[llvm] [AMDGPU] Add DS loop waitcnt optimization for GFX12+ (PR #172728)

[via llvm-commits]

github-actions[bot] wrote:

<!--LLVM CODE FORMAT COMMENT: {clang-format}-->


:warning: C/C++ code formatter, clang-format found issues in your code. :warning:

<details>
<summary>
You can test this locally with the following command:
</summary>

``````````bash
git-clang-format --diff origin/main HEAD --extensions cpp -- llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp --diff_from_common_commit
``````````

:warning:
The reproduction instructions above might return results for more than one PR
in a stack if you are using a stacked PR workflow. You can limit the results by
changing `origin/main` to the base branch/commit you want to compare against.
:warning:

</details>

<details>
<summary>
View the diff from clang-format here.
</summary>

``````````diff
diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index 6b0bcfc6b..4de55653e 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -2702,8 +2702,9 @@ bool SIInsertWaitcnts::insertWaitcntInBlock(MachineFunction &MF,
 }
 
 // Return flags indicating which counters should be flushed in the preheader.
-PreheaderFlushFlags SIInsertWaitcnts::isPreheaderToFlush(
-    MachineBasicBlock &MBB, const WaitcntBrackets &ScoreBrackets) {
+PreheaderFlushFlags
+SIInsertWaitcnts::isPreheaderToFlush(MachineBasicBlock &MBB,
+                                     const WaitcntBrackets &ScoreBrackets) {
   auto [Iterator, IsInserted] =
       PreheadersToFlush.try_emplace(&MBB, PreheaderFlushFlags());
   if (!IsInserted)
@@ -2830,14 +2831,14 @@ SIInsertWaitcnts::getPreheaderFlushFlags(MachineLoop *ML,
           // Check if loaded outside the loop via DS (not VMEM/FLAT)
           // Only consider it a DS load if there's no pending VMEM load for
           // this register, since FLAT can set both counters.
-          bool HasPendingVMEM = Brackets.getVMemScore(ID, LOAD_CNT) >
-                                    Brackets.getScoreLB(LOAD_CNT) ||
-                                Brackets.getVMemScore(ID, SAMPLE_CNT) >
-                                    Brackets.getScoreLB(SAMPLE_CNT) ||
-                                Brackets.getVMemScore(ID, BVH_CNT) >
-                                    Brackets.getScoreLB(BVH_CNT);
-          if (!HasPendingVMEM && Brackets.getVMemScore(ID, DS_CNT) >
-                                     Brackets.getScoreLB(DS_CNT)) {
+          bool HasPendingVMEM =
+              Brackets.getVMemScore(ID, LOAD_CNT) >
+                  Brackets.getScoreLB(LOAD_CNT) ||
+              Brackets.getVMemScore(ID, SAMPLE_CNT) >
+                  Brackets.getScoreLB(SAMPLE_CNT) ||
+              Brackets.getVMemScore(ID, BVH_CNT) > Brackets.getScoreLB(BVH_CNT);
+          if (!HasPendingVMEM &&
+              Brackets.getVMemScore(ID, DS_CNT) > Brackets.getScoreLB(DS_CNT)) {
             UsesVgprLoadedOutsideDS = true;
           }
         }

``````````

</details>


https://github.com/llvm/llvm-project/pull/172728