radeonsi: add a debug option to enable NGG culling for tessellation

Acked-by: Pierre-Eric Pelloux-Prayer <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5524>

4 files changed, 15 insertions, 8 deletions

diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index 27b3e9200a5..46f2ff72492 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -94,7 +94,8 @@ static const struct debug_named_value debug_options[] = {
    /* 3D engine options: */
    {"nogfx", DBG(NO_GFX), "Disable graphics. Only multimedia compute paths can be used."},
    {"nongg", DBG(NO_NGG), "Disable NGG and use the legacy pipeline."},
-   {"nggc", DBG(ALWAYS_NGG_CULLING), "Always use NGG culling even when it can hurt."},
+   {"nggc", DBG(ALWAYS_NGG_CULLING_ALL), "Always use NGG culling even when it can hurt."},
+   {"nggctess", DBG(ALWAYS_NGG_CULLING_TESS), "Always use NGG culling for tessellation."},
    {"nonggc", DBG(NO_NGG_CULLING), "Disable NGG culling."},
    {"alwayspd", DBG(ALWAYS_PD), "Always enable the primitive discard compute shader."},
    {"pd", DBG(PD), "Enable the primitive discard compute shader for large draw calls."},
@@ -1138,8 +1139,10 @@ static struct pipe_screen *radeonsi_screen_create_impl(struct radeon_winsys *ws,
    sscreen->use_ngg = sscreen->info.chip_class >= GFX10 && sscreen->info.family != CHIP_NAVI14 &&
                       !(sscreen->debug_flags & DBG(NO_NGG));
    sscreen->use_ngg_culling = sscreen->use_ngg && !(sscreen->debug_flags & DBG(NO_NGG_CULLING));
-   sscreen->always_use_ngg_culling =
-      sscreen->use_ngg_culling && sscreen->debug_flags & DBG(ALWAYS_NGG_CULLING);
+   sscreen->always_use_ngg_culling_all =
+      sscreen->use_ngg_culling && sscreen->debug_flags & DBG(ALWAYS_NGG_CULLING_ALL);
+   sscreen->always_use_ngg_culling_tess =
+      sscreen->use_ngg_culling && sscreen->debug_flags & DBG(ALWAYS_NGG_CULLING_TESS);
    sscreen->use_ngg_streamout = false;
 
    /* Only enable primitive binning on APUs by default. */
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 9ad14cab96a..467538c76cd 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -188,7 +188,8 @@ enum
    /* 3D engine options: */
    DBG_NO_GFX,
    DBG_NO_NGG,
-   DBG_ALWAYS_NGG_CULLING,
+   DBG_ALWAYS_NGG_CULLING_ALL,
+   DBG_ALWAYS_NGG_CULLING_TESS,
    DBG_NO_NGG_CULLING,
    DBG_ALWAYS_PD,
    DBG_PD,
@@ -517,7 +518,8 @@ struct si_screen {
    bool llvm_has_working_vgpr_indexing;
    bool use_ngg;
    bool use_ngg_culling;
-   bool always_use_ngg_culling;
+   bool always_use_ngg_culling_all;
+   bool always_use_ngg_culling_tess;
    bool use_ngg_streamout;
 
    struct {
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index c7223d94122..b123e610ffe 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -1864,7 +1864,8 @@ static void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *i
    /* Update NGG culling settings. */
    if (sctx->ngg && !dispatch_prim_discard_cs && rast_prim == PIPE_PRIM_TRIANGLES &&
        !sctx->gs_shader.cso && /* GS doesn't support NGG culling. */
-       (sctx->screen->always_use_ngg_culling ||
+       (sctx->screen->always_use_ngg_culling_all ||
+        (sctx->tes_shader.cso && sctx->screen->always_use_ngg_culling_tess) ||
         /* At least 1024 non-indexed vertices (8 subgroups) are needed
          * per draw call (no TES/GS) to enable NGG culling.
          */
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index b4e95b78549..d1a1aa725cd 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -2752,9 +2752,10 @@ static void *si_create_shader_selector(struct pipe_context *ctx,
       sscreen->info.chip_class >= GFX10 &&
       sscreen->info.has_dedicated_vram &&
       sscreen->use_ngg_culling &&
-      /* Disallow TES by default, because TessMark results are mixed. */
       (sel->type == PIPE_SHADER_VERTEX ||
-       (sscreen->always_use_ngg_culling && sel->type == PIPE_SHADER_TESS_EVAL)) &&
+       (sel->type == PIPE_SHADER_TESS_EVAL &&
+        (sscreen->always_use_ngg_culling_all ||
+         sscreen->always_use_ngg_culling_tess))) &&
       sel->info.writes_position &&
       !sel->info.writes_viewport_index && /* cull only against viewport 0 */
       !sel->info.writes_memory && !sel->so.num_outputs &&