aco: combine two sgprs into a VALU if they're the same

This was supposed to be done before but it wasn't done correctly and
everywhere.

pipeline-db (Navi):
Totals from affected shaders:
SGPRS: 784680 -> 786128 (0.18 %)
VGPRS: 574012 -> 573892 (-0.02 %)
Spilled SGPRs: 461 -> 461 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Code Size: 45477088 -> 45478172 (0.00 %) bytes
Max Waves: 81294 -> 81277 (-0.02 %)
Instructions: 8657970 -> 8622483 (-0.41 %)

pipeline-db (Vega):
Totals from affected shaders:
SGPRS: 780664 -> 782072 (0.18 %)
VGPRS: 573880 -> 573760 (-0.02 %)
Spilled SGPRs: 629 -> 629 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Code Size: 45445244 -> 45448340 (0.01 %) bytes
Max Waves: 81178 -> 81161 (-0.02 %)
Instructions: 8649902 -> 8614918 (-0.40 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2883>

1 files changed, 11 insertions, 5 deletions

diff --git a/src/amd/compiler/aco_optimizer.cpp b/src/amd/compiler/aco_optimizer.cpp
index e7a2da1bcb1..a589fbbe73e 100644
--- a/src/amd/compiler/aco_optimizer.cpp
+++ b/src/amd/compiler/aco_optimizer.cpp
@@ -1207,13 +1207,16 @@ bool combine_ordering_test(opt_ctx &ctx, aco_ptr<Instruction>& instr)
       Temp op1 = op_instr[i]->operands[1].getTemp();
       if (original_temp_id(ctx, op0) != original_temp_id(ctx, op1))
          return false;
-      /* shouldn't happen yet, but best to be safe */
-      if (op1.type() != RegType::vgpr)
-         return false;
 
       op[i] = op1;
    }
 
+   if (op[1].type() == RegType::sgpr)
+      std::swap(op[0], op[1]);
+   //TODO: we can use two different SGPRs on GFX10
+   if (op[0].type() == RegType::sgpr && op[1].type() == RegType::sgpr)
+      return false;
+
    ctx.uses[op[0].id()]++;
    ctx.uses[op[1].id()]++;
    decrease_uses(ctx, op_instr[0]);
@@ -1919,7 +1922,7 @@ void apply_sgprs(opt_ctx &ctx, aco_ptr<Instruction>& instr)
 
    /* we can have two sgprs on one instruction if it is the same sgpr! */
    } else if (sgpr_info_id != 0 &&
-              sgpr_ssa_id == sgpr_info_id &&
+              sgpr_ssa_id == ctx.info[sgpr_info_id].temp.id() &&
               ctx.uses[sgpr_info_id] == 1 &&
               can_use_VOP3(instr)) {
       to_VOP3(ctx, instr);
@@ -2133,6 +2136,7 @@ void combine_instruction(opt_ctx &ctx, Block& block, aco_ptr<Instruction>& instr
          bool clamp = false;
          bool need_vop3 = false;
          int num_sgpr = 0;
+         unsigned cur_sgpr = 0;
          op[0] = mul_instr->operands[0];
          op[1] = mul_instr->operands[1];
          op[2] = instr->operands[add_op_idx];
@@ -2140,8 +2144,10 @@ void combine_instruction(opt_ctx &ctx, Block& block, aco_ptr<Instruction>& instr
          {
             if (op[i].isLiteral())
                return;
-            if (op[i].isTemp() && op[i].getTemp().type() == RegType::sgpr)
+            if (op[i].isTemp() && op[i].getTemp().type() == RegType::sgpr && op[i].tempId() != cur_sgpr) {
                num_sgpr++;
+               cur_sgpr = op[i].tempId();
+            }
             if (!(i == 0 || (op[i].isTemp() && op[i].getTemp().type() == RegType::vgpr)))
                need_vop3 = true;
          }