From: Timur Kristóf <timur.kristof@gmail.com>
Date: Sat, 7 Mar 2020 00:51:39 +0000 (+0100)
Subject: aco: Only write TCS outputs to LDS when they are read by the TCS.
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=9b36d8c23ac405be98a0e83ace1bea4d7a95df82;p=mesa.git

aco: Only write TCS outputs to LDS when they are read by the TCS.

Note that tess factors are always read at the end of the shader,
so those are still always saved to LDS.

Totals from affected shaders:
VGPRS: 25244 -> 25164 (-0.32 %)
Code Size: 1768268 -> 1690804 (-4.38 %) bytes
Max Waves: 4947 -> 4953 (0.12 %)

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3964>
---

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index c363edb602d..386616f52ef 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -3334,6 +3334,29 @@ bool should_write_tcs_patch_output_to_vmem(isel_context *ctx, nir_intrinsic_inst
           (off != (tess_index_outer * 16u));
 }
 
+bool should_write_tcs_patch_output_to_lds(isel_context *ctx, nir_intrinsic_instr *instr, bool per_vertex)
+{
+   unsigned off = nir_intrinsic_base(instr) * 4u;
+   nir_src *off_src = nir_get_io_offset_src(instr);
+
+   /* Indirect offset, we can't be sure if this is read or not, always write to LDS */
+   if (!nir_src_is_const(*off_src))
+      return true;
+
+   off += nir_src_as_uint(*off_src) * 16u;
+
+   uint64_t out_rd = per_vertex
+                     ? ctx->shader->info.outputs_read
+                     : ctx->shader->info.patch_outputs_read;
+   while (out_rd) {
+      unsigned slot = u_bit_scan64(&out_rd) + (per_vertex ? 0 : VARYING_SLOT_PATCH0);
+      if (off == shader_io_get_unique_index((gl_varying_slot) slot) * 16u)
+         return true;
+   }
+
+   return false;
+}
+
 void visit_store_tcs_output(isel_context *ctx, nir_intrinsic_instr *instr, bool per_vertex)
 {
    assert(ctx->stage == tess_control_hs || ctx->stage == vertex_tess_control_hs);
@@ -3347,8 +3370,8 @@ void visit_store_tcs_output(isel_context *ctx, nir_intrinsic_instr *instr, bool
 
    /* Only write to VMEM if the output is per-vertex or it's per-patch non tess factor */
    bool write_to_vmem = per_vertex || should_write_tcs_patch_output_to_vmem(ctx, instr);
-   /* TODO: Only write to LDS if the output is read by the shader, or it's per-patch tess factor */
-   bool write_to_lds = true;
+   /* Only write to LDS if the output is read by the shader, or it's per-patch tess factor */
+   bool write_to_lds = !write_to_vmem || should_write_tcs_patch_output_to_lds(ctx, instr, per_vertex);
 
    if (write_to_vmem) {
       std::pair<Temp, unsigned> vmem_offs = per_vertex