diff options
| author | Marek Olšák <[email protected]> | 2017-08-04 17:26:58 +0200 |
|---|---|---|
| committer | Marek Olšák <[email protected]> | 2017-08-07 21:12:24 +0200 |
| commit | c441999b7abbe5e97db0e01a3a8276d291f2f4ca (patch) | |
| tree | 0449a8bf5313fa6afaa8e386108bcc9817e5145f /src/gallium | |
| parent | e887c68bd23fa6e7421279286842f15746eb5e28 (diff) | |
radeonsi: prefetch VBO descriptors after the first VGT shader
Tested-by: Dieter Nützel <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Diffstat (limited to 'src/gallium')
| -rw-r--r-- | src/gallium/drivers/radeonsi/si_cp_dma.c | 80 |
1 files changed, 63 insertions, 17 deletions
diff --git a/src/gallium/drivers/radeonsi/si_cp_dma.c b/src/gallium/drivers/radeonsi/si_cp_dma.c index 21202b3d5b5..91a6aff4c12 100644 --- a/src/gallium/drivers/radeonsi/si_cp_dma.c +++ b/src/gallium/drivers/radeonsi/si_cp_dma.c @@ -448,27 +448,73 @@ static void cik_prefetch_shader_async(struct si_context *sctx, cik_prefetch_TC_L2_async(sctx, bo, 0, bo->width0); } +static void cik_prefetch_VBO_descriptors(struct si_context *sctx) +{ + cik_prefetch_TC_L2_async(sctx, &sctx->vertex_buffers.buffer->b.b, + sctx->vertex_buffers.buffer_offset, + sctx->vertex_elements->desc_list_byte_size); +} + void cik_emit_prefetch_L2(struct si_context *sctx) { /* Prefetch shaders and VBO descriptors to TC L2. */ - if (sctx->prefetch_L2_mask & SI_PREFETCH_LS) - cik_prefetch_shader_async(sctx, sctx->queued.named.ls); - if (sctx->prefetch_L2_mask & SI_PREFETCH_HS) - cik_prefetch_shader_async(sctx, sctx->queued.named.hs); - if (sctx->prefetch_L2_mask & SI_PREFETCH_ES) - cik_prefetch_shader_async(sctx, sctx->queued.named.es); - if (sctx->prefetch_L2_mask & SI_PREFETCH_GS) - cik_prefetch_shader_async(sctx, sctx->queued.named.gs); - if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) - cik_prefetch_shader_async(sctx, sctx->queued.named.vs); - - /* Vertex buffer descriptors are uploaded uncached, so prefetch - * them right after the VS binary. */ - if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) { - cik_prefetch_TC_L2_async(sctx, &sctx->vertex_buffers.buffer->b.b, - sctx->vertex_buffers.buffer_offset, - sctx->vertex_elements->desc_list_byte_size); + if (sctx->b.chip_class >= GFX9) { + /* Choose the right spot for the VBO prefetch. */ + if (sctx->tes_shader.cso) { + if (sctx->prefetch_L2_mask & SI_PREFETCH_HS) + cik_prefetch_shader_async(sctx, sctx->queued.named.hs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) + cik_prefetch_VBO_descriptors(sctx); + if (sctx->prefetch_L2_mask & SI_PREFETCH_GS) + cik_prefetch_shader_async(sctx, sctx->queued.named.gs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) + cik_prefetch_shader_async(sctx, sctx->queued.named.vs); + } else if (sctx->gs_shader.cso) { + if (sctx->prefetch_L2_mask & SI_PREFETCH_GS) + cik_prefetch_shader_async(sctx, sctx->queued.named.gs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) + cik_prefetch_VBO_descriptors(sctx); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) + cik_prefetch_shader_async(sctx, sctx->queued.named.vs); + } else { + if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) + cik_prefetch_shader_async(sctx, sctx->queued.named.vs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) + cik_prefetch_VBO_descriptors(sctx); + } + } else { + /* SI-CI-VI */ + /* Choose the right spot for the VBO prefetch. */ + if (sctx->tes_shader.cso) { + if (sctx->prefetch_L2_mask & SI_PREFETCH_LS) + cik_prefetch_shader_async(sctx, sctx->queued.named.ls); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) + cik_prefetch_VBO_descriptors(sctx); + if (sctx->prefetch_L2_mask & SI_PREFETCH_HS) + cik_prefetch_shader_async(sctx, sctx->queued.named.hs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_ES) + cik_prefetch_shader_async(sctx, sctx->queued.named.es); + if (sctx->prefetch_L2_mask & SI_PREFETCH_GS) + cik_prefetch_shader_async(sctx, sctx->queued.named.gs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) + cik_prefetch_shader_async(sctx, sctx->queued.named.vs); + } else if (sctx->gs_shader.cso) { + if (sctx->prefetch_L2_mask & SI_PREFETCH_ES) + cik_prefetch_shader_async(sctx, sctx->queued.named.es); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) + cik_prefetch_VBO_descriptors(sctx); + if (sctx->prefetch_L2_mask & SI_PREFETCH_GS) + cik_prefetch_shader_async(sctx, sctx->queued.named.gs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) + cik_prefetch_shader_async(sctx, sctx->queued.named.vs); + } else { + if (sctx->prefetch_L2_mask & SI_PREFETCH_VS) + cik_prefetch_shader_async(sctx, sctx->queued.named.vs); + if (sctx->prefetch_L2_mask & SI_PREFETCH_VBO_DESCRIPTORS) + cik_prefetch_VBO_descriptors(sctx); + } } + if (sctx->prefetch_L2_mask & SI_PREFETCH_PS) cik_prefetch_shader_async(sctx, sctx->queued.named.ps); |