Splitting intel_miptree_map() like so should help with the yuck factor.
Though don't we also need to treat the stencil_mt to a similar treatment
to avoid slow reads?
Note the map should really record what method intel_miptree_map() used
so that is can be unwound correctly without chasing the same decision
tree (too easy for mistakes to occur).
---
src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 150 +++++++++++-------
1 file changed, 95 insertions(+), 55 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
index 89074a64930..1314547cc6c 100644
--- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
@@ -49,6 +49,17 @@
#define FILE_DEBUG_FLAG DEBUG_MIPTREE
+static void __intel_miptree_map(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ unsigned int level,
+ unsigned int slice,
+ struct intel_miptree_map *map);
+static void __intel_miptree_unmap(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ unsigned int level,
+ unsigned int slice,
+ struct intel_miptree_map *map);
+
static void *intel_miptree_map_raw(struct brw_context *brw,
struct intel_mipmap_tree *mt,
GLbitfield mode);
@@ -3441,27 +3452,31 @@ intel_miptree_map_depthstencil(struct brw_context *brw,
if (!(map->mode & GL_MAP_INVALIDATE_RANGE_BIT)) {
uint32_t *packed_map = map->ptr;
uint8_t *s_map = intel_miptree_map_raw(brw, s_mt, GL_MAP_READ_BIT);
- uint32_t *z_map = intel_miptree_map_raw(brw, z_mt, GL_MAP_READ_BIT);
unsigned int s_image_x, s_image_y;
- unsigned int z_image_x, z_image_y;
+
+ struct intel_miptree_map z_map = {
+ .mode = GL_MAP_READ_BIT | BRW_MAP_DIRECT_BIT,
+ .x = map->x,
+ .y = map->y,
+ .w = map->w,
+ .h = map->h,
+ };
+ __intel_miptree_map(brw, z_mt, level, slice, &z_map);
intel_miptree_get_image_offset(s_mt, level, slice,
&s_image_x, &s_image_y);
- intel_miptree_get_image_offset(z_mt, level, slice,
- &z_image_x, &z_image_y);
for (uint32_t y = 0; y < map->h; y++) {
+ uint32_t *z_line =
+ (uint32_t *)((uint8_t *)z_map.ptr + z_map.stride * y);
for (uint32_t x = 0; x < map->w; x++) {
int map_x = map->x + x, map_y = map->y + y;
ptrdiff_t s_offset = intel_offset_S8(s_mt->surf.row_pitch,
map_x + s_image_x,
map_y + s_image_y,
brw->has_swizzling);
- ptrdiff_t z_offset = ((map_y + z_image_y) *
- (z_mt->surf.row_pitch / 4) +
- (map_x + z_image_x));
uint8_t s = s_map[s_offset];
- uint32_t z = z_map[z_offset];
+ uint32_t z = z_line[x];
if (map_z32f_x24s8) {
packed_map[(y * map->w + x) * 2 + 0] = z;
@@ -3472,13 +3487,13 @@ intel_miptree_map_depthstencil(struct brw_context *brw,
}
}
+ __intel_miptree_unmap(brw, z_mt, level, slice, &z_map);
intel_miptree_unmap_raw(s_mt);
- intel_miptree_unmap_raw(z_mt);
DBG("%s: %d,%d %dx%d from z mt %p %d,%d, s mt %p %d,%d = %p/%d\n",
__func__,
map->x, map->y, map->w, map->h,
- z_mt, map->x + z_image_x, map->y + z_image_y,
+ z_mt, map->x, map->y,
s_mt, map->x + s_image_x, map->y + s_image_y,
map->ptr, map->stride);
} else {
@@ -3502,44 +3517,47 @@ intel_miptree_unmap_depthstencil(struct brw_context
*brw,
if (map->mode & GL_MAP_WRITE_BIT) {
uint32_t *packed_map = map->ptr;
uint8_t *s_map = intel_miptree_map_raw(brw, s_mt, GL_MAP_WRITE_BIT);
- uint32_t *z_map = intel_miptree_map_raw(brw, z_mt, GL_MAP_WRITE_BIT);
unsigned int s_image_x, s_image_y;
- unsigned int z_image_x, z_image_y;
intel_miptree_get_image_offset(s_mt, level, slice,
&s_image_x, &s_image_y);
- intel_miptree_get_image_offset(z_mt, level, slice,
- &z_image_x, &z_image_y);
+
+ struct intel_miptree_map z_map = {
+ .mode = GL_MAP_WRITE_BIT | BRW_MAP_DIRECT_BIT |
GL_MAP_INVALIDATE_RANGE_BIT,
+ .x = map->x,
+ .y = map->y,
+ .w = map->w,
+ .h = map->h,
+ };
+ __intel_miptree_map(brw, z_mt, level, slice, &z_map);
for (uint32_t y = 0; y < map->h; y++) {
+ uint32_t *z_line =
+ (uint32_t *)((uint8_t *)z_map.ptr + z_map.stride * y);
for (uint32_t x = 0; x < map->w; x++) {
ptrdiff_t s_offset = intel_offset_S8(s_mt->surf.row_pitch,
x + s_image_x + map->x,
y + s_image_y + map->y,
brw->has_swizzling);
- ptrdiff_t z_offset = ((y + z_image_y + map->y) *
- (z_mt->surf.row_pitch / 4) +
- (x + z_image_x + map->x));
if (map_z32f_x24s8) {
- z_map[z_offset] = packed_map[(y * map->w + x) * 2 + 0];
+ z_line[x] = packed_map[(y * map->w + x) * 2 + 0];
s_map[s_offset] = packed_map[(y * map->w + x) * 2 + 1];
} else {
uint32_t packed = packed_map[y * map->w + x];
s_map[s_offset] = packed >> 24;
- z_map[z_offset] = packed;
+ z_line[x] = packed;
}
}
}
+ __intel_miptree_unmap(brw, z_mt, level, slice, &z_map);
intel_miptree_unmap_raw(s_mt);
- intel_miptree_unmap_raw(z_mt);
DBG("%s: %d,%d %dx%d from z mt %p (%s) %d,%d, s mt %p %d,%d = %p/%d\n",
__func__,
map->x, map->y, map->w, map->h,
- z_mt, _mesa_get_format_name(z_mt->format),
- map->x + z_image_x, map->y + z_image_y,
+ z_mt, _mesa_get_format_name(z_mt->format), map->x, map->y,
s_mt, map->x + s_image_x, map->y + s_image_y,
map->ptr, map->stride);
}
@@ -3636,6 +3654,38 @@ use_intel_mipree_map_blit(struct brw_context *brw,
return false;
}
+static void
+__intel_miptree_map(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ unsigned int level,
+ unsigned int slice,
+ struct intel_miptree_map *map)
+{
+ assert(mt->surf.samples == 1);
+
+ intel_miptree_access_raw(brw, mt, level, slice,
+ map->mode & GL_MAP_WRITE_BIT);
+
+ if (mt->format == MESA_FORMAT_S_UINT8) {
+ intel_miptree_map_s8(brw, mt, map, level, slice);
+ } else if (mt->etc_format != MESA_FORMAT_NONE &&
+ !(map->mode & BRW_MAP_DIRECT_BIT)) {
+ intel_miptree_map_etc(brw, mt, map, level, slice);
+ } else if (mt->stencil_mt && !(map->mode & BRW_MAP_DIRECT_BIT)) {
+ intel_miptree_map_depthstencil(brw, mt, map, level, slice);
+ } else if (use_intel_mipree_map_blit(brw, mt, map->mode, level, slice)) {
+ intel_miptree_map_blit(brw, mt, map, level, slice);
+#if defined(USE_SSE41)
+ } else if (!(map->mode & GL_MAP_WRITE_BIT) &&
+ !mt->compressed && cpu_has_sse4_1 &&
+ (mt->surf.row_pitch % 16 == 0)) {
+ intel_miptree_map_movntdqa(brw, mt, map, level, slice);
+#endif
+ } else {
+ intel_miptree_map_gtt(brw, mt, map, level, slice);
+ }
+}
+
/**
* Parameter \a out_stride has type ptrdiff_t not because the buffer stride may
* exceed 32 bits but to diminish the likelihood subtle bugs in pointer
@@ -3671,27 +3721,7 @@ intel_miptree_map(struct brw_context *brw,
return;
}
- intel_miptree_access_raw(brw, mt, level, slice,
- map->mode & GL_MAP_WRITE_BIT);
-
- if (mt->format == MESA_FORMAT_S_UINT8) {
- intel_miptree_map_s8(brw, mt, map, level, slice);
- } else if (mt->etc_format != MESA_FORMAT_NONE &&
- !(mode & BRW_MAP_DIRECT_BIT)) {
- intel_miptree_map_etc(brw, mt, map, level, slice);
- } else if (mt->stencil_mt && !(mode & BRW_MAP_DIRECT_BIT)) {
- intel_miptree_map_depthstencil(brw, mt, map, level, slice);
- } else if (use_intel_mipree_map_blit(brw, mt, mode, level, slice)) {
- intel_miptree_map_blit(brw, mt, map, level, slice);
-#if defined(USE_SSE41)
- } else if (!(mode & GL_MAP_WRITE_BIT) &&
- !mt->compressed && cpu_has_sse4_1 &&
- (mt->surf.row_pitch % 16 == 0)) {
- intel_miptree_map_movntdqa(brw, mt, map, level, slice);
-#endif
- } else {
- intel_miptree_map_gtt(brw, mt, map, level, slice);
- }
+ __intel_miptree_map(brw, mt, level, slice, map);
*out_ptr = map->ptr;
*out_stride = map->stride;
@@ -3700,19 +3730,13 @@ intel_miptree_map(struct brw_context *brw,
intel_miptree_release_map(mt, level, slice);
}
-void
-intel_miptree_unmap(struct brw_context *brw,
- struct intel_mipmap_tree *mt,
- unsigned int level,
- unsigned int slice)
+static void
+__intel_miptree_unmap(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ unsigned int level,
+ unsigned int slice,
+ struct intel_miptree_map *map)
{
- struct intel_miptree_map *map = mt->level[level].slice[slice].map;
-
- assert(mt->surf.samples == 1);
-
- if (!map)
- return;
-
DBG("%s: mt %p (%s) level %d slice %d\n", __func__,
mt, _mesa_get_format_name(mt->format), level, slice);
@@ -3732,6 +3756,22 @@ intel_miptree_unmap(struct brw_context *brw,
} else {
intel_miptree_unmap_gtt(mt);
}
+}
+
+void
+intel_miptree_unmap(struct brw_context *brw,
+ struct intel_mipmap_tree *mt,
+ unsigned int level,
+ unsigned int slice)
+{
+ struct intel_miptree_map *map = mt->level[level].slice[slice].map;
+
+ assert(mt->surf.samples == 1);
+
+ if (!map)
+ return;
+
+ __intel_miptree_unmap(brw, mt, level, slice, map);
intel_miptree_release_map(mt, level, slice);
}
--
2.17.0
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
