On 03/04/2014 02:17 PM, Eric Anholt wrote:
> The meta patch supports all the same, except for the stencil case being
> unaccelerated.
> ---
> src/mesa/drivers/dri/i965/brw_blorp_blit.cpp | 105
> ---------------------------
> src/mesa/drivers/dri/i965/brw_context.h | 9 ---
> src/mesa/drivers/dri/i965/intel_tex_copy.c | 9 +--
> 3 files changed, 1 insertion(+), 122 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
> b/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
> index acdc9c7..fbed8ce 100644
> --- a/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
> +++ b/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
> @@ -334,111 +334,6 @@ try_blorp_blit(struct brw_context *brw,
> return true;
> }
>
> -bool
> -brw_blorp_copytexsubimage(struct brw_context *brw,
> - struct gl_renderbuffer *src_rb,
> - struct gl_texture_image *dst_image,
> - int slice,
> - int srcX0, int srcY0,
> - int dstX0, int dstY0,
> - int width, int height)
> -{
> - struct gl_context *ctx = &brw->ctx;
> - struct intel_renderbuffer *src_irb = intel_renderbuffer(src_rb);
> - struct intel_texture_image *intel_image = intel_texture_image(dst_image);
> -
> - /* Sync up the state of window system buffers. We need to do this before
> - * we go looking at the src renderbuffer's miptree.
> - */
> - intel_prepare_render(brw);
> -
> - struct intel_mipmap_tree *src_mt = src_irb->mt;
> - struct intel_mipmap_tree *dst_mt = intel_image->mt;
> -
> - /* BLORP is not supported before Gen6. */
> - if (brw->gen < 6 || brw->gen >= 8)
> - return false;
> -
> - if (_mesa_get_format_base_format(src_mt->format) !=
> - _mesa_get_format_base_format(dst_mt->format)) {
> - return false;
> - }
> -
> - /* We can't handle format conversions between Z24 and other formats since
> - * we have to lie about the surface format. See the comments in
> - * brw_blorp_surface_info::set().
> - */
> - if ((src_mt->format == MESA_FORMAT_Z24_UNORM_X8_UINT) !=
> - (dst_mt->format == MESA_FORMAT_Z24_UNORM_X8_UINT)) {
> - return false;
> - }
> -
> - if (!brw->format_supported_as_render_target[dst_mt->format])
> - return false;
> -
> - /* Source clipping shouldn't be necessary, since copytexsubimage (in
> - * src/mesa/main/teximage.c) calls _mesa_clip_copytexsubimage() which
> - * takes care of it.
> - *
> - * Destination clipping shouldn't be necessary since the restrictions on
> - * glCopyTexSubImage prevent the user from specifying a destination
> rectangle
> - * that falls outside the bounds of the destination texture.
> - * See error_check_subtexture_dimensions().
> - */
> -
> - int srcY1 = srcY0 + height;
> - int srcX1 = srcX0 + width;
> - int dstX1 = dstX0 + width;
> - int dstY1 = dstY0 + height;
> -
> - /* Account for the fact that in the system framebuffer, the origin is at
> - * the lower left.
> - */
> - bool mirror_y = false;
> - if (_mesa_is_winsys_fbo(ctx->ReadBuffer)) {
> - GLint tmp = src_rb->Height - srcY0;
> - srcY0 = src_rb->Height - srcY1;
> - srcY1 = tmp;
> - mirror_y = true;
> - }
> -
> - brw_blorp_blit_miptrees(brw,
> - src_mt, src_irb->mt_level, src_irb->mt_layer,
> - dst_mt, dst_image->Level, dst_image->Face + slice,
> - srcX0, srcY0, srcX1, srcY1,
> - dstX0, dstY0, dstX1, dstY1,
> - GL_NEAREST, false, mirror_y);
> -
> - /* If we're copying to a packed depth stencil texture and the source
> - * framebuffer has separate stencil, we need to also copy the stencil data
> - * over.
> - */
> - src_rb = ctx->ReadBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
> - if (_mesa_get_format_bits(dst_image->TexFormat, GL_STENCIL_BITS) > 0 &&
> - src_rb != NULL) {
> - src_irb = intel_renderbuffer(src_rb);
> - src_mt = src_irb->mt;
> -
> - if (src_mt->stencil_mt)
> - src_mt = src_mt->stencil_mt;
> - if (dst_mt->stencil_mt)
> - dst_mt = dst_mt->stencil_mt;
> -
> - if (src_mt != dst_mt) {
> - brw_blorp_blit_miptrees(brw,
> - src_mt, src_irb->mt_level,
> src_irb->mt_layer,
> - dst_mt, dst_image->Level,
> - dst_image->Face + slice,
> - srcX0, srcY0, srcX1, srcY1,
> - dstX0, dstY0, dstX1, dstY1,
> - GL_NEAREST, false, mirror_y);
> - }
> - }
> -
> - return true;
> -}
> -
> -
> GLbitfield
> brw_blorp_framebuffer(struct brw_context *brw,
> GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
> diff --git a/src/mesa/drivers/dri/i965/brw_context.h
> b/src/mesa/drivers/dri/i965/brw_context.h
> index ccbe9ea..5568fcb 100644
> --- a/src/mesa/drivers/dri/i965/brw_context.h
> +++ b/src/mesa/drivers/dri/i965/brw_context.h
> @@ -1710,15 +1710,6 @@ brw_blorp_framebuffer(struct brw_context *brw,
> GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
> GLbitfield mask, GLenum filter);
>
> -bool
> -brw_blorp_copytexsubimage(struct brw_context *brw,
> - struct gl_renderbuffer *src_rb,
> - struct gl_texture_image *dst_image,
> - int slice,
> - int srcX0, int srcY0,
> - int dstX0, int dstY0,
> - int width, int height);
> -
> /* gen6_multisample_state.c */
> unsigned
> gen6_determine_sample_mask(struct brw_context *brw);
> diff --git a/src/mesa/drivers/dri/i965/intel_tex_copy.c
> b/src/mesa/drivers/dri/i965/intel_tex_copy.c
> index ca0b5b9..082437a 100644
> --- a/src/mesa/drivers/dri/i965/intel_tex_copy.c
> +++ b/src/mesa/drivers/dri/i965/intel_tex_copy.c
> @@ -99,12 +99,7 @@ intelCopyTexSubImage(struct gl_context *ctx, GLuint dims,
> {
> struct brw_context *brw = brw_context(ctx);
>
> - /* Try BLORP first. It can handle almost everything. */
> - if (brw_blorp_copytexsubimage(brw, rb, texImage, slice, x, y,
> - xoffset, yoffset, width, height))
> - return;
> -
> - /* Next, try the BLT engine. */
> + /* Try the BLT engine. */
> if (intel_copy_texsubimage(brw,
> intel_texture_image(texImage),
> xoffset, yoffset, slice,
> @@ -112,8 +107,6 @@ intelCopyTexSubImage(struct gl_context *ctx, GLuint dims,
> return;
> }
>
> - /* Finally, fall back to meta. This will likely be slow. */
> - perf_debug("%s - fallback to swrast\n", __FUNCTION__);
> _mesa_meta_CopyTexSubImage(ctx, dims, texImage,
> xoffset, yoffset, slice,
> rb, x, y, width, height);
>
Sadly, this looks like a slight performance regression in Xonotic:
+--------------------------------------------------------------------------+
| + + + x |
| ++ + x x++x ++ * x x+*+ x x xx |
|+ + ++ x* ** +x+ *+*x+++ * * x*** x*xxx+x xxx xx +|
| |__________|_A_M_______A_M_|_______| |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 37 156.87213 159.97879 158.69534 158.58011 0.88627101
+ 37 155.08178 160.50746 157.95545 157.86242 0.99051368
Difference at 95.0% confidence
-0.717688 +/- 0.435487
-0.452571% +/- 0.274616%
(Student's t, pooled s = 0.939839)
After 37 runs my GPU hung, so I wasn't able to benchmark further.
signature.asc
Description: OpenPGP digital signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
