diff options
author | Eric Anholt <eric@anholt.net> | 2016-10-06 14:45:23 -0700 |
---|---|---|
committer | Eric Anholt <eric@anholt.net> | 2016-10-06 18:29:12 -0700 |
commit | cb328123fe2fda48add8a754b84b22e54f9c5912 (patch) | |
tree | 907a74e308eae112ab8274fa16f59be4e0090f9e /src | |
parent | bca9a58d04b556077c540984f2328eaa0ff26f15 (diff) |
vc4: Try to fix the HW-2116 workaround.
We were incrementing the count at the end of vc4_start_draw(), except that
that function returns immediately if we've already started drawing on this
batch. It also failed to count the statechanges from the GFXH-515
workaround.
This incidentally allows repeated glClear() to be coalesced, because the
fast clears aren't counted in draw_calls_queued any more. Fixes most of
the extra flushes in Processing, which emits glClear(Z|S); glClear(Z);
glClear(C) during its frame setup.
Improves performance of Processing's QuadRendering demo at 5000 quads by
3.33538% +/- 2.05846% (n=21 before, 15 after)
Diffstat (limited to 'src')
-rw-r--r-- | src/gallium/drivers/vc4/vc4_draw.c | 19 |
1 files changed, 10 insertions, 9 deletions
diff --git a/src/gallium/drivers/vc4/vc4_draw.c b/src/gallium/drivers/vc4/vc4_draw.c index 46c42f14f84..87c8a5f99c6 100644 --- a/src/gallium/drivers/vc4/vc4_draw.c +++ b/src/gallium/drivers/vc4/vc4_draw.c @@ -32,6 +32,8 @@ #include "vc4_context.h" #include "vc4_resource.h" +#define VC4_HW_2116_COUNT 0x1ef0 + static void vc4_get_draw_cl_space(struct vc4_job *job, int vert_count) { @@ -108,7 +110,6 @@ vc4_start_draw(struct vc4_context *vc4) VC4_PRIMITIVE_LIST_FORMAT_TYPE_TRIANGLES)); job->needs_flush = true; - job->draw_calls_queued++; job->draw_width = vc4->framebuffer.width; job->draw_height = vc4->framebuffer.height; @@ -264,12 +265,12 @@ vc4_emit_gl_shader_state(struct vc4_context *vc4, * tiles with VC4_PACKET_RETURN_FROM_LIST. */ static void -vc4_hw_2116_workaround(struct pipe_context *pctx) +vc4_hw_2116_workaround(struct pipe_context *pctx, int vert_count) { struct vc4_context *vc4 = vc4_context(pctx); struct vc4_job *job = vc4_get_job_for_fbo(vc4); - if (job->draw_calls_queued == 0x1ef0) { + if (job->draw_calls_queued + vert_count / 65535 >= VC4_HW_2116_COUNT) { perf_debug("Flushing batch due to HW-2116 workaround " "(too many draw calls per scene\n"); vc4_job_submit(vc4, job); @@ -294,7 +295,7 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info) vc4_predraw_check_textures(pctx, &vc4->verttex); vc4_predraw_check_textures(pctx, &vc4->fragtex); - vc4_hw_2116_workaround(pctx); + vc4_hw_2116_workaround(pctx, info->count); struct vc4_job *job = vc4_get_job_for_fbo(vc4); @@ -308,7 +309,6 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info) vc4_start_draw(vc4); vc4_update_compiled_shaders(vc4, info->mode); - uint32_t start_draw_calls_queued = job->draw_calls_queued; vc4_emit_state(pctx); if ((vc4->dirty & (VC4_DIRTY_VTXBUF | @@ -362,6 +362,7 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info) cl_u32(&bcl, info->count); cl_reloc(job, &job->bcl, &bcl, rsc->bo, offset); cl_u32(&bcl, vc4->max_index); + job->draw_calls_queued++; if (vc4->indexbuf.index_size == 4 || vc4->indexbuf.user_buffer) pipe_resource_reference(&prsc, NULL); @@ -430,6 +431,7 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info) cl_u8(&bcl, info->mode); cl_u32(&bcl, this_count); cl_u32(&bcl, start); + job->draw_calls_queued++; count -= step; extra_index_bias += start + step; @@ -438,11 +440,10 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info) } cl_end(&job->bcl, bcl); - /* No flushes of the job should have happened between when we started - * emitting state for our draw and when we just emitted our draw's - * primitives. + /* We shouldn't have tripped the HW_2116 bug with the GFXH-515 + * workaround. */ - assert(start_draw_calls_queued == job->draw_calls_queued); + assert(job->draw_calls_queued <= VC4_HW_2116_COUNT); if (vc4->zsa && vc4->framebuffer.zsbuf) { struct vc4_resource *rsc = |