xf86-video-intel: 4 commits - src/sna/gen2_render.c src/sna/gen3_render.c src/sna/kgem.h src/sna/sna_accel.c src/sna/sna_reg.h
Chris Wilson
ickle at kemper.freedesktop.org
Tue Oct 4 11:37:35 PDT 2011
src/sna/gen2_render.c | 2
src/sna/gen3_render.c | 2
src/sna/kgem.h | 2
src/sna/sna_accel.c | 248 +++++++++++++++++++++++++++++++++++++++++++++++---
src/sna/sna_reg.h | 3
5 files changed, 243 insertions(+), 14 deletions(-)
New commits:
commit c5414ec992d935e10156a2b513d5ec2dded2f689
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Sun Oct 2 12:02:41 2011 +0100
sna: Use BLT operations to avoid fallbacks in core glyph rendering
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c
index 5a94fbd..1edd3a3 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -30,6 +30,7 @@
#endif
#include "sna.h"
+#include "sna_reg.h"
#include <X11/fonts/font.h>
#include <X11/fonts/fontstruct.h>
@@ -1689,9 +1690,13 @@ sna_poly_fill_rect(DrawablePtr draw, GCPtr gc, int n, xRectangle *rect);
static bool
can_fill_spans(DrawablePtr drawable, GCPtr gc)
{
+ DBG(("%s: is-solid-mask? %d\n", __FUNCTION__,
+ PM_IS_SOLID(drawable, gc->planemask)));
if (!PM_IS_SOLID(drawable, gc->planemask))
return false;
+ DBG(("%s: non-stipple fill? %d\n", __FUNCTION__,
+ gc->fillStyle != FillStippled));
return gc->fillStyle == FillSolid || gc->fillStyle == FillTiled;
}
@@ -3045,11 +3050,198 @@ fallback:
fbPolyFillRect(draw, gc, n, rect);
}
+static uint8_t byte_reverse(uint8_t b)
+{
+ return ((b * 0x80200802ULL) & 0x0884422110ULL) * 0x0101010101ULL >> 32;
+}
+
+static uint8_t blt_depth(int depth)
+{
+ switch (depth) {
+ case 8: return 0;
+ case 15: return 0x2;
+ case 16: return 0x1;
+ default: return 0x3;
+ }
+}
+
+static bool
+sna_glyph_blt(DrawablePtr drawable, GCPtr gc,
+ int x, int y, unsigned int n,
+ CharInfoPtr *info, pointer base,
+ bool transparent,
+ const BoxRec *extents)
+{
+ struct sna *sna = to_sna_from_drawable(drawable);
+ PixmapPtr pixmap = get_drawable_pixmap(drawable);
+ struct sna_pixmap *priv = sna_pixmap(pixmap);
+ struct sna_damage **damage;
+ uint32_t *b;
+ int16_t dx, dy;
+
+ /* XXX sna_blt! */
+ static const uint8_t ROP[] = {
+ ROP_0, /* GXclear */
+ ROP_DSa, /* GXand */
+ ROP_SDna, /* GXandReverse */
+ ROP_S, /* GXcopy */
+ ROP_DSna, /* GXandInverted */
+ ROP_D, /* GXnoop */
+ ROP_DSx, /* GXxor */
+ ROP_DSo, /* GXor */
+ ROP_DSon, /* GXnor */
+ ROP_DSxn, /* GXequiv */
+ ROP_Dn, /* GXinvert */
+ ROP_SDno, /* GXorReverse */
+ ROP_Sn, /* GXcopyInverted */
+ ROP_DSno, /* GXorInverted */
+ ROP_DSan, /* GXnand */
+ ROP_1 /* GXset */
+ };
+ uint8_t rop = transparent ? ROP[gc->alu] : ROP_S;
+ RegionRec clip;
+
+ if (priv->gpu_bo->tiling == I915_TILING_Y)
+ return false;
+
+ RegionInit(&clip, (BoxPtr)extents, 1);
+ if (gc->pCompositeClip && gc->pCompositeClip->data)
+ RegionIntersect(&clip, &clip, gc->pCompositeClip);
+
+ /* XXX loop over clips using SETUP_CLIP? */
+ if (REGION_NUM_RECTS(&clip) != 1) {
+ RegionUninit(&clip);
+ return false;
+ }
+
+ damage = priv->gpu_only ? NULL :
+ reduce_damage(drawable, &priv->gpu_damage, extents),
+
+ get_drawable_deltas(drawable, pixmap, &dx, &dy);
+ x += drawable->x + dx;
+ y += drawable->y + dy;
+
+ kgem_set_mode(&sna->kgem, KGEM_BLT);
+ if (!kgem_check_batch(&sna->kgem, 16) ||
+ !kgem_check_bo_fenced(&sna->kgem, priv->gpu_bo, NULL)) {
+ _kgem_submit(&sna->kgem);
+ _kgem_set_mode(&sna->kgem, KGEM_BLT);
+ }
+ b = sna->kgem.batch + sna->kgem.nbatch;
+ b[0] = XY_SETUP_BLT | 1 << 20;
+ b[1] = priv->gpu_bo->pitch;
+ if (sna->kgem.gen >= 40) {
+ if (priv->gpu_bo->tiling)
+ b[0] |= 1 << 11;
+ b[1] >>= 2;
+ }
+ b[1] |= 1 << 30 | transparent << 29 | blt_depth(drawable->depth) << 24 | rop << 16;
+ b[2] = (dy + clip.extents.y1) << 16 | (dx + clip.extents.x1);
+ b[3] = (dy + clip.extents.y2) << 16 | (dx + clip.extents.x2);
+ b[4] = kgem_add_reloc(&sna->kgem, sna->kgem.nbatch + 4,
+ priv->gpu_bo,
+ I915_GEM_DOMAIN_RENDER << 16 |
+ I915_GEM_DOMAIN_RENDER |
+ KGEM_RELOC_FENCED,
+ 0);
+ b[5] = gc->bgPixel;
+ b[6] = gc->fgPixel;
+ b[7] = 0;
+ sna->kgem.nbatch += 8;
+
+ while (n--) {
+ CharInfoPtr c = *info++;
+ uint8_t *glyph = FONTGLYPHBITS(base, c);
+ int w = GLYPHWIDTHPIXELS(c);
+ int h = GLYPHHEIGHTPIXELS(c);
+ int stride = GLYPHWIDTHBYTESPADDED(c);
+ int w8 = (w + 7) >> 3;
+ int x1, y1, len, i;
+ uint8_t *byte;
+
+ if (w == 0 || h == 0)
+ goto skip;
+
+ len = (w8 * h + 7) >> 3 << 1;
+ DBG(("%s glyph: (%d, %d) x (%d[%d], %d), len=%d\n" ,__FUNCTION__,
+ x,y, w, w8, h, len));
+
+ x1 = x + c->metrics.leftSideBearing;
+ y1 = y - c->metrics.ascent;
+
+ if (!kgem_check_batch(&sna->kgem, 3+len)) {
+ _kgem_submit(&sna->kgem);
+ _kgem_set_mode(&sna->kgem, KGEM_BLT);
+
+ b = sna->kgem.batch + sna->kgem.nbatch;
+ b[0] = XY_SETUP_BLT | 1 << 20;
+ b[1] = priv->gpu_bo->pitch;
+ if (sna->kgem.gen >= 40) {
+ if (priv->gpu_bo->tiling)
+ b[0] |= 1 << 11;
+ b[1] >>= 2;
+ }
+ b[1] |= 1 << 30 | transparent << 29 | blt_depth(drawable->depth) << 24 | rop << 16;
+ b[2] = (dy + clip.extents.y1) << 16 | (dx + clip.extents.x1);
+ b[3] = (dy + clip.extents.y2) << 16 | (dx + clip.extents.x2);
+ b[4] = kgem_add_reloc(&sna->kgem, sna->kgem.nbatch + 4,
+ priv->gpu_bo,
+ I915_GEM_DOMAIN_RENDER << 16 |
+ I915_GEM_DOMAIN_RENDER |
+ KGEM_RELOC_FENCED,
+ 0);
+ b[5] = gc->bgPixel;
+ b[6] = gc->fgPixel;
+ b[7] = 0;
+ sna->kgem.nbatch += 8;
+ }
+
+ b = sna->kgem.batch + sna->kgem.nbatch;
+ b[0] = XY_TEXT_IMMEDIATE_BLT | (1 + len);
+ if (priv->gpu_bo->tiling && sna->kgem.gen >= 40)
+ b[0] |= 1 << 11;
+ b[1] = (uint16_t)y1 << 16 | (uint16_t)x1;
+ b[2] = (uint16_t)(y1+h) << 16 | (uint16_t)(x1+w);
+
+ byte = (uint8_t *)&b[3];
+ stride -= w8;
+ do {
+ i = w8;
+ do {
+ *byte++ = byte_reverse(*glyph++);
+ } while (--i);
+ glyph += stride;
+ } while (--h);
+ while ((byte - (uint8_t *)&b[3]) & 7)
+ *byte++ = 0;
+ sna->kgem.nbatch += 3 + len;
+ assert((uint32_t *)byte == sna->kgem.batch + sna->kgem.nbatch);
+
+ if (damage) {
+ BoxRec r;
+
+ r.x1 = x1;
+ r.y1 = y1;
+ r.x2 = x1 + w;
+ r.y2 = y1 + h;
+ if (box_intersect(&r, extents))
+ sna_damage_add_box(damage, &r);
+ }
+skip:
+ x += c->metrics.characterWidth;
+ }
+
+ RegionUninit(&clip);
+ sna->blt_state.fill_bo = 0;
+ return true;
+}
+
static void
sna_image_glyph(DrawablePtr drawable, GCPtr gc,
int x, int y, unsigned int n,
CharInfoPtr *info, pointer base)
{
+ struct sna *sna = to_sna_from_drawable(drawable);
ExtentInfoRec extents;
BoxRec box;
RegionRec region;
@@ -3058,15 +3250,11 @@ sna_image_glyph(DrawablePtr drawable, GCPtr gc,
return;
QueryGlyphExtents(gc->font, info, n, &extents);
- if (extents.overallWidth >= 0) {
- box.x1 = x;
- box.x2 = x + extents.overallWidth;
- } else {
- box.x2 = x;
- box.x1 = x + extents.overallWidth;
- }
- box.y1 = y - FONTASCENT(gc->font);
- box.y2 = y + FONTDESCENT(gc->font);
+ box.x1 = x + extents.overallLeft;
+ box.y1 = y - extents.overallAscent;
+ box.x2 = x + extents.overallRight;
+ box.y2 = y + extents.overallDescent;
+
trim_box(&box, drawable);
if (box_empty(&box))
return;
@@ -3075,6 +3263,19 @@ sna_image_glyph(DrawablePtr drawable, GCPtr gc,
DBG(("%s: extents(%d, %d), (%d, %d)\n",
__FUNCTION__, box.x1, box.y1, box.x2, box.y2));
+ if (FORCE_FALLBACK)
+ goto fallback;
+
+ if (sna->kgem.wedged) {
+ DBG(("%s: fallback -- wedged\n", __FUNCTION__));
+ goto fallback;
+ }
+
+ if (sna_drawable_use_gpu_bo(drawable, &box) &&
+ sna_glyph_blt(drawable, gc, x, y, n, info, base, false, &box))
+ return;
+
+fallback:
RegionInit(®ion, &box, 1);
if (gc->pCompositeClip)
RegionIntersect(®ion, ®ion, gc->pCompositeClip);
@@ -3085,6 +3286,7 @@ sna_image_glyph(DrawablePtr drawable, GCPtr gc,
sna_drawable_move_region_to_cpu(drawable, ®ion, true);
RegionUninit(®ion);
+ DBG(("%s: fallback -- fbImageGlyphBlt\n", __FUNCTION__));
fbImageGlyphBlt(drawable, gc, x, y, n, info, base);
}
@@ -3093,6 +3295,7 @@ sna_poly_glyph(DrawablePtr drawable, GCPtr gc,
int x, int y, unsigned int n,
CharInfoPtr *info, pointer base)
{
+ struct sna *sna = to_sna_from_drawable(drawable);
ExtentInfoRec extents;
BoxRec box;
RegionRec region;
@@ -3114,6 +3317,20 @@ sna_poly_glyph(DrawablePtr drawable, GCPtr gc,
DBG(("%s: extents(%d, %d), (%d, %d)\n",
__FUNCTION__, box.x1, box.y1, box.x2, box.y2));
+ if (FORCE_FALLBACK)
+ goto fallback;
+
+ if (sna->kgem.wedged) {
+ DBG(("%s: fallback -- wedged\n", __FUNCTION__));
+ goto fallback;
+ }
+
+ if (sna_drawable_use_gpu_bo(drawable, &box) &&
+ sna_glyph_blt(drawable, gc, x, y, n, info, base, true, &box))
+
+ return;
+
+fallback:
RegionInit(®ion, &box, 1);
if (gc->pCompositeClip)
RegionIntersect(®ion, ®ion, gc->pCompositeClip);
@@ -3124,6 +3341,7 @@ sna_poly_glyph(DrawablePtr drawable, GCPtr gc,
sna_drawable_move_region_to_cpu(drawable, ®ion, true);
RegionUninit(®ion);
+ DBG(("%s: fallback -- fbPolyGlyphBlt\n", __FUNCTION__));
fbPolyGlyphBlt(drawable, gc, x, y, n, info, base);
}
diff --git a/src/sna/sna_reg.h b/src/sna/sna_reg.h
index f16c64d..8ca9089 100644
--- a/src/sna/sna_reg.h
+++ b/src/sna/sna_reg.h
@@ -78,10 +78,11 @@
#define COLOR_BLT_CMD ((2<<29)|(0x40<<22)|(0x3))
#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|(0x4))
-#define XY_SETUP_BLT_CMD ((2<<29)|(1<<22)|6)
+#define XY_SETUP_BLT ((2<<29)|(1<<22)|6)
#define XY_SETUP_MONO_PATTERN_SL_BLT ((2<<29)|(0x11<<22)|7)
#define XY_SETUP_CLIP_BLT_CMD ((2<<29)|(3<<22)|1)
#define XY_SCANLINE_BLT ((2<<29)|(0x25<<22)|1)
+#define XY_TEXT_IMMEDIATE_BLT ((2<<29)|(0x31<<22)|(1<<16))
#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6)
#define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|0x4)
#define XY_PAT_BLT_IMMEDIATE ((2<<29)|(0x72<<22))
commit 6b62b9d7c4301457a024f840383fb4515704ebdb
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Tue Oct 4 19:30:06 2011 +0100
sna: Increase reserved space in batch to accommodate gen5 workaround
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/kgem.h b/src/sna/kgem.h
index af6203c..49ca3d7 100644
--- a/src/sna/kgem.h
+++ b/src/sna/kgem.h
@@ -120,7 +120,7 @@ struct kgem {
struct drm_i915_gem_relocation_entry reloc[384];
};
-#define KGEM_BATCH_RESERVED 2
+#define KGEM_BATCH_RESERVED 4 /* need a bit of extra room for workarounds */
#define KGEM_RELOC_RESERVED 4
#define KGEM_EXEC_RESERVED 1
commit 6351d8eb823dbfffb4792af8592d1ce565211e23
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Tue Oct 4 19:29:46 2011 +0100
sna/gen[23]: Fix compilation with debugging enabled
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/gen2_render.c b/src/sna/gen2_render.c
index 43aaef0..935dffe 100644
--- a/src/sna/gen2_render.c
+++ b/src/sna/gen2_render.c
@@ -2072,7 +2072,7 @@ gen2_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
gen2_emit_fill_state(sna, &tmp);
gen2_get_rectangles(sna, &tmp, 1);
- DBG((" (%d, %d), (%d, %d): %x\n", x1, y1, x2, y2, pixel));
+ DBG((" (%d, %d), (%d, %d): %x\n", x1, y1, x2, y2, color));
OUT_VERTEX(x2);
OUT_VERTEX(y2);
OUT_VERTEX(x1);
diff --git a/src/sna/gen3_render.c b/src/sna/gen3_render.c
index 65176ed..57c57b9 100644
--- a/src/sna/gen3_render.c
+++ b/src/sna/gen3_render.c
@@ -3731,7 +3731,7 @@ gen3_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
gen3_emit_composite_state(sna, &tmp);
gen3_align_vertex(sna, &tmp);
gen3_get_rectangles(sna, &tmp, 1);
- DBG((" (%d, %d), (%d, %d): %x\n", x1, y1, x2, y2, pixel));
+ DBG((" (%d, %d), (%d, %d): %x\n", x1, y1, x2, y2, color));
OUT_VERTEX(x2);
OUT_VERTEX(y2);
OUT_VERTEX(x1);
commit 1ec6a0e27753027a68beeaff0276ae65c658d468
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Mon Oct 3 16:14:18 2011 +0100
sna: Move the source to the GPU if it is reused
We attempt to skip upload a source pixmap to the GPU in the event it is
used only once (for example during image upload by firefox). However, if
we continue to use the CPU source pixmap then it obviously was worth
uploading to the GPU. So if we use the CPU pixmap a second time, do the
upload and then blit.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c
index a6e639b..5a94fbd 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -1142,6 +1142,15 @@ sna_put_image(DrawablePtr drawable, GCPtr gc, int depth,
RegionUninit(®ion);
}
+static Bool
+move_to_gpu(PixmapPtr pixmap, struct sna_pixmap *priv, const BoxRec *box)
+{
+ int w = box->x2 - box->x1;
+ int h = box->y2 - box->y1;
+
+ return ++priv->source_count * w*h >= 2 * pixmap->drawable.width * pixmap->drawable.height;
+}
+
static void
sna_copy_boxes(DrawablePtr src, DrawablePtr dst, GCPtr gc,
BoxPtr box, int n,
@@ -1233,7 +1242,8 @@ sna_copy_boxes(DrawablePtr src, DrawablePtr dst, GCPtr gc,
goto fallback;
}
- if (src_priv && src_priv->gpu_bo &&
+ if (src_priv &&
+ move_to_gpu(src_pixmap, src_priv, ®ion.extents) &&
sna_pixmap_move_to_gpu(src_pixmap)) {
if (!sna->render.copy_boxes(sna, alu,
src_pixmap, src_priv->gpu_bo, src_dx, src_dy,
More information about the xorg-commit
mailing list