radeonsi: prefetch VBO descriptors after the first VGT shader

Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

diff --git a/src/gallium/drivers/radeonsi/si_cp_dma.c b/src/gallium/drivers/radeonsi/si_cp_dma.c
index 21202b3d5b5a9d49fc73a896dbfdcf1b5230600e..91a6aff4c12d00fdcdf9c033704ff31ca5087324 100644 (file)
--- 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);