diff options
| author | Marek Olšák <marek.olsak@amd.com> | 2021-06-01 01:09:57 -0400 |
|---|---|---|
| committer | Marek Olšák <marek.olsak@amd.com> | 2021-06-20 01:22:01 -0400 |
| commit | 12d2df15f178591d45193b6cc3e093281e2ea0aa (patch) | |
| tree | 3ba1f522c3aa34d1fbfc814fea5e494da34dd8fb /src/gallium | |
| parent | 18055726948f24670a31bb421cc9fa07140c1a5e (diff) | |
ac/llvm: add a callback to ac_cull_triangle to generate code in inner-most block
This will reduce jumps in culling code.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11486>
Diffstat (limited to 'src/gallium')
| -rw-r--r-- | src/gallium/drivers/radeonsi/gfx10_shader_ngg.c | 2 | ||||
| -rw-r--r-- | src/gallium/drivers/radeonsi/si_compute_prim_discard.c | 3 |
2 files changed, 3 insertions, 2 deletions
diff --git a/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c b/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c index d72c72e748a..0bcd8b4b3d9 100644 --- a/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c +++ b/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c @@ -985,7 +985,7 @@ void gfx10_emit_ngg_culling_epilogue(struct ac_shader_abi *abi, unsigned max_out /* Tell ES threads whether their vertex survived. */ ac_build_ifcc(&ctx->ac, ac_cull_triangle(&ctx->ac, pos, ctx->ac.i1true, vp_scale, vp_translate, - small_prim_precision, &options), + small_prim_precision, &options, NULL, NULL), 16003); { LLVMBuildStore(builder, ctx->ac.i32_1, gs_accepted); diff --git a/src/gallium/drivers/radeonsi/si_compute_prim_discard.c b/src/gallium/drivers/radeonsi/si_compute_prim_discard.c index c830a857f66..a75086606f4 100644 --- a/src/gallium/drivers/radeonsi/si_compute_prim_discard.c +++ b/src/gallium/drivers/radeonsi/si_compute_prim_discard.c @@ -673,7 +673,8 @@ void si_build_prim_discard_compute_shader(struct si_shader_context *ctx) LLVMValueRef accepted = ac_cull_triangle(&ctx->ac, pos, prim_restart_accepted, vp_scale, vp_translate, - ac_get_arg(&ctx->ac, param_smallprim_precision), &options); + ac_get_arg(&ctx->ac, param_smallprim_precision), &options, + NULL, NULL); ac_build_optimization_barrier(&ctx->ac, &accepted, false); LLVMValueRef accepted_threadmask = ac_get_i1_sgpr_mask(&ctx->ac, accepted); |