From ee0f7848581c813eb16d31a0da552c7797bfac13 Mon Sep 17 00:00:00 2001
From: Natalie Vock <natalie.vock@gmx.de>
Date: Sat, 5 Apr 2025 13:04:19 +0200
Subject: [PATCH] aco/ra: Don't consider precolored ops/defs in get_reg_impl

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34273>
---
 src/amd/compiler/aco_register_allocation.cpp | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/src/amd/compiler/aco_register_allocation.cpp b/src/amd/compiler/aco_register_allocation.cpp
index 70638fe6636..0a9083f7a11 100644
--- a/src/amd/compiler/aco_register_allocation.cpp
+++ b/src/amd/compiler/aco_register_allocation.cpp
@@ -1280,8 +1280,15 @@ get_reg_impl(ra_ctx& ctx, const RegisterFile& reg_file, std::vector<parallelcopy
    /* mark and count killed operands */
    unsigned killed_ops = 0;
    std::bitset<256> is_killed_operand; /* per-register */
+   std::bitset<256> is_precolored;     /* per-register */
    for (unsigned j = 0; !is_phi(instr) && j < instr->operands.size(); j++) {
       Operand& op = instr->operands[j];
+      if (op.isTemp() && op.isPrecolored() && !op.isFirstKillBeforeDef() &&
+          bounds.contains(op.physReg())) {
+         for (unsigned i = 0; i < op.size(); ++i) {
+            is_precolored[(op.physReg() & 0xff) + i] = true;
+         }
+      }
       if (op.isTemp() && op.isFirstKillBeforeDef() && bounds.contains(op.physReg()) &&
           !reg_file.test(PhysReg{op.physReg().reg()}, align(op.bytes() + op.physReg().byte(), 4))) {
          assert(op.isFixed());
@@ -1293,6 +1300,14 @@ get_reg_impl(ra_ctx& ctx, const RegisterFile& reg_file, std::vector<parallelcopy
          killed_ops += op.getTemp().size();
       }
    }
+   for (unsigned j = 0; !is_phi(instr) && j < instr->definitions.size(); j++) {
+      Definition& def = instr->definitions[j];
+      if (def.isTemp() && def.isPrecolored() && bounds.contains(def.physReg())) {
+         for (unsigned i = 0; i < def.size(); ++i) {
+            is_precolored[(def.physReg() & 0xff) + i] = true;
+         }
+      }
+   }
 
    assert((regs_free + ctx.num_linear_vgprs) >= size);
 
@@ -1337,6 +1352,10 @@ get_reg_impl(ra_ctx& ctx, const RegisterFile& reg_file, std::vector<parallelcopy
             }
             continue;
          }
+         if (is_precolored[j & 0xFF]) {
+            found = false;
+            break;
+         }
 
          if (reg_file[j] == 0 || reg_file[j] == last_var)
             continue;