xf86-video-intel: 5 commits - src/sna/gen7_render.c src/sna/kgem.c src/sna/sna_accel.c src/sna/sna_display.c
Chris Wilson
ickle at kemper.freedesktop.org
Wed Apr 25 10:55:16 PDT 2012
src/sna/gen7_render.c | 368 +++++++++++++++++++++++---------------------------
src/sna/kgem.c | 3
src/sna/sna_accel.c | 2
src/sna/sna_display.c | 5
4 files changed, 179 insertions(+), 199 deletions(-)
New commits:
commit 5967d76ca09a257ec9db66ea664158e1dfd083ba
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Wed Apr 25 17:15:37 2012 +0100
sna: Fixup broken assertion
It is valid for the cpu_bo to be NULL, as we may be choosing to free the
large shadow pixel buffer instead.
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 382c671..dc084f0 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -2256,7 +2256,7 @@ sna_pixmap_force_to_gpu(PixmapPtr pixmap, unsigned flags)
sna_damage_destroy(&priv->cpu_damage);
priv->undamaged = false;
list_del(&priv->list);
- assert(!priv->cpu_bo->sync);
+ assert(priv->cpu_bo == NULL || !priv->cpu_bo->sync);
sna_pixmap_free_cpu(to_sna_from_pixmap(pixmap), priv);
}
commit 860d3859b586939cd52e45b944cb6abd2a2ca71b
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Wed Apr 25 16:04:33 2012 +0100
sna/gen7: Add CS stall before changing WM binding table
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/gen7_render.c b/src/sna/gen7_render.c
index e2741c4..2228873 100644
--- a/src/sna/gen7_render.c
+++ b/src/sna/gen7_render.c
@@ -837,11 +837,11 @@ gen7_emit_wm(struct sna *sna, unsigned int kernel, int nr_surfaces, int nr_input
OUT_BATCH(0); /* kernel 2 */
}
-static void
+static bool
gen7_emit_binding_table(struct sna *sna, uint16_t offset)
{
if (sna->render_state.gen7.surface_table == offset)
- return;
+ return false;
/* Binding table pointers */
assert(is_aligned(4*offset, 32));
@@ -849,6 +849,7 @@ gen7_emit_binding_table(struct sna *sna, uint16_t offset)
OUT_BATCH(offset*4);
sna->render_state.gen7.surface_table = offset;
+ return true;
}
static void
@@ -970,11 +971,7 @@ gen7_emit_state(struct sna *sna,
const struct sna_composite_op *op,
uint16_t wm_binding_table)
{
- if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) {
- gen7_emit_flush(sna);
- kgem_clear_dirty(&sna->kgem);
- kgem_bo_mark_dirty(op->dst.bo);
- }
+ bool need_stall = false;
gen7_emit_cc(sna,
gen7_get_blend(op->op,
@@ -1001,8 +998,22 @@ gen7_emit_state(struct sna *sna,
op->u.gen7.nr_inputs);
gen7_emit_vertex_elements(sna, op);
- gen7_emit_binding_table(sna, wm_binding_table);
+ need_stall |= gen7_emit_binding_table(sna, wm_binding_table);
gen7_emit_drawing_rectangle(sna, op);
+
+ if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) {
+ gen7_emit_flush(sna);
+ kgem_clear_dirty(&sna->kgem);
+ kgem_bo_mark_dirty(op->dst.bo);
+ need_stall = false;
+ }
+ if (need_stall) {
+ OUT_BATCH(GEN7_PIPE_CONTROL | (4 - 2));
+ OUT_BATCH(GEN7_PIPE_CONTROL_CS_STALL |
+ GEN7_PIPE_CONTROL_STALL_AT_SCOREBOARD);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ }
}
static void gen7_magic_ca_pass(struct sna *sna,
commit c219283460c0f2dfdb823e0cb139d05075c6afce
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Wed Apr 25 15:04:01 2012 +0100
sna/gen7: Apply more recent improvements from SNB perf tuning
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/gen7_render.c b/src/sna/gen7_render.c
index 7dff02f..e2741c4 100644
--- a/src/sna/gen7_render.c
+++ b/src/sna/gen7_render.c
@@ -57,6 +57,8 @@
#define NO_FILL_BOXES 0
#define NO_CLEAR 0
+#define NO_RING_SWITCH 1
+
#define GEN7_MAX_SIZE 16384
/* XXX Todo
@@ -967,7 +969,6 @@ static void
gen7_emit_state(struct sna *sna,
const struct sna_composite_op *op,
uint16_t wm_binding_table)
-
{
if (kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) {
gen7_emit_flush(sna);
@@ -1353,18 +1354,13 @@ gen7_emit_composite_primitive_solid(struct sna *sna,
dst.p.x = r->dst.x + r->width;
dst.p.y = r->dst.y + r->height;
v[0] = dst.f;
- v[1] = 1.;
- v[2] = 1.;
-
dst.p.x = r->dst.x;
v[3] = dst.f;
- v[4] = 0.;
- v[5] = 1.;
-
dst.p.y = r->dst.y;
v[6] = dst.f;
- v[7] = 0.;
- v[8] = 0.;
+
+ v[5] = v[2] = v[1] = 1.;
+ v[8] = v[7] = v[4] = 0.;
}
fastcall static void
@@ -1397,6 +1393,44 @@ gen7_emit_composite_primitive_identity_source(struct sna *sna,
}
fastcall static void
+gen7_emit_composite_primitive_simple_source(struct sna *sna,
+ const struct sna_composite_op *op,
+ const struct sna_composite_rectangles *r)
+{
+ float *v;
+ union {
+ struct sna_coordinate p;
+ float f;
+ } dst;
+
+ float xx = op->src.transform->matrix[0][0];
+ float x0 = op->src.transform->matrix[0][2];
+ float yy = op->src.transform->matrix[1][1];
+ float y0 = op->src.transform->matrix[1][2];
+ float sx = op->src.scale[0];
+ float sy = op->src.scale[1];
+ int16_t tx = op->src.offset[0];
+ int16_t ty = op->src.offset[1];
+
+ v = sna->render.vertices + sna->render.vertex_used;
+ sna->render.vertex_used += 3*3;
+
+ dst.p.x = r->dst.x + r->width;
+ dst.p.y = r->dst.y + r->height;
+ v[0] = dst.f;
+ v[1] = ((r->src.x + r->width + tx) * xx + x0) * sx;
+ v[5] = v[2] = ((r->src.y + r->height + ty) * yy + y0) * sy;
+
+ dst.p.x = r->dst.x;
+ v[3] = dst.f;
+ v[7] = v[4] = ((r->src.x + tx) * xx + x0) * sx;
+
+ dst.p.y = r->dst.y;
+ v[6] = dst.f;
+ v[8] = ((r->src.y + ty) * yy + y0) * sy;
+}
+
+fastcall static void
gen7_emit_composite_primitive_affine_source(struct sna *sna,
const struct sna_composite_op *op,
const struct sna_composite_rectangles *r)
@@ -1486,141 +1520,63 @@ gen7_emit_composite_primitive_identity_source_mask(struct sna *sna,
v[14] = msk_y * op->mask.scale[1];
}
-fastcall static void
-gen7_emit_composite_primitive(struct sna *sna,
- const struct sna_composite_op *op,
- const struct sna_composite_rectangles *r)
+inline static void
+gen7_emit_composite_texcoord(struct sna *sna,
+ const struct sna_composite_channel *channel,
+ int16_t x, int16_t y)
{
- float src_x[3], src_y[3], src_w[3], mask_x[3], mask_y[3], mask_w[3];
- Bool is_affine = op->is_affine;
- const float *src_sf = op->src.scale;
- const float *mask_sf = op->mask.scale;
-
- if (is_affine) {
- sna_get_transformed_coordinates(r->src.x + op->src.offset[0],
- r->src.y + op->src.offset[1],
- op->src.transform,
- &src_x[0],
- &src_y[0]);
-
- sna_get_transformed_coordinates(r->src.x + op->src.offset[0],
- r->src.y + op->src.offset[1] + r->height,
- op->src.transform,
- &src_x[1],
- &src_y[1]);
-
- sna_get_transformed_coordinates(r->src.x + op->src.offset[0] + r->width,
- r->src.y + op->src.offset[1] + r->height,
- op->src.transform,
- &src_x[2],
- &src_y[2]);
- } else {
- if (!sna_get_transformed_coordinates_3d(r->src.x + op->src.offset[0],
- r->src.y + op->src.offset[1],
- op->src.transform,
- &src_x[0],
- &src_y[0],
- &src_w[0]))
- return;
-
- if (!sna_get_transformed_coordinates_3d(r->src.x + op->src.offset[0],
- r->src.y + op->src.offset[1] + r->height,
- op->src.transform,
- &src_x[1],
- &src_y[1],
- &src_w[1]))
- return;
-
- if (!sna_get_transformed_coordinates_3d(r->src.x + op->src.offset[0] + r->width,
- r->src.y + op->src.offset[1] + r->height,
- op->src.transform,
- &src_x[2],
- &src_y[2],
- &src_w[2]))
- return;
- }
+ x += channel->offset[0];
+ y += channel->offset[1];
- if (op->mask.bo) {
- if (is_affine) {
- sna_get_transformed_coordinates(r->mask.x + op->mask.offset[0],
- r->mask.y + op->mask.offset[1],
- op->mask.transform,
- &mask_x[0],
- &mask_y[0]);
-
- sna_get_transformed_coordinates(r->mask.x + op->mask.offset[0],
- r->mask.y + op->mask.offset[1] + r->height,
- op->mask.transform,
- &mask_x[1],
- &mask_y[1]);
-
- sna_get_transformed_coordinates(r->mask.x + op->mask.offset[0] + r->width,
- r->mask.y + op->mask.offset[1] + r->height,
- op->mask.transform,
- &mask_x[2],
- &mask_y[2]);
- } else {
- if (!sna_get_transformed_coordinates_3d(r->mask.x + op->mask.offset[0],
- r->mask.y + op->mask.offset[1],
- op->mask.transform,
- &mask_x[0],
- &mask_y[0],
- &mask_w[0]))
- return;
-
- if (!sna_get_transformed_coordinates_3d(r->mask.x + op->mask.offset[0],
- r->mask.y + op->mask.offset[1] + r->height,
- op->mask.transform,
- &mask_x[1],
- &mask_y[1],
- &mask_w[1]))
- return;
-
- if (!sna_get_transformed_coordinates_3d(r->mask.x + op->mask.offset[0] + r->width,
- r->mask.y + op->mask.offset[1] + r->height,
- op->mask.transform,
- &mask_x[2],
- &mask_y[2],
- &mask_w[2]))
- return;
- }
- }
+ if (channel->is_affine) {
+ float s, t;
- OUT_VERTEX(r->dst.x + r->width, r->dst.y + r->height);
- OUT_VERTEX_F(src_x[2] * src_sf[0]);
- OUT_VERTEX_F(src_y[2] * src_sf[1]);
- if (!is_affine)
- OUT_VERTEX_F(src_w[2]);
- if (op->mask.bo) {
- OUT_VERTEX_F(mask_x[2] * mask_sf[0]);
- OUT_VERTEX_F(mask_y[2] * mask_sf[1]);
- if (!is_affine)
- OUT_VERTEX_F(mask_w[2]);
- }
+ sna_get_transformed_coordinates(x, y,
+ channel->transform,
+ &s, &t);
+ OUT_VERTEX_F(s * channel->scale[0]);
+ OUT_VERTEX_F(t * channel->scale[1]);
+ } else {
+ float s, t, w;
- OUT_VERTEX(r->dst.x, r->dst.y + r->height);
- OUT_VERTEX_F(src_x[1] * src_sf[0]);
- OUT_VERTEX_F(src_y[1] * src_sf[1]);
- if (!is_affine)
- OUT_VERTEX_F(src_w[1]);
- if (op->mask.bo) {
- OUT_VERTEX_F(mask_x[1] * mask_sf[0]);
- OUT_VERTEX_F(mask_y[1] * mask_sf[1]);
- if (!is_affine)
- OUT_VERTEX_F(mask_w[1]);
+ sna_get_transformed_coordinates_3d(x, y,
+ channel->transform,
+ &s, &t, &w);
+ OUT_VERTEX_F(s * channel->scale[0]);
+ OUT_VERTEX_F(t * channel->scale[1]);
+ OUT_VERTEX_F(w);
}
+}
- OUT_VERTEX(r->dst.x, r->dst.y);
- OUT_VERTEX_F(src_x[0] * src_sf[0]);
- OUT_VERTEX_F(src_y[0] * src_sf[1]);
- if (!is_affine)
- OUT_VERTEX_F(src_w[0]);
- if (op->mask.bo) {
- OUT_VERTEX_F(mask_x[0] * mask_sf[0]);
- OUT_VERTEX_F(mask_y[0] * mask_sf[1]);
- if (!is_affine)
- OUT_VERTEX_F(mask_w[0]);
- }
+static void
+gen7_emit_composite_vertex(struct sna *sna,
+ const struct sna_composite_op *op,
+ int16_t srcX, int16_t srcY,
+ int16_t mskX, int16_t mskY,
+ int16_t dstX, int16_t dstY)
+{
+ OUT_VERTEX(dstX, dstY);
+ gen7_emit_composite_texcoord(sna, &op->src, srcX, srcY);
+ gen7_emit_composite_texcoord(sna, &op->mask, mskX, mskY);
+}
+
+fastcall static void
+gen7_emit_composite_primitive(struct sna *sna,
+ const struct sna_composite_op *op,
+ const struct sna_composite_rectangles *r)
+{
+ gen7_emit_composite_vertex(sna, op,
+ r->src.x + r->width, r->src.y + r->height,
+ r->mask.x + r->width, r->mask.y + r->height,
+ r->dst.x + r->width, r->dst.y + r->height);
+ gen7_emit_composite_vertex(sna, op,
+ r->src.x, r->src.y + r->height,
+ r->mask.x, r->mask.y + r->height,
+ r->dst.x, r->dst.y + r->height);
+ gen7_emit_composite_vertex(sna, op,
+ r->src.x, r->src.y,
+ r->mask.x, r->mask.y,
+ r->dst.x, r->dst.y);
}
static void gen7_emit_vertex_buffer(struct sna *sna,
@@ -2064,7 +2020,7 @@ gen7_render_video(struct sna *sna,
tmp.dst.pixmap = pixmap;
tmp.dst.width = pixmap->drawable.width;
tmp.dst.height = pixmap->drawable.height;
- tmp.dst.format = sna_format_for_depth(pixmap->drawable.depth);
+ tmp.dst.format = sna_render_format_for_depth(pixmap->drawable.depth);
tmp.dst.bo = priv->gpu_bo;
tmp.src.bo = frame->bo;
@@ -2163,6 +2119,7 @@ gen7_composite_solid_init(struct sna *sna,
channel->repeat = RepeatNormal;
channel->is_affine = TRUE;
channel->is_solid = TRUE;
+ channel->is_opaque = (color >> 24) == 0xff;
channel->transform = NULL;
channel->width = 1;
channel->height = 1;
@@ -2366,7 +2323,7 @@ gen7_composite_picture(struct sna *sna,
channel->transform = picture->transform;
channel->card_format = gen7_get_card_format(picture->format);
- if (channel->card_format == -1)
+ if (channel->card_format == (unsigned)-1)
return sna_render_picture_convert(sna, picture, channel, pixmap,
x, y, w, h, dst_x, dst_y);
@@ -2411,9 +2368,6 @@ gen7_composite_set_target(struct sna *sna, struct sna_composite_op *op, PictureP
{
struct sna_pixmap *priv;
- if (!gen7_check_dst_format(dst->format))
- return FALSE;
-
op->dst.pixmap = get_drawable_pixmap(dst->pDrawable);
op->dst.width = op->dst.pixmap->drawable.width;
op->dst.height = op->dst.pixmap->drawable.height;
@@ -2453,12 +2407,22 @@ gen7_composite_set_target(struct sna *sna, struct sna_composite_op *op, PictureP
return TRUE;
}
+static bool prefer_blt_ring(struct sna *sna)
+{
+ return sna->kgem.ring != KGEM_RENDER;
+}
+
+static bool can_switch_rings(struct sna *sna)
+{
+ return sna->kgem.has_semaphores && !NO_RING_SWITCH;
+}
+
static Bool
try_blt(struct sna *sna,
PicturePtr dst, PicturePtr src,
int width, int height)
{
- if (sna->kgem.ring != KGEM_RENDER) {
+ if (prefer_blt_ring(sna)) {
DBG(("%s: already performing BLT\n", __FUNCTION__));
return TRUE;
}
@@ -2484,11 +2448,16 @@ try_blt(struct sna *sna,
return TRUE;
}
+ if (can_switch_rings(sna)) {
+ if (sna_picture_is_solid(src, NULL))
+ return TRUE;
+ }
+
return FALSE;
}
static bool
-is_gradient(PicturePtr picture)
+check_gradient(PicturePtr picture)
{
if (picture->pDrawable)
return FALSE;
@@ -2542,9 +2511,10 @@ source_fallback(PicturePtr p, PixmapPtr pixmap)
if (sna_picture_is_solid(p, NULL))
return false;
- if (is_gradient(p) ||
- !gen7_check_repeat(p) ||
- !gen7_check_format(p->format))
+ if (p->pSourcePict)
+ return check_gradient(p);
+
+ if (!gen7_check_repeat(p) || !gen7_check_format(p->format))
return true;
if (pixmap && source_is_busy(pixmap))
@@ -2578,7 +2548,7 @@ gen7_composite_fallback(struct sna *sna,
if (mask) {
mask_pixmap = mask->pDrawable ? get_drawable_pixmap(mask->pDrawable) : NULL;
- mask_fallback = source_fallback(src, mask_pixmap);
+ mask_fallback = source_fallback(mask, mask_pixmap);
} else {
mask_pixmap = NULL;
mask_fallback = false;
@@ -2743,6 +2713,8 @@ gen7_render_composite(struct sna *sna,
width, height,
tmp);
+ if (op == PictOpClear)
+ op = PictOpSrc;
tmp->op = op;
if (!gen7_composite_set_target(sna, tmp, dst))
return FALSE;
@@ -2842,12 +2814,21 @@ gen7_render_composite(struct sna *sna,
tmp->floats_per_vertex = 5 + 2 * !tmp->is_affine;
} else {
- if (tmp->src.is_solid)
+ if (tmp->src.is_solid) {
tmp->prim_emit = gen7_emit_composite_primitive_solid;
- else if (tmp->src.transform == NULL)
+ if (tmp->src.is_opaque && op == PictOpOver)
+ tmp->op = PictOpSrc;
+ } else if (tmp->src.transform == NULL)
tmp->prim_emit = gen7_emit_composite_primitive_identity_source;
- else if (tmp->src.is_affine)
- tmp->prim_emit = gen7_emit_composite_primitive_affine_source;
+ else if (tmp->src.is_affine) {
+ if (tmp->src.transform->matrix[0][1] == 0 &&
+ tmp->src.transform->matrix[1][0] == 0) {
+ tmp->src.scale[0] /= tmp->src.transform->matrix[2][2];
+ tmp->src.scale[1] /= tmp->src.transform->matrix[2][2];
+ tmp->prim_emit = gen7_emit_composite_primitive_simple_source;
+ } else
+ tmp->prim_emit = gen7_emit_composite_primitive_affine_source;
+ }
tmp->floats_per_vertex = 3 + !tmp->is_affine;
}
@@ -2920,32 +2901,6 @@ gen7_composite_alpha_gradient_init(struct sna *sna,
}
inline static void
-gen7_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
gen7_emit_composite_texcoord_affine(struct sna *sna,
const struct sna_composite_channel *channel,
int16_t x, int16_t y)
@@ -3344,7 +3299,7 @@ static inline bool prefer_blt_copy(struct sna *sna,
PixmapPtr src, struct kgem_bo *src_bo,
PixmapPtr dst, struct kgem_bo *dst_bo)
{
- return (sna->kgem.ring != KGEM_RENDER ||
+ return (sna->kgem.ring == KGEM_BLT ||
prefer_blt_bo(sna, src, src_bo) ||
prefer_blt_bo(sna, dst, dst_bo));
}
@@ -3667,7 +3622,7 @@ fallback:
if (!gen7_check_format(op->base.src.pict_format))
goto fallback;
- op->base.op = alu == GXcopy ? PictOpSrc : PictOpClear;
+ op->base.op = PictOpSrc;
op->base.dst.pixmap = dst;
op->base.dst.width = dst->drawable.width;
@@ -3751,7 +3706,9 @@ gen7_emit_fill_state(struct sna *sna, const struct sna_composite_op *op)
static inline bool prefer_blt_fill(struct sna *sna,
struct kgem_bo *bo)
{
- return sna->kgem.ring != KGEM_RENDER || untiled_tlb_miss(bo);
+ return (can_switch_rings(sna) ||
+ prefer_blt_ring(sna) ||
+ untiled_tlb_miss(bo));
}
static Bool
@@ -3810,9 +3767,10 @@ gen7_render_fill_boxes(struct sna *sna,
return FALSE;
#endif
- if (op == PictOpClear)
+ if (op == PictOpClear) {
pixel = 0;
- else if (!sna_get_pixel_from_rgba(&pixel,
+ op = PictOpSrc;
+ } else if (!sna_get_pixel_from_rgba(&pixel,
color->red,
color->green,
color->blue,
@@ -3824,8 +3782,6 @@ gen7_render_fill_boxes(struct sna *sna,
__FUNCTION__, pixel, n,
box[0].x1, box[0].y1, box[0].x2, box[0].y2));
- memset(&tmp, 0, sizeof(tmp));
-
tmp.op = op;
tmp.dst.pixmap = dst;
@@ -3833,14 +3789,21 @@ gen7_render_fill_boxes(struct sna *sna,
tmp.dst.height = dst->drawable.height;
tmp.dst.format = format;
tmp.dst.bo = dst_bo;
+ tmp.dst.x = tmp.dst.y = 0;
tmp.src.bo = sna_render_get_solid(sna, pixel);
tmp.src.filter = SAMPLER_FILTER_NEAREST;
tmp.src.repeat = SAMPLER_EXTEND_REPEAT;
+ tmp.mask.bo = NULL;
+ tmp.mask.filter = SAMPLER_FILTER_NEAREST;
+ tmp.mask.repeat = SAMPLER_EXTEND_NONE;
+
tmp.is_affine = TRUE;
tmp.floats_per_vertex = 3;
tmp.floats_per_rect = 9;
+ tmp.has_component_alpha = FALSE;
+ tmp.need_magic_ca_pass = FALSE;
tmp.u.gen7.wm_kernel = GEN7_WM_KERNEL_NOMASK;
tmp.u.gen7.nr_surfaces = 2;
@@ -4004,7 +3967,7 @@ gen7_render_fill(struct sna *sna, uint8_t alu,
if (alu == GXclear)
color = 0;
- op->base.op = color == 0 ? PictOpClear : PictOpSrc;
+ op->base.op = PictOpSrc;
op->base.dst.pixmap = dst;
op->base.dst.width = dst->drawable.width;
@@ -4097,7 +4060,7 @@ gen7_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
if (alu == GXclear)
color = 0;
- tmp.op = color == 0 ? PictOpClear : PictOpSrc;
+ tmp.op = PictOpSrc;
tmp.dst.pixmap = dst;
tmp.dst.width = dst->drawable.width;
@@ -4195,7 +4158,7 @@ gen7_render_clear(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo)
if (too_large(dst->drawable.width, dst->drawable.height))
return gen7_render_clear_try_blt(sna, dst, bo);
- tmp.op = PictOpClear;
+ tmp.op = PictOpSrc;
tmp.dst.pixmap = dst;
tmp.dst.width = dst->drawable.width;
commit a3371613c9bf577a69cdf811ca1bebaea46bbe95
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Wed Apr 25 11:09:35 2012 +0100
sna: Do not automagically convert GTT mappings on untiled scanout to CPU
The likelihood of an untiled mapping of the scanout is slim, except for
gen3 with large desktops, and there it should never be in the CPU
domain...
The issue is that we may perform an operation "inplace", yet incoherent
with the display engine, and never flush the CPU cache, resulting in
render corruption. In theory at least!
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/kgem.c b/src/sna/kgem.c
index d97f559..4c4aa7c 100644
--- a/src/sna/kgem.c
+++ b/src/sna/kgem.c
@@ -3167,7 +3167,7 @@ void *kgem_bo_map(struct kgem *kgem, struct kgem_bo *bo)
assert(bo->exec == NULL);
assert(list_is_empty(&bo->list));
- if (bo->tiling == I915_TILING_NONE &&
+ if (bo->tiling == I915_TILING_NONE && !bo->scanout &&
(kgem->has_llc || bo->domain == DOMAIN_CPU)) {
DBG(("%s: converting request for GTT map into CPU map\n",
__FUNCTION__));
@@ -3274,6 +3274,7 @@ void *kgem_bo_map__cpu(struct kgem *kgem, struct kgem_bo *bo)
DBG(("%s(handle=%d, size=%d)\n", __FUNCTION__, bo->handle, bytes(bo)));
assert(!bo->purged);
assert(list_is_empty(&bo->list));
+ assert(!bo->scanout);
if (IS_CPU_MAP(bo->map))
return MAP(bo->map);
commit 1abd92cd012ee46d44ed4873a5e750d56ae6668f
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date: Wed Apr 25 11:16:30 2012 +0100
sna: Clear the domain tracking after attaching the bo to scanout
This is basically to make sure we don't continue treating it as CPU
coherent.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
diff --git a/src/sna/sna_display.c b/src/sna/sna_display.c
index cb051b0..5275d4a 100644
--- a/src/sna/sna_display.c
+++ b/src/sna/sna_display.c
@@ -436,6 +436,7 @@ sna_crtc_restore(struct sna *sna)
assert(bo->tiling != I915_TILING_Y);
bo->scanout = true;
+ bo->domain = DOMAIN_NONE;
DBG(("%s: create fb %dx%d@%d/%d\n",
__FUNCTION__,
@@ -673,6 +674,7 @@ sna_crtc_set_mode_major(xf86CrtcPtr crtc, DisplayModePtr mode,
DBG(("%s: handle %d attached to fb %d\n",
__FUNCTION__, bo->handle, sna_mode->fb_id));
bo->scanout = true;
+ bo->domain = DOMAIN_NONE;
sna_mode->fb_pixmap = sna->front->drawable.serialNumber;
}
@@ -787,6 +789,7 @@ sna_crtc_shadow_allocate(xf86CrtcPtr crtc, int width, int height)
DBG(("%s: attached handle %d to fb %d\n",
__FUNCTION__, bo->handle, sna_crtc->shadow_fb_id));
bo->scanout = true;
+ bo->domain = DOMAIN_NONE;
return sna_crtc->shadow = shadow;
}
@@ -1725,6 +1728,7 @@ sna_crtc_resize(ScrnInfoPtr scrn, int width, int height)
if (!sna_crtc_apply(crtc))
goto fail;
}
+ bo->domain = DOMAIN_NONE;
scrn->virtualX = width;
scrn->virtualY = height;
@@ -1869,6 +1873,7 @@ sna_page_flip(struct sna *sna,
DBG(("%s: page flipped %d crtcs\n", __FUNCTION__, count));
if (count) {
bo->scanout = true;
+ bo->domain = DOMAIN_NONE;
} else {
drmModeRmFB(sna->kgem.fd, mode->fb_id);
mode->fb_id = *old_fb;
More information about the xorg-commit
mailing list