* For textured video, we don't double buffer, and instead just wait for + * acceleration to finish before writing the new video data into + * framebuffer. + */ + if (pI830->AccelInfoRec && pI830->AccelInfoRec->NeedToSync) { + (*pI830->AccelInfoRec->Sync)(pScrn); + pI830->AccelInfoRec->NeedToSync = FALSE; + } + } + switch (id) { case FOURCC_YV12: case FOURCC_I420: top &= ~1; nlines = ((((y2 + 0xffff) >> 16) + 1) & ~1) - top; - I830CopyPlanarData(pScrn, buf, srcPitch, srcPitch2, dstPitch, height, top, left, - nlines, npixels, id); + if (pPriv->textured) { + I830CopyPlanarToPackedData(pScrn, pPriv, buf, srcPitch, srcPitch2, + dstPitch, height, top, left, nlines, + npixels, id); + } else { + I830CopyPlanarData(pScrn, pPriv, buf, srcPitch, srcPitch2, dstPitch, + height, top, left, nlines, npixels, id); + } break; case FOURCC_UYVY: case FOURCC_YUY2: default: nlines = ((y2 + 0xffff) >> 16) - top; - I830CopyPackedData(pScrn, buf, srcPitch, dstPitch, top, left, nlines, - npixels); + I830CopyPackedData(pScrn, pPriv, buf, srcPitch, dstPitch, top, left, + nlines, npixels); break; } - /* update cliplist */ - if (!RegionsEqual(&pPriv->clip, clipBoxes)) { - REGION_COPY(pScrn->pScreen, &pPriv->clip, clipBoxes); - xf86XVFillKeyHelper(pScreen, pPriv->colorKey, clipBoxes); - } - - I830DisplayVideo(pScrn, id, width, height, dstPitch, - x1, y1, x2, y2, &dstBox, src_w, src_h, drw_w, drw_h); + if (!pPriv->textured) { + /* update cliplist */ + if (!RegionsEqual(&pPriv->clip, clipBoxes)) { + REGION_COPY(pScrn->pScreen, &pPriv->clip, clipBoxes); + xf86XVFillKeyHelper(pScreen, pPriv->colorKey, clipBoxes); + } + I830DisplayVideo(pScrn, destId, width, height, dstPitch, + x1, y1, x2, y2, &dstBox, src_w, src_h, drw_w, drw_h); + } else if (IS_I965G(pI830)) { + BroadwaterDisplayVideoTextured (pScrn, pPriv, destId, clipBoxes, width, height, + dstPitch, x1, y1, x2, y2, + src_w, src_h, drw_w, drw_h, pDraw); + } else { + I915DisplayVideoTextured(pScrn, pPriv, destId, clipBoxes, width, height, + dstPitch, x1, y1, x2, y2, + src_w, src_h, drw_w, drw_h, pDraw); + } pPriv->videoStatus = CLIENT_VIDEO_ON; return Success; @@ -2109,23 +3142,27 @@ static int I830QueryImageAttributes(ScrnInfoPtr pScrn, int id, unsigned short *w, unsigned short *h, - int *pitches, int *offsets) + int *pitches, int *offsets, Bool textured) { I830Ptr pI830 = I830PTR(pScrn); int size, tmp; +#if 0 ErrorF("I830QueryImageAttributes: w is %d, h is %d\n", *w, *h); +#endif - if (IS_845G(pI830) || IS_I830(pI830)) { - if (*w > IMAGE_MAX_WIDTH_LEGACY) - *w = IMAGE_MAX_WIDTH_LEGACY; - if (*h > IMAGE_MAX_HEIGHT_LEGACY) - *h = IMAGE_MAX_HEIGHT_LEGACY; - } else { - if (*w > IMAGE_MAX_WIDTH) - *w = IMAGE_MAX_WIDTH; - if (*h > IMAGE_MAX_HEIGHT) - *h = IMAGE_MAX_HEIGHT; + if (!textured) { + if (IS_845G(pI830) || IS_I830(pI830)) { + if (*w > IMAGE_MAX_WIDTH_LEGACY) + *w = IMAGE_MAX_WIDTH_LEGACY; + if (*h > IMAGE_MAX_HEIGHT_LEGACY) + *h = IMAGE_MAX_HEIGHT_LEGACY; + } else { + if (*w > IMAGE_MAX_WIDTH) + *w = IMAGE_MAX_WIDTH; + if (*h > IMAGE_MAX_HEIGHT) + *h = IMAGE_MAX_HEIGHT; + } } *w = (*w + 1) & ~1; @@ -2178,6 +3215,24 @@ I830QueryImageAttributes(ScrnInfoPtr pScrn, return size; } +static int +I830QueryImageAttributesOverlay(ScrnInfoPtr pScrn, + int id, + unsigned short *w, unsigned short *h, + int *pitches, int *offsets) +{ + return I830QueryImageAttributes(pScrn, id, w, h, pitches, offsets, FALSE); +} + +static int +I830QueryImageAttributesTextured(ScrnInfoPtr pScrn, + int id, + unsigned short *w, unsigned short *h, + int *pitches, int *offsets) +{ + return I830QueryImageAttributes(pScrn, id, w, h, pitches, offsets, TRUE); +} + static void I830BlockHandler(int i, pointer blockData, pointer pTimeout, pointer pReadmask) @@ -2196,7 +3251,11 @@ I830BlockHandler(int i, pScreen->BlockHandler = I830BlockHandler; if (pPriv->videoStatus & TIMER_MASK) { +#if 1 Time now = currentTime.milliseconds; +#else + UpdateCurrentTime(); +#endif if (pPriv->videoStatus & OFF_TIMER) { if (pPriv->offTime < now) { /* Turn off the overlay */ @@ -2296,7 +3355,11 @@ I830AllocateSurface(ScrnInfoPtr pScrn, surface->offsets[0] = linear->offset * bpp; surface->devPrivate.ptr = (pointer) pPriv; +#if 0 + memset(pI830->FbBase + pScrn->fbOffset + surface->offsets[0], 0, size); +#else memset(pI830->FbBase + pI830->FrontBuffer.Start + surface->offsets[0], 0, size); +#endif return Success; } @@ -2542,37 +3605,39 @@ I830VideoSwitchModeAfter(ScrnInfoPtr pScrn, DisplayModePtr mode) } } - if (pPriv->pipe == 0) { - if (INREG(PIPEACONF) & PIPEACONF_DOUBLE_WIDE) { - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Disabling XVideo output because Pipe A is in double-wide mode.\n"); - pPriv->overlayOK = FALSE; - } else if (!pPriv->overlayOK) { - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Re-enabling XVideo output because Pipe A is now in single-wide mode.\n"); - pPriv->overlayOK = TRUE; + if (!IS_I965G(pI830)) { + if (pPriv->pipe == 0) { + if (INREG(PIPEACONF) & PIPEACONF_DOUBLE_WIDE) { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Disabling XVideo output because Pipe A is in double-wide mode.\n"); + pPriv->overlayOK = FALSE; + } else if (!pPriv->overlayOK) { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Re-enabling XVideo output because Pipe A is now in single-wide mode.\n"); + pPriv->overlayOK = TRUE; + } } - } - if (pPriv->pipe == 1) { - if (INREG(PIPEBCONF) & PIPEBCONF_DOUBLE_WIDE) { - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Disabling XVideo output because Pipe B is in double-wide mode.\n"); - pPriv->overlayOK = FALSE; - } else if (!pPriv->overlayOK) { - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Re-enabling XVideo output because Pipe B is now in single-wide mode.\n"); - pPriv->overlayOK = TRUE; + if (pPriv->pipe == 1) { + if (INREG(PIPEBCONF) & PIPEBCONF_DOUBLE_WIDE) { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Disabling XVideo output because Pipe B is in double-wide mode.\n"); + pPriv->overlayOK = FALSE; + } else if (!pPriv->overlayOK) { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Re-enabling XVideo output because Pipe B is now in single-wide mode.\n"); + pPriv->overlayOK = TRUE; + } } } /* Check we have an LFP connected */ if ((pPriv->pipe == 1 && pI830->operatingDevices & (PIPE_LFP << 8)) || (pPriv->pipe == 0 && pI830->operatingDevices & PIPE_LFP) ) { - size = pI830->pipe ? INREG(PIPEBSRC) : INREG(PIPEASRC); + size = pPriv->pipe ? INREG(PIPEBSRC) : INREG(PIPEASRC); hsize = (size >> 16) & 0x7FF; vsize = size & 0x7FF; - active = pI830->pipe ? (INREG(VTOTAL_B) & 0x7FF) : (INREG(VTOTAL_A) & 0x7FF); + active = pPriv->pipe ? (INREG(VTOTAL_B) & 0x7FF) : (INREG(VTOTAL_A) & 0x7FF); if (vsize < active && hsize > 1024) I830SetOneLineModeRatio(pScrn); |