| Message ID | 1252092408-29419-1-git-send-email-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Fri, 2009-09-04 at 20:26 +0100, Chris Wilson wrote: > While not yet fully accelerating gradient patterns, by using pixman to > compute the gradient image and copying that to a pixmap to use as the > source, we avoid incurring the GPU stall suffered currently from > reading back the destination surface. The destination Picture (the one passed in to our uxa hooks) is supposed to always have a pDrawable, and if it doesn't, that should have been caught at a higher level. > > Speedups on i915: > firefox-talos-svg: 710378.14 -> 549262.96: 1.29x speedup > > No slowdowns. > > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > --- > src/i830_render.c | 14 ++- > src/i915_render.c | 12 ++- > src/i965_render.c | 14 ++- > uxa/uxa-render.c | 261 ++++++++++++++++++++++++++++++++++++++++------------- > 4 files changed, 223 insertions(+), 78 deletions(-) > > diff --git a/src/i830_render.c b/src/i830_render.c > index e4c4623..8418bd7 100644 > --- a/src/i830_render.c > +++ b/src/i830_render.c > @@ -220,11 +220,15 @@ static Bool i830_get_blend_cntl(ScrnInfoPtr pScrn, int op, PicturePtr pMask, > > static Bool i830_check_composite_texture(PicturePtr pPict, int unit) > { > - ScrnInfoPtr pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; > - int w = pPict->pDrawable->width; > - int h = pPict->pDrawable->height; > - int i; > + ScrnInfoPtr pScrn; > + int w, h, i; > > + if (!pPict->pDrawable) > + return TRUE; > + > + pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; > + w = pPict->pDrawable->width; > + h = pPict->pDrawable->height; > if ((w > 2048) || (h > 2048)) > I830FALLBACK("Picture w/h too large (%dx%d)\n", w, h); > > @@ -393,7 +397,7 @@ i830_prepare_composite(int op, PicturePtr pSrcPicture, > PicturePtr pMaskPicture, PicturePtr pDstPicture, > PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst) > { > - ScrnInfoPtr pScrn = xf86Screens[pSrcPicture->pDrawable->pScreen->myNum]; > + ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum]; > I830Ptr pI830 = I830PTR(pScrn); > Bool is_affine_src, is_affine_mask; > Bool is_nearest = FALSE; > diff --git a/src/i915_render.c b/src/i915_render.c > index c81366a..bb755c5 100644 > --- a/src/i915_render.c > +++ b/src/i915_render.c > @@ -169,11 +169,15 @@ static Bool i915_get_dest_format(PicturePtr pDstPicture, uint32_t *dst_format) > > static Bool i915_check_composite_texture(PicturePtr pPict, int unit) > { > - ScrnInfoPtr pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; > - int w = pPict->pDrawable->width; > - int h = pPict->pDrawable->height; > - int i; > + ScrnInfoPtr pScrn; > + int w, h, i; > + > + if (!pPict->pDrawable) > + return TRUE; > > + pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; > + w = pPict->pDrawable->width; > + h = pPict->pDrawable->height; > if ((w > 2048) || (h > 2048)) > I830FALLBACK("Picture w/h too large (%dx%d)\n", w, h); > > diff --git a/src/i965_render.c b/src/i965_render.c > index 1a8075b..1d88af8 100644 > --- a/src/i965_render.c > +++ b/src/i965_render.c > @@ -183,11 +183,15 @@ static Bool i965_get_dest_format(PicturePtr pDstPicture, uint32_t *dst_format) > > static Bool i965_check_composite_texture(PicturePtr pPict, int unit) > { > - ScrnInfoPtr pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; > - int w = pPict->pDrawable->width; > - int h = pPict->pDrawable->height; > - int i; > + ScrnInfoPtr pScrn; > + int w, h, i; > + > + if (!pPict->pDrawable) > + return TRUE; > > + pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; > + w = pPict->pDrawable->width; > + h = pPict->pDrawable->height; > if ((w > 8192) || (h > 8192)) > I830FALLBACK("Picture w/h too large (%dx%d)\n", w, h); > > @@ -1438,7 +1442,7 @@ i965_prepare_composite(int op, PicturePtr pSrcPicture, > PicturePtr pMaskPicture, PicturePtr pDstPicture, > PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst) > { > - ScrnInfoPtr pScrn = xf86Screens[pSrcPicture->pDrawable->pScreen->myNum]; > + ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum]; > I830Ptr pI830 = I830PTR(pScrn); > struct gen4_render_state *render_state= pI830->gen4_render_state; > gen4_composite_op *composite_op = &render_state->composite_op; > diff --git a/uxa/uxa-render.c b/uxa/uxa-render.c > index 13128ed..fde10b7 100644 > --- a/uxa/uxa-render.c > +++ b/uxa/uxa-render.c > @@ -314,6 +314,131 @@ uxa_try_driver_solid_fill(PicturePtr pSrc, > return 1; > } > > +static PicturePtr > +uxa_picture_from_pixman_image (ScreenPtr pScreen, > + pixman_image_t *image, > + pixman_format_code_t format) > +{ > + PicturePtr pPicture; > + PixmapPtr pPixmap; > + GCPtr pGC; > + int width, height, depth; > + int error; > + > + width = pixman_image_get_width (image); > + height = pixman_image_get_height (image); > + depth = pixman_image_get_depth (image); > + > + pPixmap = (*pScreen->CreatePixmap) (pScreen, width, height, depth, > + UXA_CREATE_PIXMAP_FOR_MAP); > + if (!pPixmap) > + return 0; > + > + pPicture = CreatePicture (0, &pPixmap->drawable, > + PictureMatchFormat (pScreen, depth, format), > + 0, 0, serverClient, &error); > + (*pScreen->DestroyPixmap) (pPixmap); > + > + > + pPixmap = GetScratchPixmapHeader(pScreen, width, height, depth, > + BitsPerPixel (depth), > + pixman_image_get_stride (image), > + pixman_image_get_data (image)); > + if (!pPixmap) > + { > + FreePicture (pPicture, 0); > + return 0; > + } > + > + pGC = GetScratchGC (depth, pScreen); > + if (!pGC) > + { > + FreeScratchPixmapHeader (pPixmap); > + FreePicture (pPicture, 0); > + return 0; > + } > + ValidateGC (pPicture->pDrawable, pGC); > + > + (*pGC->ops->CopyArea) (&pPixmap->drawable, pPicture->pDrawable, > + pGC, 0, 0, width, height, 0, 0); > + > + FreeScratchGC (pGC); > + FreeScratchPixmapHeader (pPixmap); > + > + return pPicture; > +} > + > +static PicturePtr > +uxa_get_pattern (ScreenPtr pScreen, > + PicturePtr pPict, > + pixman_format_code_t format, > + INT16 x, INT16 y, > + CARD16 width, CARD16 height) > +{ > + pixman_image_t *source, *image; > + > + source = image_from_pict (pPict, 0, 0); > + if (!source) > + return 0; > + > + image = pixman_image_create_bits (format, width, height, NULL, 0); > + if (!image) { > + pixman_image_unref (source); > + return 0; > + } > + > + pixman_image_composite (PIXMAN_OP_SRC, > + source, NULL, image, > + x, y, > + 0, 0, > + 0, 0, > + width, height); > + pixman_image_unref (source); > + > + pPict = uxa_picture_from_pixman_image (pScreen, image, format); > + pixman_image_unref (image); > + > + return pPict; > +} > + > +static PicturePtr > +uxa_get_source (ScreenPtr pScreen, > + PicturePtr pPict, > + INT16 x, INT16 y, > + CARD16 width, CARD16 height, > + INT16 *out_x, INT16 *out_y) > +{ > + if (pPict->pDrawable) { > + *out_x = x + pPict->pDrawable->x; > + *out_y = y + pPict->pDrawable->y; > + return pPict; > + } > + > + *out_x = 0; > + *out_y = 0; > + return uxa_get_pattern (pScreen, pPict, > + PICT_a8r8g8b8, x, y, width, height); > +} > + > +static PicturePtr > +uxa_get_mask (ScreenPtr pScreen, > + PicturePtr pPict, > + INT16 x, INT16 y, > + INT16 width, INT16 height, > + INT16 *out_x, INT16 *out_y) > +{ > + if (pPict->pDrawable) { > + *out_x = x + pPict->pDrawable->x; > + *out_y = y + pPict->pDrawable->y; > + return pPict; > + } > + > + *out_x = 0; > + *out_y = 0; > + return uxa_get_pattern (pScreen, pPict, > + PICT_a8, x, y, width, height); > +} > + > static int > uxa_try_driver_composite_rects(CARD8 op, > PicturePtr pSrc, > @@ -333,7 +458,7 @@ uxa_try_driver_composite_rects(CARD8 op, > { > return -1; > } > - > + > pDstPix = uxa_get_offscreen_pixmap(pDst->pDrawable, &dst_off_x, &dst_off_y); > if (!pDstPix) > return 0; > @@ -453,48 +578,84 @@ uxa_try_driver_composite(CARD8 op, > int nbox; > int src_off_x, src_off_y, mask_off_x, mask_off_y, dst_off_x, dst_off_y; > PixmapPtr pSrcPix, pMaskPix = NULL, pDstPix; > + PicturePtr localSrc, localMask = NULL; > > xDst += pDst->pDrawable->x; > yDst += pDst->pDrawable->y; > > + localSrc = uxa_get_source (pDst->pDrawable->pScreen, > + pSrc, xSrc, ySrc, width, height, > + &xSrc, &ySrc); > + if (! localSrc) > + return 0; > + > if (pMask) { > - xMask += pMask->pDrawable->x; > - yMask += pMask->pDrawable->y; > - } > + localMask = uxa_get_mask (pDst->pDrawable->pScreen, > + pMask, xMask, yMask, width, height, > + &xMask, &yMask); > + if (! localMask) { > + if (localSrc != pSrc) > + FreePicture (localSrc, 0); > > - xSrc += pSrc->pDrawable->x; > - ySrc += pSrc->pDrawable->y; > + return 0; > + } > + } > > if (uxa_screen->info->check_composite && > - !(*uxa_screen->info->check_composite) (op, pSrc, pMask, pDst)) > + !(*uxa_screen->info->check_composite) (op, localSrc, localMask, pDst)) > { > + if (localSrc != pSrc) > + FreePicture (localSrc, 0); > + if (localMask && localMask != pMask) > + FreePicture (localMask, 0); > + > return -1; > } > > - if (!miComputeCompositeRegion (®ion, pSrc, pMask, pDst, > + if (!miComputeCompositeRegion (®ion, localSrc, localMask, pDst, > xSrc, ySrc, xMask, yMask, xDst, yDst, > width, height)) > + { > + if (localSrc != pSrc) > + FreePicture (localSrc, 0); > + if (localMask && localMask != pMask) > + FreePicture (localMask, 0); > + > return 1; > + } > > pDstPix = uxa_get_offscreen_pixmap (pDst->pDrawable, &dst_off_x, &dst_off_y); > > - pSrcPix = uxa_get_offscreen_pixmap (pSrc->pDrawable, &src_off_x, &src_off_y); > + pSrcPix = uxa_get_offscreen_pixmap (localSrc->pDrawable, > + &src_off_x, &src_off_y); > > - if (pMask) > - pMaskPix = uxa_get_offscreen_pixmap (pMask->pDrawable, &mask_off_x, > - &mask_off_y); > + if (localMask) > + pMaskPix = uxa_get_offscreen_pixmap (localMask->pDrawable, > + &mask_off_x, &mask_off_y); > > - if (!pDstPix || !pSrcPix || (pMask && !pMaskPix)) { > + if (!pDstPix || !pSrcPix || (localMask && !pMaskPix)) { > REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); > + > + if (localSrc != pSrc) > + FreePicture (localSrc, 0); > + if (localMask && localMask != pMask) > + FreePicture (localMask, 0); > + > return 0; > } > > REGION_TRANSLATE(pScreen, ®ion, dst_off_x, dst_off_y); > > - if (!(*uxa_screen->info->prepare_composite) (op, pSrc, pMask, pDst, pSrcPix, > - pMaskPix, pDstPix)) > + if (!(*uxa_screen->info->prepare_composite) (op, localSrc, localMask, pDst, > + pSrcPix, pMaskPix, pDstPix)) > { > REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); > + > + if (localSrc != pSrc) > + FreePicture (localSrc, 0); > + if (localMask && localMask != pMask) > + FreePicture (localMask, 0); > + > return -1; > } > > @@ -523,6 +684,12 @@ uxa_try_driver_composite(CARD8 op, > (*uxa_screen->info->done_composite) (pDstPix); > > REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); > + > + if (localSrc != pSrc) > + FreePicture (localSrc, 0); > + if (localMask && localMask != pMask) > + FreePicture (localMask, 0); > + > return 1; > } > > @@ -636,14 +803,11 @@ uxa_composite(CARD8 op, > Bool saveMaskRepeat = pMask ? pMask->repeat : 0; > RegionRec region; > > - /* We currently don't support acceleration of gradients, or other pictures > - * with a NULL pDrawable. > - */ > - if (uxa_screen->swappedOut || > - pSrc->pDrawable == NULL || (pMask != NULL && pMask->pDrawable == NULL)) > - { > + if (uxa_screen->swappedOut) > goto fallback; > - } > + > + if (pSrc->pDrawable == NULL || (pMask && pMask->pDrawable == NULL)) > + goto composite; > > /* Remove repeat in source if useless */ > if (pSrc->repeat && !pSrc->transform && xSrc >= 0 && > @@ -744,6 +908,7 @@ uxa_composite(CARD8 op, > (yMask + height) <= pMask->pDrawable->height) > pMask->repeat = 0; > > +composite: > if (uxa_screen->info->prepare_composite && > !pSrc->alphaMap && (!pMask || !pMask->alphaMap) && !pDst->alphaMap) > { > @@ -757,7 +922,8 @@ uxa_composite(CARD8 op, > /* For generic masks and solid src pictures, mach64 can do Over in two > * passes, similar to the component-alpha case. > */ > - isSrcSolid = pSrc->pDrawable->width == 1 && > + isSrcSolid = pSrc->pDrawable && > + pSrc->pDrawable->width == 1 && > pSrc->pDrawable->height == 1 && > pSrc->repeat; > > @@ -888,62 +1054,29 @@ uxa_trapezoids (CARD8 op, PicturePtr pSrc, PicturePtr pDst, > PicturePtr pPicture; > INT16 xDst, yDst; > INT16 xRel, yRel; > - int width, height, stride; > - PixmapPtr pPixmap; > - GCPtr pGC; > + int width, height; > pixman_image_t *image; > + pixman_format_code_t format; > > xDst = traps[0].left.p1.x >> 16; > yDst = traps[0].left.p1.y >> 16; > > width = bounds.x2 - bounds.x1; > height = bounds.y2 - bounds.y1; > - stride = (width * BitsPerPixel (maskFormat->depth) + 7) / 8; > > - pPicture = uxa_create_alpha_picture (pScreen, pDst, maskFormat, > - width, height); > - if (!pPicture) > + format = maskFormat->format | (BitsPerPixel (maskFormat->depth) << 24); > + image = pixman_image_create_bits (format, width, height, NULL, 0); > + if (!image) > return; > > - image = pixman_image_create_bits (pPicture->format, > - width, height, > - NULL, stride); > - if (!image) { > - FreePicture (pPicture, 0); > - return; > - } > - > for (; ntrap; ntrap--, traps++) > pixman_rasterize_trapezoid (image, (pixman_trapezoid_t *) traps, > -bounds.x1, -bounds.y1); > > - pPixmap = GetScratchPixmapHeader(pScreen, width, height, > - maskFormat->depth, > - BitsPerPixel (maskFormat->depth), > - PixmapBytePad (width, maskFormat->depth), > - pixman_image_get_data (image)); > - if (!pPixmap) { > - FreePicture (pPicture, 0); > - pixman_image_unref (image); > - return; > - } > - > - pGC = GetScratchGC (pPicture->pDrawable->depth, pScreen); > - if (!pGC) > - { > - FreeScratchPixmapHeader (pPixmap); > - pixman_image_unref (image); > - FreePicture (pPicture, 0); > - return; > - } > - ValidateGC (pPicture->pDrawable, pGC); > - > - (*pGC->ops->CopyArea) (&pPixmap->drawable, pPicture->pDrawable, > - pGC, 0, 0, width, height, 0, 0); > - > - FreeScratchGC (pGC); > - FreeScratchPixmapHeader (pPixmap); > + pPicture = uxa_picture_from_pixman_image (pScreen, image, format); > pixman_image_unref (image); > + if (!pPicture) > + return; > > xRel = bounds.x1 + xSrc - xDst; > yRel = bounds.y1 + ySrc - yDst;
Excerpts from Eric Anholt's message of Fri Sep 04 20:35:29 +0100 2009: > On Fri, 2009-09-04 at 20:26 +0100, Chris Wilson wrote: > > While not yet fully accelerating gradient patterns, by using pixman to > > compute the gradient image and copying that to a pixmap to use as the > > source, we avoid incurring the GPU stall suffered currently from > > reading back the destination surface. > > The destination Picture (the one passed in to our uxa hooks) is supposed > to always have a pDrawable, and if it doesn't, that should have been > caught at a higher level. Hmm, didn't explain the problem well enough (or at all in this case). The issue is that if asked to composite using a gradient source, and/or mask, we fallback to fbComposite, which causes the readback of the destination. All this patch does, like uxa_trapezoids(), is to allocate a scratch pixmap for the source (or mask) using pixman to fill it and then uses CopyArea to transfer that to an offscreen pixmap. -ickle
On Fri, 2009-09-04 at 20:51 +0100, Chris Wilson wrote: > Excerpts from Eric Anholt's message of Fri Sep 04 20:35:29 +0100 2009: > > On Fri, 2009-09-04 at 20:26 +0100, Chris Wilson wrote: > > > While not yet fully accelerating gradient patterns, by using pixman to > > > compute the gradient image and copying that to a pixmap to use as the > > > source, we avoid incurring the GPU stall suffered currently from > > > reading back the destination surface. > > > > The destination Picture (the one passed in to our uxa hooks) is supposed > > to always have a pDrawable, and if it doesn't, that should have been > > caught at a higher level. > > Hmm, didn't explain the problem well enough (or at all in this case). > The issue is that if asked to composite using a gradient source, and/or > mask, we fallback to fbComposite, which causes the readback of the > destination. All this patch does, like uxa_trapezoids(), is to allocate > a scratch pixmap for the source (or mask) using pixman to fill it and then > uses CopyArea to transfer that to an offscreen pixmap. Sorry, I missed that they were in the texture-setup functions. Looks good.
Great work, thanks a lot :) - Clemens
diff --git a/src/i830_render.c b/src/i830_render.c index e4c4623..8418bd7 100644 --- a/src/i830_render.c +++ b/src/i830_render.c @@ -220,11 +220,15 @@ static Bool i830_get_blend_cntl(ScrnInfoPtr pScrn, int op, PicturePtr pMask, static Bool i830_check_composite_texture(PicturePtr pPict, int unit) { - ScrnInfoPtr pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; - int w = pPict->pDrawable->width; - int h = pPict->pDrawable->height; - int i; + ScrnInfoPtr pScrn; + int w, h, i; + if (!pPict->pDrawable) + return TRUE; + + pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; + w = pPict->pDrawable->width; + h = pPict->pDrawable->height; if ((w > 2048) || (h > 2048)) I830FALLBACK("Picture w/h too large (%dx%d)\n", w, h); @@ -393,7 +397,7 @@ i830_prepare_composite(int op, PicturePtr pSrcPicture, PicturePtr pMaskPicture, PicturePtr pDstPicture, PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst) { - ScrnInfoPtr pScrn = xf86Screens[pSrcPicture->pDrawable->pScreen->myNum]; + ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum]; I830Ptr pI830 = I830PTR(pScrn); Bool is_affine_src, is_affine_mask; Bool is_nearest = FALSE; diff --git a/src/i915_render.c b/src/i915_render.c index c81366a..bb755c5 100644 --- a/src/i915_render.c +++ b/src/i915_render.c @@ -169,11 +169,15 @@ static Bool i915_get_dest_format(PicturePtr pDstPicture, uint32_t *dst_format) static Bool i915_check_composite_texture(PicturePtr pPict, int unit) { - ScrnInfoPtr pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; - int w = pPict->pDrawable->width; - int h = pPict->pDrawable->height; - int i; + ScrnInfoPtr pScrn; + int w, h, i; + + if (!pPict->pDrawable) + return TRUE; + pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; + w = pPict->pDrawable->width; + h = pPict->pDrawable->height; if ((w > 2048) || (h > 2048)) I830FALLBACK("Picture w/h too large (%dx%d)\n", w, h); diff --git a/src/i965_render.c b/src/i965_render.c index 1a8075b..1d88af8 100644 --- a/src/i965_render.c +++ b/src/i965_render.c @@ -183,11 +183,15 @@ static Bool i965_get_dest_format(PicturePtr pDstPicture, uint32_t *dst_format) static Bool i965_check_composite_texture(PicturePtr pPict, int unit) { - ScrnInfoPtr pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; - int w = pPict->pDrawable->width; - int h = pPict->pDrawable->height; - int i; + ScrnInfoPtr pScrn; + int w, h, i; + + if (!pPict->pDrawable) + return TRUE; + pScrn = xf86Screens[pPict->pDrawable->pScreen->myNum]; + w = pPict->pDrawable->width; + h = pPict->pDrawable->height; if ((w > 8192) || (h > 8192)) I830FALLBACK("Picture w/h too large (%dx%d)\n", w, h); @@ -1438,7 +1442,7 @@ i965_prepare_composite(int op, PicturePtr pSrcPicture, PicturePtr pMaskPicture, PicturePtr pDstPicture, PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst) { - ScrnInfoPtr pScrn = xf86Screens[pSrcPicture->pDrawable->pScreen->myNum]; + ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum]; I830Ptr pI830 = I830PTR(pScrn); struct gen4_render_state *render_state= pI830->gen4_render_state; gen4_composite_op *composite_op = &render_state->composite_op; diff --git a/uxa/uxa-render.c b/uxa/uxa-render.c index 13128ed..fde10b7 100644 --- a/uxa/uxa-render.c +++ b/uxa/uxa-render.c @@ -314,6 +314,131 @@ uxa_try_driver_solid_fill(PicturePtr pSrc, return 1; } +static PicturePtr +uxa_picture_from_pixman_image (ScreenPtr pScreen, + pixman_image_t *image, + pixman_format_code_t format) +{ + PicturePtr pPicture; + PixmapPtr pPixmap; + GCPtr pGC; + int width, height, depth; + int error; + + width = pixman_image_get_width (image); + height = pixman_image_get_height (image); + depth = pixman_image_get_depth (image); + + pPixmap = (*pScreen->CreatePixmap) (pScreen, width, height, depth, + UXA_CREATE_PIXMAP_FOR_MAP); + if (!pPixmap) + return 0; + + pPicture = CreatePicture (0, &pPixmap->drawable, + PictureMatchFormat (pScreen, depth, format), + 0, 0, serverClient, &error); + (*pScreen->DestroyPixmap) (pPixmap); + + + pPixmap = GetScratchPixmapHeader(pScreen, width, height, depth, + BitsPerPixel (depth), + pixman_image_get_stride (image), + pixman_image_get_data (image)); + if (!pPixmap) + { + FreePicture (pPicture, 0); + return 0; + } + + pGC = GetScratchGC (depth, pScreen); + if (!pGC) + { + FreeScratchPixmapHeader (pPixmap); + FreePicture (pPicture, 0); + return 0; + } + ValidateGC (pPicture->pDrawable, pGC); + + (*pGC->ops->CopyArea) (&pPixmap->drawable, pPicture->pDrawable, + pGC, 0, 0, width, height, 0, 0); + + FreeScratchGC (pGC); + FreeScratchPixmapHeader (pPixmap); + + return pPicture; +} + +static PicturePtr +uxa_get_pattern (ScreenPtr pScreen, + PicturePtr pPict, + pixman_format_code_t format, + INT16 x, INT16 y, + CARD16 width, CARD16 height) +{ + pixman_image_t *source, *image; + + source = image_from_pict (pPict, 0, 0); + if (!source) + return 0; + + image = pixman_image_create_bits (format, width, height, NULL, 0); + if (!image) { + pixman_image_unref (source); + return 0; + } + + pixman_image_composite (PIXMAN_OP_SRC, + source, NULL, image, + x, y, + 0, 0, + 0, 0, + width, height); + pixman_image_unref (source); + + pPict = uxa_picture_from_pixman_image (pScreen, image, format); + pixman_image_unref (image); + + return pPict; +} + +static PicturePtr +uxa_get_source (ScreenPtr pScreen, + PicturePtr pPict, + INT16 x, INT16 y, + CARD16 width, CARD16 height, + INT16 *out_x, INT16 *out_y) +{ + if (pPict->pDrawable) { + *out_x = x + pPict->pDrawable->x; + *out_y = y + pPict->pDrawable->y; + return pPict; + } + + *out_x = 0; + *out_y = 0; + return uxa_get_pattern (pScreen, pPict, + PICT_a8r8g8b8, x, y, width, height); +} + +static PicturePtr +uxa_get_mask (ScreenPtr pScreen, + PicturePtr pPict, + INT16 x, INT16 y, + INT16 width, INT16 height, + INT16 *out_x, INT16 *out_y) +{ + if (pPict->pDrawable) { + *out_x = x + pPict->pDrawable->x; + *out_y = y + pPict->pDrawable->y; + return pPict; + } + + *out_x = 0; + *out_y = 0; + return uxa_get_pattern (pScreen, pPict, + PICT_a8, x, y, width, height); +} + static int uxa_try_driver_composite_rects(CARD8 op, PicturePtr pSrc, @@ -333,7 +458,7 @@ uxa_try_driver_composite_rects(CARD8 op, { return -1; } - + pDstPix = uxa_get_offscreen_pixmap(pDst->pDrawable, &dst_off_x, &dst_off_y); if (!pDstPix) return 0; @@ -453,48 +578,84 @@ uxa_try_driver_composite(CARD8 op, int nbox; int src_off_x, src_off_y, mask_off_x, mask_off_y, dst_off_x, dst_off_y; PixmapPtr pSrcPix, pMaskPix = NULL, pDstPix; + PicturePtr localSrc, localMask = NULL; xDst += pDst->pDrawable->x; yDst += pDst->pDrawable->y; + localSrc = uxa_get_source (pDst->pDrawable->pScreen, + pSrc, xSrc, ySrc, width, height, + &xSrc, &ySrc); + if (! localSrc) + return 0; + if (pMask) { - xMask += pMask->pDrawable->x; - yMask += pMask->pDrawable->y; - } + localMask = uxa_get_mask (pDst->pDrawable->pScreen, + pMask, xMask, yMask, width, height, + &xMask, &yMask); + if (! localMask) { + if (localSrc != pSrc) + FreePicture (localSrc, 0); - xSrc += pSrc->pDrawable->x; - ySrc += pSrc->pDrawable->y; + return 0; + } + } if (uxa_screen->info->check_composite && - !(*uxa_screen->info->check_composite) (op, pSrc, pMask, pDst)) + !(*uxa_screen->info->check_composite) (op, localSrc, localMask, pDst)) { + if (localSrc != pSrc) + FreePicture (localSrc, 0); + if (localMask && localMask != pMask) + FreePicture (localMask, 0); + return -1; } - if (!miComputeCompositeRegion (®ion, pSrc, pMask, pDst, + if (!miComputeCompositeRegion (®ion, localSrc, localMask, pDst, xSrc, ySrc, xMask, yMask, xDst, yDst, width, height)) + { + if (localSrc != pSrc) + FreePicture (localSrc, 0); + if (localMask && localMask != pMask) + FreePicture (localMask, 0); + return 1; + } pDstPix = uxa_get_offscreen_pixmap (pDst->pDrawable, &dst_off_x, &dst_off_y); - pSrcPix = uxa_get_offscreen_pixmap (pSrc->pDrawable, &src_off_x, &src_off_y); + pSrcPix = uxa_get_offscreen_pixmap (localSrc->pDrawable, + &src_off_x, &src_off_y); - if (pMask) - pMaskPix = uxa_get_offscreen_pixmap (pMask->pDrawable, &mask_off_x, - &mask_off_y); + if (localMask) + pMaskPix = uxa_get_offscreen_pixmap (localMask->pDrawable, + &mask_off_x, &mask_off_y); - if (!pDstPix || !pSrcPix || (pMask && !pMaskPix)) { + if (!pDstPix || !pSrcPix || (localMask && !pMaskPix)) { REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); + + if (localSrc != pSrc) + FreePicture (localSrc, 0); + if (localMask && localMask != pMask) + FreePicture (localMask, 0); + return 0; } REGION_TRANSLATE(pScreen, ®ion, dst_off_x, dst_off_y); - if (!(*uxa_screen->info->prepare_composite) (op, pSrc, pMask, pDst, pSrcPix, - pMaskPix, pDstPix)) + if (!(*uxa_screen->info->prepare_composite) (op, localSrc, localMask, pDst, + pSrcPix, pMaskPix, pDstPix)) { REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); + + if (localSrc != pSrc) + FreePicture (localSrc, 0); + if (localMask && localMask != pMask) + FreePicture (localMask, 0); + return -1; } @@ -523,6 +684,12 @@ uxa_try_driver_composite(CARD8 op, (*uxa_screen->info->done_composite) (pDstPix); REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); + + if (localSrc != pSrc) + FreePicture (localSrc, 0); + if (localMask && localMask != pMask) + FreePicture (localMask, 0); + return 1; } @@ -636,14 +803,11 @@ uxa_composite(CARD8 op, Bool saveMaskRepeat = pMask ? pMask->repeat : 0; RegionRec region; - /* We currently don't support acceleration of gradients, or other pictures - * with a NULL pDrawable. - */ - if (uxa_screen->swappedOut || - pSrc->pDrawable == NULL || (pMask != NULL && pMask->pDrawable == NULL)) - { + if (uxa_screen->swappedOut) goto fallback; - } + + if (pSrc->pDrawable == NULL || (pMask && pMask->pDrawable == NULL)) + goto composite; /* Remove repeat in source if useless */ if (pSrc->repeat && !pSrc->transform && xSrc >= 0 && @@ -744,6 +908,7 @@ uxa_composite(CARD8 op, (yMask + height) <= pMask->pDrawable->height) pMask->repeat = 0; +composite: if (uxa_screen->info->prepare_composite && !pSrc->alphaMap && (!pMask || !pMask->alphaMap) && !pDst->alphaMap) { @@ -757,7 +922,8 @@ uxa_composite(CARD8 op, /* For generic masks and solid src pictures, mach64 can do Over in two * passes, similar to the component-alpha case. */ - isSrcSolid = pSrc->pDrawable->width == 1 && + isSrcSolid = pSrc->pDrawable && + pSrc->pDrawable->width == 1 && pSrc->pDrawable->height == 1 && pSrc->repeat; @@ -888,62 +1054,29 @@ uxa_trapezoids (CARD8 op, PicturePtr pSrc, PicturePtr pDst, PicturePtr pPicture; INT16 xDst, yDst; INT16 xRel, yRel; - int width, height, stride; - PixmapPtr pPixmap; - GCPtr pGC; + int width, height; pixman_image_t *image; + pixman_format_code_t format; xDst = traps[0].left.p1.x >> 16; yDst = traps[0].left.p1.y >> 16; width = bounds.x2 - bounds.x1; height = bounds.y2 - bounds.y1; - stride = (width * BitsPerPixel (maskFormat->depth) + 7) / 8; - pPicture = uxa_create_alpha_picture (pScreen, pDst, maskFormat, - width, height); - if (!pPicture) + format = maskFormat->format | (BitsPerPixel (maskFormat->depth) << 24); + image = pixman_image_create_bits (format, width, height, NULL, 0); + if (!image) return; - image = pixman_image_create_bits (pPicture->format, - width, height, - NULL, stride); - if (!image) { - FreePicture (pPicture, 0); - return; - } - for (; ntrap; ntrap--, traps++) pixman_rasterize_trapezoid (image, (pixman_trapezoid_t *) traps, -bounds.x1, -bounds.y1); - pPixmap = GetScratchPixmapHeader(pScreen, width, height, - maskFormat->depth, - BitsPerPixel (maskFormat->depth), - PixmapBytePad (width, maskFormat->depth), - pixman_image_get_data (image)); - if (!pPixmap) { - FreePicture (pPicture, 0); - pixman_image_unref (image); - return; - } - - pGC = GetScratchGC (pPicture->pDrawable->depth, pScreen); - if (!pGC) - { - FreeScratchPixmapHeader (pPixmap); - pixman_image_unref (image); - FreePicture (pPicture, 0); - return; - } - ValidateGC (pPicture->pDrawable, pGC); - - (*pGC->ops->CopyArea) (&pPixmap->drawable, pPicture->pDrawable, - pGC, 0, 0, width, height, 0, 0); - - FreeScratchGC (pGC); - FreeScratchPixmapHeader (pPixmap); + pPicture = uxa_picture_from_pixman_image (pScreen, image, format); pixman_image_unref (image); + if (!pPicture) + return; xRel = bounds.x1 + xSrc - xDst; yRel = bounds.y1 + ySrc - yDst;
While not yet fully accelerating gradient patterns, by using pixman to compute the gradient image and copying that to a pixmap to use as the source, we avoid incurring the GPU stall suffered currently from reading back the destination surface. Speedups on i915: firefox-talos-svg: 710378.14 -> 549262.96: 1.29x speedup No slowdowns. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> --- src/i830_render.c | 14 ++- src/i915_render.c | 12 ++- src/i965_render.c | 14 ++- uxa/uxa-render.c | 261 ++++++++++++++++++++++++++++++++++++++++------------- 4 files changed, 223 insertions(+), 78 deletions(-)