xf86-video-intel: 2 commits - src/sna/gen2_render.c src/sna/gen3_render.c src/sna/gen4_render.c src/sna/gen5_render.c src/sna/gen6_render.c src/sna/gen7_render.c src/sna/sna_accel.c src/sna/sna_render.c src/sna/sna_render.h src/sna/sna_trapezoids.c
Chris Wilson
ickle at kemper.freedesktop.org
Thu Jul 19 03:55:51 PDT 2012
src/sna/gen2_render.c | 46 +++---
src/sna/gen3_render.c | 45 +++---
src/sna/gen4_render.c | 324 +++++++++++++++++++++++++++++++++++++++++++++++
src/sna/gen5_render.c | 42 +++---
src/sna/gen6_render.c | 42 +++---
src/sna/gen7_render.c | 36 +++--
src/sna/sna_accel.c | 10 -
src/sna/sna_render.c | 51 ++++---
src/sna/sna_render.h | 3
src/sna/sna_trapezoids.c | 68 ++++++---
10 files changed, 540 insertions(+), 127 deletions(-)
New commits:
commit e4fce3b7801038e4f64d848a0995f4b441b4d2aa
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Thu Jul 19 10:50:09 2012 +0100
sna/gen4: Hookup composite spans
Due to the unresolved flushing bug it is no faster (so only enable when
we definitely can't do the operation inplace), however it does eliminate
a chunk of CPU overhead.
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 622ba1f..04c351c 100644
--- a/src/sna/gen2_render.c
+++ b/src/sna/gen2_render.c
@@ -2192,6 +2192,28 @@ gen2_render_composite_spans_done(struct sna *sna,
}
static bool
+gen2_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height, unsigned flags)
+{
+ if (op >= ARRAY_SIZE(gen2_blend_op))
+ return false;
+
+ if (gen2_composite_fallback(sna, src, NULL, dst))
+ return false;
+
+ if (need_tiling(sna, width, height)) {
+ if (!is_gpu(dst->pDrawable)) {
+ DBG(("%s: fallback, tiled operation not on GPU\n",
+ __FUNCTION__));
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static bool
gen2_render_composite_spans(struct sna *sna,
uint8_t op,
PicturePtr src,
@@ -2205,29 +2227,10 @@ gen2_render_composite_spans(struct sna *sna,
DBG(("%s(src=(%d, %d), dst=(%d, %d), size=(%d, %d))\n", __FUNCTION__,
src_x, src_y, dst_x, dst_y, width, height));
-#if NO_COMPOSITE_SPANS
- return false;
-#endif
-
- if (op >= ARRAY_SIZE(gen2_blend_op)) {
- DBG(("%s: fallback due to unhandled blend op: %d\n",
- __FUNCTION__, op));
- return false;
- }
-
- if (gen2_composite_fallback(sna, src, NULL, dst))
- return false;
-
+ assert(gen2_check_composite_spans(sna, op, src, dst, width, height, flags));
if (need_tiling(sna, width, height)) {
DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
__FUNCTION__, width, height));
-
- if (!is_gpu(dst->pDrawable)) {
- DBG(("%s: fallback, tiled operation not on GPU\n",
- __FUNCTION__));
- return false;
- }
-
return sna_tiling_composite_spans(op, src, dst,
src_x, src_y, dst_x, dst_y,
width, height, flags, tmp);
@@ -3134,7 +3137,10 @@ bool gen2_render_init(struct sna *sna)
* use the texture combiners.
*/
render->composite = gen2_render_composite;
+#if !NO_COMPOSITE_SPANS
+ render->check_composite_spans = gen2_check_composite_spans;
render->composite_spans = gen2_render_composite_spans;
+#endif
render->fill_boxes = gen2_render_fill_boxes;
render->fill = gen2_render_fill;
render->fill_one = gen2_render_fill_one;
diff --git a/src/sna/gen3_render.c b/src/sna/gen3_render.c
index e02eb89..1f6c1aa 100644
--- a/src/sna/gen3_render.c
+++ b/src/sna/gen3_render.c
@@ -3368,6 +3368,28 @@ gen3_render_composite_spans_done(struct sna *sna,
}
static bool
+gen3_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height, unsigned flags)
+{
+ if (op >= ARRAY_SIZE(gen3_blend_op))
+ return false;
+
+ if (gen3_composite_fallback(sna, op, src, NULL, dst))
+ return false;
+
+ if (need_tiling(sna, width, height)) {
+ if (!is_gpu(dst->pDrawable)) {
+ DBG(("%s: fallback, tiled operation not on GPU\n",
+ __FUNCTION__));
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static bool
gen3_render_composite_spans(struct sna *sna,
uint8_t op,
PicturePtr src,
@@ -3383,29 +3405,11 @@ gen3_render_composite_spans(struct sna *sna,
DBG(("%s(src=(%d, %d), dst=(%d, %d), size=(%d, %d))\n", __FUNCTION__,
src_x, src_y, dst_x, dst_y, width, height));
-#if NO_COMPOSITE_SPANS
- return false;
-#endif
-
- if (op >= ARRAY_SIZE(gen3_blend_op)) {
- DBG(("%s: fallback due to unhandled blend op: %d\n",
- __FUNCTION__, op));
- return false;
- }
-
- if (gen3_composite_fallback(sna, op, src, NULL, dst))
- return false;
+ assert(gen3_check_composite_spans(sna, op, src, dst, width, height, flags));
if (need_tiling(sna, width, height)) {
DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
__FUNCTION__, width, height));
-
- if (!is_gpu(dst->pDrawable)) {
- DBG(("%s: fallback, tiled operation not on GPU\n",
- __FUNCTION__));
- return false;
- }
-
return sna_tiling_composite_spans(op, src, dst,
src_x, src_y, dst_x, dst_y,
width, height, flags, tmp);
@@ -4666,7 +4670,10 @@ bool gen3_render_init(struct sna *sna)
struct sna_render *render = &sna->render;
render->composite = gen3_render_composite;
+#if !NO_COMPOSITE_SPANS
+ render->check_composite_spans = gen3_check_composite_spans;
render->composite_spans = gen3_render_composite_spans;
+#endif
render->video = gen3_render_video;
diff --git a/src/sna/gen4_render.c b/src/sna/gen4_render.c
index 4a67728..6fcce71 100644
--- a/src/sna/gen4_render.c
+++ b/src/sna/gen4_render.c
@@ -50,6 +50,7 @@
#define FLUSH_EVERY_VERTEX 1
#define NO_COMPOSITE 0
+#define NO_COMPOSITE_SPANS 0
#define NO_COPY 0
#define NO_COPY_BOXES 0
#define NO_FILL 0
@@ -61,8 +62,13 @@
gen4_magic_ca_pass(sna, OP); \
OUT_BATCH(MI_FLUSH | MI_INHIBIT_RENDER_CACHE_FLUSH); \
} while (0)
+#define FLUSH_NOCA() do { \
+ gen4_vertex_flush(sna); \
+ OUT_BATCH(MI_FLUSH | MI_INHIBIT_RENDER_CACHE_FLUSH); \
+} while (0)
#else
#define FLUSH(OP)
+#define FLUSH_NOCA()
#endif
#define GEN4_GRF_BLOCKS(nreg) ((nreg + 15) / 16 - 1)
@@ -2494,6 +2500,320 @@ cleanup_dst:
return false;
}
+/* A poor man's span interface. But better than nothing? */
+#if !NO_COMPOSITE_SPANS
+static bool
+gen4_composite_alpha_gradient_init(struct sna *sna,
+ struct sna_composite_channel *channel)
+{
+ DBG(("%s\n", __FUNCTION__));
+
+ channel->filter = PictFilterNearest;
+ channel->repeat = RepeatPad;
+ channel->is_affine = true;
+ channel->is_solid = false;
+ channel->transform = NULL;
+ channel->width = 256;
+ channel->height = 1;
+ channel->card_format = GEN4_SURFACEFORMAT_B8G8R8A8_UNORM;
+
+ channel->bo = sna_render_get_alpha_gradient(sna);
+
+ channel->scale[0] = channel->scale[1] = 1;
+ channel->offset[0] = channel->offset[1] = 0;
+ return channel->bo != NULL;
+}
+
+inline static void
+gen4_emit_composite_texcoord(struct sna *sna,
+ const struct sna_composite_channel *channel,
+ int16_t x, int16_t y)
+{
+ float t[3];
+
+ if (channel->is_affine) {
+ sna_get_transformed_coordinates(x + channel->offset[0],
+ y + channel->offset[1],
+ channel->transform,
+ &t[0], &t[1]);
+ OUT_VERTEX_F(t[0] * channel->scale[0]);
+ OUT_VERTEX_F(t[1] * channel->scale[1]);
+ } else {
+ t[0] = t[1] = 0; t[2] = 1;
+ sna_get_transformed_coordinates_3d(x + channel->offset[0],
+ y + channel->offset[1],
+ channel->transform,
+ &t[0], &t[1], &t[2]);
+ OUT_VERTEX_F(t[0] * channel->scale[0]);
+ OUT_VERTEX_F(t[1] * channel->scale[1]);
+ OUT_VERTEX_F(t[2]);
+ }
+}
+
+inline static void
+gen4_emit_composite_texcoord_affine(struct sna *sna,
+ const struct sna_composite_channel *channel,
+ int16_t x, int16_t y)
+{
+ float t[2];
+
+ sna_get_transformed_coordinates(x + channel->offset[0],
+ y + channel->offset[1],
+ channel->transform,
+ &t[0], &t[1]);
+ OUT_VERTEX_F(t[0] * channel->scale[0]);
+ OUT_VERTEX_F(t[1] * channel->scale[1]);
+}
+
+inline static void
+gen4_emit_composite_spans_vertex(struct sna *sna,
+ const struct sna_composite_spans_op *op,
+ int16_t x, int16_t y)
+{
+ OUT_VERTEX(x, y);
+ gen4_emit_composite_texcoord(sna, &op->base.src, x, y);
+}
+
+fastcall static void
+gen4_emit_composite_spans_primitive(struct sna *sna,
+ const struct sna_composite_spans_op *op,
+ const BoxRec *box,
+ float opacity)
+{
+ gen4_emit_composite_spans_vertex(sna, op, box->x2, box->y2);
+ OUT_VERTEX_F(opacity);
+ OUT_VERTEX_F(1);
+ if (!op->base.is_affine)
+ OUT_VERTEX_F(1);
+
+ gen4_emit_composite_spans_vertex(sna, op, box->x1, box->y2);
+ OUT_VERTEX_F(opacity);
+ OUT_VERTEX_F(1);
+ if (!op->base.is_affine)
+ OUT_VERTEX_F(1);
+
+ gen4_emit_composite_spans_vertex(sna, op, box->x1, box->y1);
+ OUT_VERTEX_F(opacity);
+ OUT_VERTEX_F(0);
+ if (!op->base.is_affine)
+ OUT_VERTEX_F(1);
+}
+
+fastcall static void
+gen4_emit_composite_spans_solid(struct sna *sna,
+ const struct sna_composite_spans_op *op,
+ const BoxRec *box,
+ float opacity)
+{
+ OUT_VERTEX(box->x2, box->y2);
+ OUT_VERTEX_F(1); OUT_VERTEX_F(1);
+ OUT_VERTEX_F(opacity); OUT_VERTEX_F(1);
+
+ OUT_VERTEX(box->x1, box->y2);
+ OUT_VERTEX_F(0); OUT_VERTEX_F(1);
+ OUT_VERTEX_F(opacity); OUT_VERTEX_F(1);
+
+ OUT_VERTEX(box->x1, box->y1);
+ OUT_VERTEX_F(0); OUT_VERTEX_F(0);
+ OUT_VERTEX_F(opacity); OUT_VERTEX_F(0);
+}
+
+fastcall static void
+gen4_emit_composite_spans_affine(struct sna *sna,
+ const struct sna_composite_spans_op *op,
+ const BoxRec *box,
+ float opacity)
+{
+ OUT_VERTEX(box->x2, box->y2);
+ gen4_emit_composite_texcoord_affine(sna, &op->base.src,
+ box->x2, box->y2);
+ OUT_VERTEX_F(opacity);
+ OUT_VERTEX_F(1);
+
+ OUT_VERTEX(box->x1, box->y2);
+ gen4_emit_composite_texcoord_affine(sna, &op->base.src,
+ box->x1, box->y2);
+ OUT_VERTEX_F(opacity);
+ OUT_VERTEX_F(1);
+
+ OUT_VERTEX(box->x1, box->y1);
+ gen4_emit_composite_texcoord_affine(sna, &op->base.src,
+ box->x1, box->y1);
+ OUT_VERTEX_F(opacity);
+ OUT_VERTEX_F(0);
+}
+
+fastcall static void
+gen4_render_composite_spans_box(struct sna *sna,
+ const struct sna_composite_spans_op *op,
+ const BoxRec *box, float opacity)
+{
+ DBG(("%s: src=+(%d, %d), opacity=%f, dst=+(%d, %d), box=(%d, %d) x (%d, %d)\n",
+ __FUNCTION__,
+ op->base.src.offset[0], op->base.src.offset[1],
+ opacity,
+ op->base.dst.x, op->base.dst.y,
+ box->x1, box->y1,
+ box->x2 - box->x1,
+ box->y2 - box->y1));
+
+ gen4_get_rectangles(sna, &op->base, 1, gen4_bind_surfaces);
+ op->prim_emit(sna, op, box, opacity);
+ FLUSH_NOCA();
+}
+
+static void
+gen4_render_composite_spans_boxes(struct sna *sna,
+ const struct sna_composite_spans_op *op,
+ const BoxRec *box, int nbox,
+ float opacity)
+{
+ DBG(("%s: nbox=%d, src=+(%d, %d), opacity=%f, dst=+(%d, %d)\n",
+ __FUNCTION__, nbox,
+ op->base.src.offset[0], op->base.src.offset[1],
+ opacity,
+ op->base.dst.x, op->base.dst.y));
+
+ do {
+ gen4_render_composite_spans_box(sna, op, box++, opacity);
+ } while (--nbox);
+}
+
+fastcall static void
+gen4_render_composite_spans_done(struct sna *sna,
+ const struct sna_composite_spans_op *op)
+{
+ if (sna->render_state.gen4.vertex_offset)
+ gen4_vertex_flush(sna);
+
+ DBG(("%s()\n", __FUNCTION__));
+
+ if (op->base.src.bo)
+ kgem_bo_destroy(&sna->kgem, op->base.src.bo);
+
+ sna_render_composite_redirect_done(sna, &op->base);
+}
+
+static bool
+gen4_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height,
+ unsigned flags)
+{
+ if ((flags & COMPOSITE_SPANS_RECTILINEAR) == 0)
+ return false;
+
+ if (op >= ARRAY_SIZE(gen4_blend_op))
+ return false;
+
+ if (gen4_composite_fallback(sna, src, NULL, dst))
+ return false;
+
+ if (!is_gpu(dst->pDrawable))
+ return false;
+
+ return true;
+}
+
+static bool
+gen4_render_composite_spans(struct sna *sna,
+ uint8_t op,
+ PicturePtr src,
+ PicturePtr dst,
+ int16_t src_x, int16_t src_y,
+ int16_t dst_x, int16_t dst_y,
+ int16_t width, int16_t height,
+ unsigned flags,
+ struct sna_composite_spans_op *tmp)
+{
+ DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__,
+ width, height, flags, sna->kgem.ring));
+
+ assert(gen4_check_composite_spans(sna, op, src, dst, width, height, flags));
+
+ if (need_tiling(sna, width, height)) {
+ DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
+ __FUNCTION__, width, height));
+ return sna_tiling_composite_spans(op, src, dst,
+ src_x, src_y, dst_x, dst_y,
+ width, height, flags, tmp);
+ }
+
+ tmp->base.op = op;
+ if (!gen4_composite_set_target(dst, &tmp->base))
+ return false;
+ sna_render_reduce_damage(&tmp->base, dst_x, dst_y, width, height);
+
+ if (too_large(tmp->base.dst.width, tmp->base.dst.height)) {
+ if (!sna_render_composite_redirect(sna, &tmp->base,
+ dst_x, dst_y, width, height))
+ return false;
+ }
+
+ switch (gen4_composite_picture(sna, src, &tmp->base.src,
+ src_x, src_y,
+ width, height,
+ dst_x, dst_y,
+ dst->polyMode == PolyModePrecise)) {
+ case -1:
+ goto cleanup_dst;
+ case 0:
+ gen4_composite_solid_init(sna, &tmp->base.src, 0);
+ /* fall through to fixup */
+ case 1:
+ gen4_composite_channel_convert(&tmp->base.src);
+ break;
+ }
+
+ tmp->base.mask.bo = NULL;
+ tmp->base.is_affine = tmp->base.src.is_affine;
+ tmp->base.has_component_alpha = false;
+ tmp->base.need_magic_ca_pass = false;
+
+ gen4_composite_alpha_gradient_init(sna, &tmp->base.mask);
+
+ tmp->prim_emit = gen4_emit_composite_spans_primitive;
+ if (tmp->base.src.is_solid)
+ tmp->prim_emit = gen4_emit_composite_spans_solid;
+ else if (tmp->base.is_affine)
+ tmp->prim_emit = gen4_emit_composite_spans_affine;
+ tmp->base.floats_per_vertex = 5 + 2*!tmp->base.is_affine;
+ tmp->base.floats_per_rect = 3 * tmp->base.floats_per_vertex;
+
+ tmp->base.u.gen4.wm_kernel =
+ gen4_choose_composite_kernel(tmp->base.op,
+ true, false,
+ tmp->base.is_affine);
+ tmp->base.u.gen4.ve_id = 1 << 1 | tmp->base.is_affine;
+
+ tmp->box = gen4_render_composite_spans_box;
+ tmp->boxes = gen4_render_composite_spans_boxes;
+ tmp->done = gen4_render_composite_spans_done;
+
+ if (!kgem_check_bo(&sna->kgem,
+ tmp->base.dst.bo, tmp->base.src.bo,
+ NULL)) {
+ kgem_submit(&sna->kgem);
+ if (!kgem_check_bo(&sna->kgem,
+ tmp->base.dst.bo, tmp->base.src.bo,
+ NULL))
+ goto cleanup_src;
+ }
+
+ gen4_bind_surfaces(sna, &tmp->base);
+ gen4_align_vertex(sna, &tmp->base);
+ return true;
+
+cleanup_src:
+ if (tmp->base.src.bo)
+ kgem_bo_destroy(&sna->kgem, tmp->base.src.bo);
+cleanup_dst:
+ if (tmp->base.redirect.real_bo)
+ kgem_bo_destroy(&sna->kgem, tmp->base.dst.bo);
+ return false;
+}
+#endif
+
static void
gen4_copy_bind_surfaces(struct sna *sna, const struct sna_composite_op *op)
{
@@ -3472,6 +3792,10 @@ bool gen4_render_init(struct sna *sna)
return false;
sna->render.composite = gen4_render_composite;
+#if !NO_COMPOSITE_SPANS
+ sna->render.check_composite_spans = gen4_check_composite_spans;
+ sna->render.composite_spans = gen4_render_composite_spans;
+#endif
sna->render.video = gen4_render_video;
sna->render.copy_boxes = gen4_render_copy_boxes;
diff --git a/src/sna/gen5_render.c b/src/sna/gen5_render.c
index 51614d4..d776e77 100644
--- a/src/sna/gen5_render.c
+++ b/src/sna/gen5_render.c
@@ -2729,19 +2729,10 @@ gen5_render_composite_spans_done(struct sna *sna,
}
static bool
-gen5_render_composite_spans(struct sna *sna,
- uint8_t op,
- PicturePtr src,
- PicturePtr dst,
- int16_t src_x, int16_t src_y,
- int16_t dst_x, int16_t dst_y,
- int16_t width, int16_t height,
- unsigned flags,
- struct sna_composite_spans_op *tmp)
+gen5_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height, unsigned flags)
{
- DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__,
- width, height, flags, sna->kgem.ring));
-
if ((flags & COMPOSITE_SPANS_RECTILINEAR) == 0)
return false;
@@ -2752,15 +2743,35 @@ gen5_render_composite_spans(struct sna *sna,
return false;
if (need_tiling(sna, width, height)) {
- DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
- __FUNCTION__, width, height));
-
if (!is_gpu(dst->pDrawable)) {
DBG(("%s: fallback, tiled operation not on GPU\n",
__FUNCTION__));
return false;
}
+ }
+
+ return true;
+}
+static bool
+gen5_render_composite_spans(struct sna *sna,
+ uint8_t op,
+ PicturePtr src,
+ PicturePtr dst,
+ int16_t src_x, int16_t src_y,
+ int16_t dst_x, int16_t dst_y,
+ int16_t width, int16_t height,
+ unsigned flags,
+ struct sna_composite_spans_op *tmp)
+{
+ DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__,
+ width, height, flags, sna->kgem.ring));
+
+ assert(gen5_check_composite_spans(sna, op, src, dst, width, height, flags));
+
+ if (need_tiling(sna, width, height)) {
+ DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
+ __FUNCTION__, width, height));
return sna_tiling_composite_spans(op, src, dst,
src_x, src_y, dst_x, dst_y,
width, height, flags, tmp);
@@ -3924,6 +3935,7 @@ bool gen5_render_init(struct sna *sna)
sna->render.composite = gen5_render_composite;
#if !NO_COMPOSITE_SPANS
+ sna->render.check_composite_spans = gen5_check_composite_spans;
sna->render.composite_spans = gen5_render_composite_spans;
#endif
sna->render.video = gen5_render_video;
diff --git a/src/sna/gen6_render.c b/src/sna/gen6_render.c
index 834cab3..389d002 100644
--- a/src/sna/gen6_render.c
+++ b/src/sna/gen6_render.c
@@ -3084,19 +3084,10 @@ gen6_render_composite_spans_done(struct sna *sna,
}
static bool
-gen6_render_composite_spans(struct sna *sna,
- uint8_t op,
- PicturePtr src,
- PicturePtr dst,
- int16_t src_x, int16_t src_y,
- int16_t dst_x, int16_t dst_y,
- int16_t width, int16_t height,
- unsigned flags,
- struct sna_composite_spans_op *tmp)
+gen6_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height, unsigned flags)
{
- DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__,
- width, height, flags, sna->kgem.ring));
-
if ((flags & COMPOSITE_SPANS_RECTILINEAR) == 0)
return false;
@@ -3107,15 +3098,35 @@ gen6_render_composite_spans(struct sna *sna,
return false;
if (need_tiling(sna, width, height)) {
- DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
- __FUNCTION__, width, height));
-
if (!is_gpu(dst->pDrawable)) {
DBG(("%s: fallback, tiled operation not on GPU\n",
__FUNCTION__));
return false;
}
+ }
+
+ return true;
+}
+static bool
+gen6_render_composite_spans(struct sna *sna,
+ uint8_t op,
+ PicturePtr src,
+ PicturePtr dst,
+ int16_t src_x, int16_t src_y,
+ int16_t dst_x, int16_t dst_y,
+ int16_t width, int16_t height,
+ unsigned flags,
+ struct sna_composite_spans_op *tmp)
+{
+ DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__,
+ width, height, flags, sna->kgem.ring));
+
+ assert(gen6_check_composite_spans(sna, op, src, dst, width, height, flags));
+
+ if (need_tiling(sna, width, height)) {
+ DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
+ __FUNCTION__, width, height));
return sna_tiling_composite_spans(op, src, dst,
src_x, src_y, dst_x, dst_y,
width, height, flags, tmp);
@@ -4232,6 +4243,7 @@ bool gen6_render_init(struct sna *sna)
sna->render.composite = gen6_render_composite;
#endif
#if !NO_COMPOSITE_SPANS
+ sna->render.check_composite_spans = gen6_check_composite_spans;
sna->render.composite_spans = gen6_render_composite_spans;
#endif
sna->render.video = gen6_render_video;
diff --git a/src/sna/gen7_render.c b/src/sna/gen7_render.c
index aedc934..18ba826 100644
--- a/src/sna/gen7_render.c
+++ b/src/sna/gen7_render.c
@@ -3166,6 +3166,28 @@ gen7_render_composite_spans_done(struct sna *sna,
}
static bool
+gen7_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height, unsigned flags)
+{
+ if (op >= ARRAY_SIZE(gen7_blend_op))
+ return false;
+
+ if (gen7_composite_fallback(sna, src, NULL, dst))
+ return false;
+
+ if (need_tiling(sna, width, height)) {
+ if (!is_gpu(dst->pDrawable)) {
+ DBG(("%s: fallback, tiled operation not on GPU\n",
+ __FUNCTION__));
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static bool
gen7_render_composite_spans(struct sna *sna,
uint8_t op,
PicturePtr src,
@@ -3179,22 +3201,11 @@ gen7_render_composite_spans(struct sna *sna,
DBG(("%s: %dx%d with flags=%x, current mode=%d\n", __FUNCTION__,
width, height, flags, sna->kgem.ring));
- if (op >= ARRAY_SIZE(gen7_blend_op))
- return false;
-
- if (gen7_composite_fallback(sna, src, NULL, dst))
- return false;
+ assert(gen7_check_composite_spans(sna, op, src, dst, width, height, flags));
if (need_tiling(sna, width, height)) {
DBG(("%s: tiling, operation (%dx%d) too wide for pipeline\n",
__FUNCTION__, width, height));
-
- if (!is_gpu(dst->pDrawable)) {
- DBG(("%s: fallback, tiled operation not on GPU\n",
- __FUNCTION__));
- return false;
- }
-
return sna_tiling_composite_spans(op, src, dst,
src_x, src_y, dst_x, dst_y,
width, height, flags, tmp);
@@ -4313,6 +4324,7 @@ bool gen7_render_init(struct sna *sna)
sna->render.composite = gen7_render_composite;
#endif
#if !NO_COMPOSITE_SPANS
+ sna->render.check_composite_spans = gen7_check_composite_spans;
sna->render.composite_spans = gen7_render_composite_spans;
#endif
sna->render.video = gen7_render_video;
diff --git a/src/sna/sna_render.c b/src/sna/sna_render.c
index cb40cd3..fd105f4 100644
--- a/src/sna/sna_render.c
+++ b/src/sna/sna_render.c
@@ -91,14 +91,22 @@ no_render_composite(struct sna *sna,
dst_x, dst_y,
width, height,
tmp))
- return TRUE;
+ return true;
- return FALSE;
+ return false;
(void)mask_x;
(void)mask_y;
}
static bool
+no_render_check_composite_spans(struct sna *sna,
+ uint8_t op, PicturePtr src, PicturePtr dst,
+ int16_t width, int16_t height, unsigned flags)
+{
+ return false;
+}
+
+static bool
no_render_copy_boxes(struct sna *sna, uint8_t alu,
PixmapPtr src, struct kgem_bo *src_bo, int16_t src_dx, int16_t src_dy,
PixmapPtr dst, struct kgem_bo *dst_bo, int16_t dst_dx, int16_t dst_dy,
@@ -107,7 +115,7 @@ no_render_copy_boxes(struct sna *sna, uint8_t alu,
DBG(("%s (n=%d)\n", __FUNCTION__, n));
if (!sna_blt_compare_depth(&src->drawable, &dst->drawable))
- return FALSE;
+ return false;
return sna_blt_copy_boxes(sna, alu,
src_bo, src_dx, src_dy,
@@ -128,9 +136,9 @@ no_render_copy(struct sna *sna, uint8_t alu,
sna_blt_copy(sna, alu,
src_bo, dst_bo, dst->drawable.bitsPerPixel,
tmp))
- return TRUE;
+ return true;
- return FALSE;
+ return false;
}
static bool
@@ -160,7 +168,7 @@ no_render_fill_boxes(struct sna *sna,
}
if (op != PictOpSrc)
- return FALSE;
+ return false;
if (alu == GXcopy &&
!sna_get_pixel_from_rgba(&pixel,
@@ -169,7 +177,7 @@ no_render_fill_boxes(struct sna *sna,
color->blue,
color->alpha,
format))
- return FALSE;
+ return false;
return sna_blt_fill_boxes(sna, alu,
dst_bo, dst->drawable.bitsPerPixel,
@@ -267,6 +275,7 @@ void no_render_init(struct sna *sna)
render->vertex_size = ARRAY_SIZE(render->vertex_data);
render->composite = no_render_composite;
+ render->check_composite_spans = no_render_check_composite_spans;
render->copy_boxes = no_render_copy_boxes;
render->copy = no_render_copy;
@@ -1314,7 +1323,7 @@ sna_render_picture_convolve(struct sna *sna,
channel->width = w;
channel->filter = PictFilterNearest;
channel->repeat = RepeatNone;
- channel->is_affine = TRUE;
+ channel->is_affine = true;
channel->transform = NULL;
channel->scale[0] = 1.f / w;
channel->scale[1] = 1.f / h;
@@ -1381,7 +1390,7 @@ sna_render_picture_flatten(struct sna *sna,
channel->filter = PictFilterNearest;
channel->repeat = RepeatNone;
channel->pict_format = PIXMAN_a8r8g8b8;
- channel->is_affine = TRUE;
+ channel->is_affine = true;
channel->transform = NULL;
channel->scale[0] = 1.f / w;
channel->scale[1] = 1.f / h;
@@ -1440,7 +1449,7 @@ sna_render_picture_approximate_gradient(struct sna *sna,
return 0;
}
- src = image_from_pict(picture, FALSE, &dx, &dy);
+ src = image_from_pict(picture, false, &dx, &dy);
if (src == NULL) {
pixman_image_unref(dst);
kgem_bo_destroy(&sna->kgem, channel->bo);
@@ -1468,7 +1477,7 @@ sna_render_picture_approximate_gradient(struct sna *sna,
channel->filter = PictFilterNearest;
channel->repeat = RepeatNone;
- channel->is_affine = TRUE;
+ channel->is_affine = true;
channel->scale[0] = 1.f/w;
channel->scale[1] = 1.f/h;
@@ -1561,7 +1570,7 @@ do_fixup:
return 0;
}
- src = image_from_pict(picture, FALSE, &dx, &dy);
+ src = image_from_pict(picture, false, &dx, &dy);
if (src == NULL) {
pixman_image_unref(dst);
kgem_bo_destroy(&sna->kgem, channel->bo);
@@ -1606,7 +1615,7 @@ do_fixup:
channel->filter = PictFilterNearest;
channel->repeat = RepeatNone;
- channel->is_affine = TRUE;
+ channel->is_affine = true;
channel->scale[0] = 1.f/w;
channel->scale[1] = 1.f/h;
@@ -1750,7 +1759,7 @@ sna_render_composite_redirect(struct sna *sna,
struct kgem_bo *bo;
#if NO_REDIRECT
- return FALSE;
+ return false;
#endif
DBG(("%s: target too large (%dx%d), copying to temporary %dx%d, max %d\n",
@@ -1760,11 +1769,11 @@ sna_render_composite_redirect(struct sna *sna,
sna->render.max_3d_size));
if (!width || !height)
- return FALSE;
+ return false;
if (width > sna->render.max_3d_size ||
height > sna->render.max_3d_size)
- return FALSE;
+ return false;
if (op->dst.bo->pitch <= sna->render.max_3d_pitch) {
BoxRec box;
@@ -1842,7 +1851,7 @@ sna_render_composite_redirect(struct sna *sna,
t->real_bo = NULL;
if (t->damage)
__sna_damage_destroy(t->damage);
- return FALSE;
+ return false;
}
assert(op->dst.bo != t->real_bo);
@@ -1852,7 +1861,7 @@ sna_render_composite_redirect(struct sna *sna,
op->dst.y -= box.y1;
op->dst.width = w;
op->dst.height = h;
- return TRUE;
+ return true;
}
}
@@ -1866,7 +1875,7 @@ sna_render_composite_redirect(struct sna *sna,
width, height, bpp),
CREATE_TEMPORARY);
if (!bo)
- return FALSE;
+ return false;
t->box.x1 = x + op->dst.x;
t->box.y1 = y + op->dst.y;
@@ -1881,7 +1890,7 @@ sna_render_composite_redirect(struct sna *sna,
bo, -t->box.x1, -t->box.y1,
bpp, &t->box, 1)) {
kgem_bo_destroy(&sna->kgem, bo);
- return FALSE;
+ return false;
}
t->real_bo = op->dst.bo;
@@ -1897,7 +1906,7 @@ sna_render_composite_redirect(struct sna *sna,
op->dst.y = -y;
op->dst.width = width;
op->dst.height = height;
- return TRUE;
+ return true;
}
void
diff --git a/src/sna/sna_render.h b/src/sna/sna_render.h
index 41cb02b..f0f4a2d 100644
--- a/src/sna/sna_render.h
+++ b/src/sna/sna_render.h
@@ -194,6 +194,9 @@ struct sna_render {
int16_t w, int16_t h,
struct sna_composite_op *tmp);
+ bool (*check_composite_spans)(struct sna *sna, uint8_t op,
+ PicturePtr dst, PicturePtr src,
+ int16_t w, int16_t h, unsigned flags);
bool (*composite_spans)(struct sna *sna, uint8_t op,
PicturePtr dst, PicturePtr src,
int16_t src_x, int16_t src_y,
diff --git a/src/sna/sna_trapezoids.c b/src/sna/sna_trapezoids.c
index c3a5447..efb53dd 100644
--- a/src/sna/sna_trapezoids.c
+++ b/src/sna/sna_trapezoids.c
@@ -3673,7 +3673,9 @@ composite_unaligned_boxes(struct sna *sna,
if (ntrap > 1 && maskFormat)
return false;
- if (force_fallback || !sna->render.composite_spans) {
+ if (force_fallback ||
+ !sna->render.check_composite_spans(sna, op, src, dst, 0, 0,
+ COMPOSITE_SPANS_RECTILINEAR)) {
fallback:
return composite_unaligned_boxes_fallback(op, src, dst,
src_x, src_y,
@@ -3721,6 +3723,15 @@ fallback:
return true;
}
+ if (!sna->render.check_composite_spans(sna, op, src, dst,
+ clip.extents.x2 - clip.extents.x1,
+ clip.extents.y2 - clip.extents.y1,
+ COMPOSITE_SPANS_RECTILINEAR)) {
+ DBG(("%s: fallback -- composite spans not supported\n",
+ __FUNCTION__));
+ goto fallback;
+ }
+
c = NULL;
if (extents.x2 - extents.x1 > clip.extents.x2 - clip.extents.x1 ||
extents.y2 - extents.y1 > clip.extents.y2 - clip.extents.y1)
@@ -4016,7 +4027,7 @@ trapezoid_span_converter(CARD8 op, PicturePtr src, PicturePtr dst,
}
sna = to_sna_from_drawable(dst->pDrawable);
- if (!sna->render.composite_spans) {
+ if (!sna->render.check_composite_spans(sna, op, src, dst, 0, 0, flags)) {
DBG(("%s: fallback -- composite spans not supported\n",
__FUNCTION__));
return false;
@@ -4053,6 +4064,15 @@ trapezoid_span_converter(CARD8 op, PicturePtr src, PicturePtr dst,
return true;
}
+ if (!sna->render.check_composite_spans(sna, op, src, dst,
+ clip.extents.x2 - clip.extents.x1,
+ clip.extents.y2 - clip.extents.y1,
+ flags)) {
+ DBG(("%s: fallback -- composite spans not supported\n",
+ __FUNCTION__));
+ return false;
+ }
+
extents = *RegionExtents(&clip);
dx = dst->pDrawable->x;
dy = dst->pDrawable->y;
@@ -5633,11 +5653,8 @@ trap_span_converter(PicturePtr dst,
struct sna_composite_spans_op tmp;
struct tor tor;
BoxRec extents;
- PicturePtr src;
- xRenderColor white;
pixman_region16_t *clip;
- int dx, dy;
- int n, error;
+ int dx, dy, n;
if (NO_SCAN_CONVERTER)
return false;
@@ -5649,15 +5666,15 @@ trap_span_converter(PicturePtr dst,
return mono_trap_span_converter(dst, src_x, src_y, ntrap, trap);
sna = to_sna_from_drawable(dst->pDrawable);
- if (!sna->render.composite_spans) {
+ if (!sna->render.check_composite_spans(sna, PictOpAdd, sna->render.white_picture, dst,
+ dst->pCompositeClip->extents.x2 - dst->pCompositeClip->extents.x1,
+ dst->pCompositeClip->extents.y2 - dst->pCompositeClip->extents.y1,
+ 0)) {
DBG(("%s: fallback -- composite spans not supported\n",
__FUNCTION__));
return false;
}
- DBG(("%s: extents (%d, %d), (%d, %d)\n",
- __FUNCTION__, extents.x1, extents.y1, extents.x2, extents.y2));
-
clip = dst->pCompositeClip;
extents = *RegionExtents(clip);
dx = dst->pDrawable->x;
@@ -5669,13 +5686,8 @@ trap_span_converter(PicturePtr dst,
extents.x2, extents.y2,
dx, dy));
- white.red = white.green = white.blue = white.alpha = 0xffff;
- src = CreateSolidPicture(0, &white, &error);
- if (src == NULL)
- return true;
-
memset(&tmp, 0, sizeof(tmp));
- if (!sna->render.composite_spans(sna, PictOpAdd, src, dst,
+ if (!sna->render.composite_spans(sna, PictOpAdd, sna->render.white_picture, dst,
0, 0,
extents.x1, extents.y1,
extents.x2 - extents.x1,
@@ -5684,7 +5696,6 @@ trap_span_converter(PicturePtr dst,
&tmp)) {
DBG(("%s: fallback -- composite spans render op not supported\n",
__FUNCTION__));
- FreePicture(src, 0);
return false;
}
@@ -5723,7 +5734,6 @@ trap_span_converter(PicturePtr dst,
skip:
tor_fini(&tor);
tmp.done(sna, &tmp);
- FreePicture(src, 0);
return true;
}
@@ -6175,7 +6185,7 @@ triangles_span_converter(CARD8 op, PicturePtr src, PicturePtr dst,
}
sna = to_sna_from_drawable(dst->pDrawable);
- if (!sna->render.composite_spans) {
+ if (!sna->render.check_composite_spans(sna, op, src, dst, 0, 0, 0)) {
DBG(("%s: fallback -- composite spans not supported\n",
__FUNCTION__));
return false;
@@ -6212,6 +6222,15 @@ triangles_span_converter(CARD8 op, PicturePtr src, PicturePtr dst,
return true;
}
+ if (!sna->render.check_composite_spans(sna, op, src, dst,
+ clip.extents.x2 - clip.extents.x1,
+ clip.extents.y2 - clip.extents.y1,
+ 0)) {
+ DBG(("%s: fallback -- composite spans not supported\n",
+ __FUNCTION__));
+ return false;
+ }
+
extents = *RegionExtents(&clip);
dx = dst->pDrawable->x;
dy = dst->pDrawable->y;
@@ -6531,7 +6550,7 @@ tristrip_span_converter(CARD8 op, PicturePtr src, PicturePtr dst,
}
sna = to_sna_from_drawable(dst->pDrawable);
- if (!sna->render.composite_spans) {
+ if (!sna->render.check_composite_spans(sna, op, src, dst, 0, 0, 0)) {
DBG(("%s: fallback -- composite spans not supported\n",
__FUNCTION__));
return false;
@@ -6568,6 +6587,15 @@ tristrip_span_converter(CARD8 op, PicturePtr src, PicturePtr dst,
return true;
}
+ if (!sna->render.check_composite_spans(sna, op, src, dst,
+ clip.extents.x2 - clip.extents.x1,
+ clip.extents.y2 - clip.extents.y1,
+ 0)) {
+ DBG(("%s: fallback -- composite spans not supported\n",
+ __FUNCTION__));
+ return false;
+ }
+
extents = *RegionExtents(&clip);
dx = dst->pDrawable->x;
dy = dst->pDrawable->y;
commit 5f138176bf15682324d2e8cfa9fac3e49604bf8f
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Thu Jul 19 10:32:33 2012 +0100
sna: Tweak order of screen re-initialisation
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 1b8c8ac..d3dad62 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -12990,6 +12990,11 @@ bool sna_accel_init(ScreenPtr screen, struct sna *sna)
assert(screen->SetWindowPixmap == NULL);
screen->SetWindowPixmap = sna_set_window_pixmap;
+ if (USE_SHM_VMAP && sna->kgem.has_vmap)
+ ShmRegisterFuncs(screen, &shm_funcs);
+ else
+ ShmRegisterFbFuncs(screen);
+
if (!sna_picture_init(screen))
return false;
@@ -13033,11 +13038,6 @@ bool sna_accel_init(ScreenPtr screen, struct sna *sna)
bool sna_accel_create(ScreenPtr screen, struct sna *sna)
{
- if (USE_SHM_VMAP && sna->kgem.has_vmap)
- ShmRegisterFuncs(screen, &shm_funcs);
- else
- ShmRegisterFbFuncs(screen);
-
if (!sna_glyphs_create(sna))
return false;
More information about the xorg-commit
mailing list