From b7af10449bc237b54d27833e65a25ce80dab8167 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Daniel=20Sch=C3=BCrmann?= <daniel@schuermann.dev>
Date: Fri, 10 Sep 2021 16:00:25 +0200
Subject: [PATCH] aco/ra: create nested affinities for loop header phis

Totals from 875 (0.58% of 150170) affected shaders: (GFX10.3)
CodeSize: 6084528 -> 6066628 (-0.29%); split: -0.32%, +0.02%
Instrs: 1136497 -> 1133565 (-0.26%); split: -0.28%, +0.02%
Latency: 23355051 -> 22952592 (-1.72%); split: -1.83%, +0.10%
InvThroughput: 13028151 -> 12859628 (-1.29%); split: -1.38%, +0.09%
Copies: 85673 -> 82790 (-3.37%); split: -3.62%, +0.26%
Branches: 25049 -> 25098 (+0.20%); split: -0.08%, +0.28%

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836>
---
 src/amd/compiler/aco_register_allocation.cpp | 36 ++++++++++++++++++--
 1 file changed, 34 insertions(+), 2 deletions(-)

diff --git a/src/amd/compiler/aco_register_allocation.cpp b/src/amd/compiler/aco_register_allocation.cpp
index 829104cf6ed..a8d4668426d 100644
--- a/src/amd/compiler/aco_register_allocation.cpp
+++ b/src/amd/compiler/aco_register_allocation.cpp
@@ -2386,14 +2386,46 @@ get_affinities(ra_ctx& ctx, std::vector<IDSet>& live_out_per_block)
             affinity_related = &phi_ressources.back();
          }
 
-         for (unsigned i = 0; i < instr->operands.size(); i++) {
-            const Operand& op = instr->operands[i];
+         for (const Operand& op : instr->operands) {
             if (op.isTemp() && op.isKill() && op.regClass() == instr->definitions[0].regClass()) {
                affinity_related->emplace_back(op.getTemp());
+               if (block.kind & block_kind_loop_header)
+                  continue;
                temp_to_phi_ressources[op.tempId()] = index;
             }
          }
       }
+
+      /* visit the loop header phis first in order to create nested affinities */
+      if (block.kind & block_kind_loop_exit) {
+         /* find loop header */
+         auto header_rit = block_rit;
+         while ((header_rit + 1)->loop_nest_depth > block.loop_nest_depth)
+            header_rit++;
+
+         for (aco_ptr<Instruction>& phi : header_rit->instructions) {
+            if (!is_phi(phi))
+               break;
+            if (phi->definitions[0].isKill() || phi->definitions[0].isFixed())
+               continue;
+
+            /* create an (empty) merge-set for the phi-related variables */
+            auto it = temp_to_phi_ressources.find(phi->definitions[0].tempId());
+            unsigned index = phi_ressources.size();
+            if (it == temp_to_phi_ressources.end()) {
+               temp_to_phi_ressources[phi->definitions[0].tempId()] = index;
+               phi_ressources.emplace_back(std::vector<Temp>{phi->definitions[0].getTemp()});
+            } else {
+               index = it->second;
+            }
+            for (unsigned i = 1; i < phi->operands.size(); i++) {
+               const Operand& op = phi->operands[i];
+               if (op.isTemp() && op.isKill() && op.regClass() == phi->definitions[0].regClass()) {
+                  temp_to_phi_ressources[op.tempId()] = index;
+               }
+            }
+         }
+      }
    }
    /* create affinities */
    for (std::vector<Temp>& vec : phi_ressources) {