Uncommonly we may be able to blit into the texture where we cannot
perform the tiled memcpy fast path, for example on older generations and
non-LLC architectures (though those restrictions may be lifted in
future). Using the GPU blit, even with a linear source and forced stall,
is still much faster than the mesa fallback paths, make it so.
Signed-off-by: Chris Wilson <ch...@chris-wilson.co.uk>
---
src/mesa/drivers/dri/i965/intel_tex.h | 6 +
src/mesa/drivers/dri/i965/intel_tex_image.c | 36 +++--
src/mesa/drivers/dri/i965/intel_tex_subimage.c | 197 +++++++++++++++++++++++--
3 files changed, 204 insertions(+), 35 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/intel_tex.h
b/src/mesa/drivers/dri/i965/intel_tex.h
index 376f075c9e..5e2f9297eb 100644
--- a/src/mesa/drivers/dri/i965/intel_tex.h
+++ b/src/mesa/drivers/dri/i965/intel_tex.h
@@ -63,6 +63,12 @@ intel_texsubimage_tiled_memcpy(struct gl_context *ctx,
const GLvoid *pixels,
const struct gl_pixelstore_attrib *packing,
bool for_glTexImage);
+bool
+intel_texsubimage_pbo(struct gl_context *ctx, GLuint dims,
+ struct gl_texture_image *texImage,
+ GLint x, GLint y, GLint z,
+ GLsizei width, GLsizei height, GLsizei depth,
+ GLenum format, GLenum type, const GLvoid *pixels);
bool
intel_gettexsubimage_tiled_memcpy(struct gl_context *ctx,
diff --git a/src/mesa/drivers/dri/i965/intel_tex_image.c
b/src/mesa/drivers/dri/i965/intel_tex_image.c
index 334e323d74..7cd8a9b66a 100644
--- a/src/mesa/drivers/dri/i965/intel_tex_image.c
+++ b/src/mesa/drivers/dri/i965/intel_tex_image.c
@@ -137,9 +137,6 @@ intelTexImage(struct gl_context * ctx,
const struct gl_pixelstore_attrib *unpack)
{
struct intel_texture_image *intelImage = intel_texture_image(texImage);
- bool ok;
-
- bool tex_busy = intelImage->mt && brw_bo_busy(intelImage->mt->bo);
DBG("%s mesa_format %s target %s format %s type %s level %d %dx%dx%d\n",
__func__, _mesa_get_format_name(texImage->TexFormat),
@@ -158,22 +155,23 @@ intelTexImage(struct gl_context * ctx,
if (intelImage->mt->format == MESA_FORMAT_S_UINT8)
intelImage->mt->r8stencil_needs_update = true;
- ok = _mesa_meta_pbo_TexSubImage(ctx, dims, texImage, 0, 0, 0,
- texImage->Width, texImage->Height,
- texImage->Depth,
- format, type, pixels,
- tex_busy, unpack);
- if (ok)
- return;
+ if (!_mesa_is_bufferobj(unpack->BufferObj)) {
+ if (intel_texsubimage_tiled_memcpy(ctx, dims, texImage,
+ 0, 0, 0, /*x,y,z offsets*/
+ texImage->Width,
+ texImage->Height,
+ texImage->Depth,
+ format, type, pixels, unpack,
+ false /*allocate_storage*/))
+ return;
+ }
- ok = intel_texsubimage_tiled_memcpy(ctx, dims, texImage,
- 0, 0, 0, /*x,y,z offsets*/
- texImage->Width,
- texImage->Height,
- texImage->Depth,
- format, type, pixels, unpack,
- false /*allocate_storage*/);
- if (ok)
+ if (intel_texsubimage_pbo(ctx, dims, texImage,
+ 0, 0, 0, /*x,y,z offsets*/
+ texImage->Width,
+ texImage->Height,
+ texImage->Depth,
+ format, type, pixels))
return;
DBG("%s: upload image %dx%dx%d pixels %p\n",
@@ -735,7 +733,7 @@ intel_gettexsubimage_pbo(struct gl_context *ctx,
intel_buffer_object(pack->BufferObj),
dst_offset,
dst_size,
- false);
+ true);
} else {
dst_offset = (GLintptr)pixels & 4095;
if ((dst_offset | dst_stride) % bpp)
diff --git a/src/mesa/drivers/dri/i965/intel_tex_subimage.c
b/src/mesa/drivers/dri/i965/intel_tex_subimage.c
index e7486300ab..77db18242c 100644
--- a/src/mesa/drivers/dri/i965/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/i965/intel_tex_subimage.c
@@ -25,21 +25,25 @@
#include "main/bufferobj.h"
#include "main/image.h"
+#include "main/glformats.h"
#include "main/macros.h"
#include "main/mtypes.h"
#include "main/glformats.h"
#include "main/pbo.h"
+#include "main/readpix.h"
#include "main/texobj.h"
#include "main/texstore.h"
#include "main/texcompress.h"
#include "main/enums.h"
#include "drivers/common/meta.h"
+#include "brw_blorp.h"
#include "brw_context.h"
#include "intel_batchbuffer.h"
#include "intel_tex.h"
#include "intel_mipmap_tree.h"
#include "intel_blit.h"
+#include "intel_buffer_objects.h"
#include "intel_tiled_memcpy.h"
#define FILE_DEBUG_FLAG DEBUG_TEXTURE
@@ -212,6 +216,169 @@ intel_texsubimage_tiled_memcpy(struct gl_context * ctx,
return true;
}
+bool
+intel_texsubimage_pbo(struct gl_context *ctx, GLuint dims,
+ struct gl_texture_image *tex,
+ GLint x, GLint y, GLint z,
+ GLsizei width, GLsizei height, GLsizei depth,
+ GLenum format, GLenum type, const GLvoid *pixels)
+{
+ struct brw_context *brw = brw_context(ctx);
+ struct intel_texture_image *image = intel_texture_image(tex);
+ const struct gl_pixelstore_attrib *pack = &ctx->Unpack;
+ bool ok;
+
+ if (ctx->_ImageTransferState)
+ return false;
+
+ if (pack->SwapBytes || pack->LsbFirst) {
+ DBG("%s: bad packing params\n", __func__);
+ return false;
+ }
+
+ if (format == GL_COLOR_INDEX)
+ return false;
+
+ if (_mesa_need_rgb_to_luminance_conversion(format, tex->_BaseFormat))
+ return false;
+
+ struct intel_mipmap_tree *dst = image->mt;
+ if (format == GL_STENCIL_INDEX && dst->stencil_mt)
+ dst = dst->stencil_mt;
+ if (!brw->mesa_format_supports_render[dst->format])
+ return false;
+
+ mesa_format src_format = _mesa_format_from_format_and_type(format, type);
+ if (_mesa_format_is_mesa_array_format(src_format))
+ src_format = _mesa_format_from_array_format(src_format);
+ if (!brw->screen->mesa_format_supports_texture[src_format])
+ return false;
+
+ bool src_flip = false;
+ int src_stride = _mesa_image_row_stride(pack, width, format, type);
+ if (pack->Invert) {
+ src_stride = -src_stride;
+ src_flip = true;
+ }
+
+ int image_height = pack->ImageHeight == 0 ? height : pack->ImageHeight;
+ int full_height = image_height * (depth - 1) + height;
+ int bpp = _mesa_bytes_per_pixel(format, type);
+ pixels += (pack->SkipImages * image_height + pack->SkipRows) * src_stride;
+ pixels += pack->SkipPixels * bpp;
+ int src_size = (full_height - 1) * src_stride + width * bpp;
+
+ struct brw_bo *src_buffer;
+ int src_offset;
+ if (_mesa_is_bufferobj(pack->BufferObj)) {
+ src_offset = (GLintptr)pixels,
+ src_buffer = intel_bufferobj_buffer(brw,
+
intel_buffer_object(pack->BufferObj),
+ src_offset,
+ src_size,
+ true);
+ } else if (brw_bo_busy(dst->bo)) {
+ src_offset = 0;
+ src_buffer = brw_bo_alloc(brw->bufmgr, "texsubimage", src_size, 0);
+ if (!src_buffer)
+ return false;
+
+ void *map = brw_bo_map(brw, src_buffer, MAP_WRITE);
+ if (!map)
+ goto err_src_buffer;
+
+ memcpy(map, pixels, src_size);
+ brw_bo_unmap(src_buffer);
+ } else {
+ src_offset = (GLintptr)pixels & 4095;
+ src_buffer = brw_bo_alloc_userptr(brw->bufmgr, "texsubimage(userptr)",
+ (void *)((GLintptr)pixels & -4096),
+ ALIGN(src_size + src_offset, 4096), 0);
+ }
+ if (!src_buffer)
+ return false;
+
+ if (tex->TexObject->Target == GL_TEXTURE_1D_ARRAY) {
+ assert(depth == 1);
+ assert(z == 0);
+ depth = height;
+ height = 1;
+ image_height = 1;
+ z = y;
+ y = 0;
+ }
+
+ struct intel_mipmap_tree *src =
+ intel_miptree_create_for_bo(brw,
+ src_buffer,
+ src_format,
+ src_offset,
+ width, full_height, 1,
+ src_stride,
+ MIPTREE_LAYOUT_TILING_NONE |
+ MIPTREE_LAYOUT_DISABLE_AUX |
+ MIPTREE_LAYOUT_FOR_TEXTURE);
+ if (!src)
+ goto err_src_buffer;
+
+ int dst_level = tex->Level + tex->TexObject->MinLevel;
+ int dst_slice = tex->Face + tex->TexObject->MinLayer + z;
+
+ /* blorp restrictions; see try_blorp_blit() */
+ mesa_format dst_format = dst->format;
+ switch (format) {
+ case GL_STENCIL_INDEX:
+ if (src_format != dst_format)
+ goto err_src_mt;
+
+ /* fallthrough */
+ case GL_DEPTH_COMPONENT:
+ case GL_DEPTH_STENCIL_EXT:
+ if ((src_format == MESA_FORMAT_Z24_UNORM_X8_UINT) !=
+ (dst_format == MESA_FORMAT_Z24_UNORM_X8_UINT))
+ goto err_src_mt;
+
+ if (_mesa_get_format_base_format(src_format) == GL_DEPTH_STENCIL ||
+ _mesa_get_format_base_format(dst_format) == GL_DEPTH_STENCIL)
+ goto err_src_mt;
+
+ src_format = MESA_FORMAT_NONE;
+ dst_format = MESA_FORMAT_NONE;
+ break;
+
+ default:
+ break;
+ }
+
+ for (z = 0; depth--; z++) {
+ if (!intel_miptree_blit(brw,
+ src, 0, 0, 0, z * image_height, src_flip,
+ dst, dst_level, dst_slice + z, x, y, false,
+ width, height, GL_COPY)) {
+ brw_blorp_blit_miptrees(brw,
+ src, 0, 0,
+ src_format, SWIZZLE_XYZW,
+ dst, dst_level, dst_slice + z, dst_format,
+ 0, z * image_height, width, z * image_height
+ height,
+ x, y, x + width, y + height,
+ GL_NEAREST, false, src_flip,
+ false, false);
+ }
+ }
+ if (src_buffer->userptr) {
+ intel_batchbuffer_flush(brw);
+ brw_bo_wait_rendering(src_buffer);
+ }
+
+ ok = true;
+err_src_mt:
+ intel_miptree_release(&src);
+err_src_buffer:
+ if (!_mesa_is_bufferobj(pack->BufferObj))
+ brw_bo_unreference(src_buffer);
+ return ok;
+}
+
static void
intelTexSubImage(struct gl_context * ctx,
GLuint dims,
@@ -223,9 +390,6 @@ intelTexSubImage(struct gl_context * ctx,
const struct gl_pixelstore_attrib *packing)
{
struct intel_mipmap_tree *mt = intel_texture_image(texImage)->mt;
- bool ok;
-
- bool tex_busy = mt && brw_bo_busy(mt->bo);
if (mt && mt->format == MESA_FORMAT_S_UINT8)
mt->r8stencil_needs_update = true;
@@ -236,20 +400,21 @@ intelTexSubImage(struct gl_context * ctx,
_mesa_enum_to_string(format), _mesa_enum_to_string(type),
texImage->Level, texImage->Width, texImage->Height, texImage->Depth);
- ok = _mesa_meta_pbo_TexSubImage(ctx, dims, texImage,
- xoffset, yoffset, zoffset,
- width, height, depth, format, type,
- pixels, tex_busy, packing);
- if (ok)
- return;
+ if (!_mesa_is_bufferobj(packing->BufferObj)) {
+ if (intel_texsubimage_tiled_memcpy(ctx, dims, texImage,
+ xoffset, yoffset, zoffset,
+ width, height, depth,
+ format, type, pixels, packing,
+ false /*for_glTexImage*/))
+ return;
+ }
- ok = intel_texsubimage_tiled_memcpy(ctx, dims, texImage,
- xoffset, yoffset, zoffset,
- width, height, depth,
- format, type, pixels, packing,
- false /*for_glTexImage*/);
- if (ok)
- return;
+
+ if (intel_texsubimage_pbo(ctx, dims, texImage,
+ xoffset, yoffset, zoffset,
+ width, height, depth,
+ format, type, pixels))
+ return;
_mesa_store_texsubimage(ctx, dims, texImage,
xoffset, yoffset, zoffset,
--
2.13.3
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
