[llvm-branch-commits] [llvm] [AMDGPU][SIInsertWaitcnts][NFC] Introduce Counter::getWait() (PR #193374)

[via llvm-branch-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-amdgpu

Author: vporpo (vporpo)

<details>
<summary>Changes</summary>

This patch moves the logic that calculates the amount of wait for a specific score into the Counter class.

This helps us remove some raw UB and LB accesses.


---
Previous PRs:

0. https://github.com/llvm/llvm-project/pull/193368
1. https://github.com/llvm/llvm-project/pull/193369



---
Full diff: https://github.com/llvm/llvm-project/pull/193374.diff


1 Files Affected:

- (modified) llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp (+8-7) 


``````````diff
diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index e43181a76fc5c..cfa6ec691002a 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -783,6 +783,9 @@ class WaitcntBrackets {
           : CntT(CntT), Limits(&Limits) {}
       /// \returns the count of outstanding instrs tracked by this counter.
       unsigned getCount() const { return UB - LB; }
+      /// \returns how much we should wait for the instruction corresponding to
+      /// \p Score to complete, assuming in-order completion.
+      unsigned getWait(unsigned Score) const { return UB - Score; }
       // TODO: Make private: we should not provide raw access to the internals.
       void setLB(unsigned NewLB) { LB = NewLB; }
       // TODO: Make private: we should not provide raw access to the internals.
@@ -910,7 +913,7 @@ class WaitcntBrackets {
   }
 
   unsigned getPendingGDSWait() const {
-    return std::min(getScoreUB(AMDGPU::DS_CNT) - LastGDS,
+    return std::min(Counters[AMDGPU::DS_CNT].getWait(LastGDS),
                     getWaitCountMax(Context->getLimits(), AMDGPU::DS_CNT) - 1);
   }
 
@@ -1612,11 +1615,8 @@ void WaitcntBrackets::purgeEmptyTrackingData() {
 void WaitcntBrackets::determineWaitForScore(AMDGPU::InstCounterType T,
                                             unsigned ScoreToWait,
                                             AMDGPU::Waitcnt &Wait) const {
-  const unsigned LB = getScoreLB(T);
-  const unsigned UB = getScoreUB(T);
-
   // If the score falls within the bracket, we need a waitcnt.
-  if ((UB >= ScoreToWait) && (ScoreToWait > LB)) {
+  if (Counters[T].contains(ScoreToWait)) {
     if ((T == AMDGPU::LOAD_CNT || T == AMDGPU::DS_CNT) && hasPendingFlat() &&
         !Context->ST.hasFlatLgkmVMemCountInOrder()) {
       // If there is a pending FLAT operation, and this is a VMem or LGKM
@@ -1631,8 +1631,9 @@ void WaitcntBrackets::determineWaitForScore(AMDGPU::InstCounterType T,
     } else {
       // If a counter has been maxed out avoid overflow by waiting for
       // MAX(CounterType) - 1 instead.
-      unsigned NeededWait = std::min(
-          UB - ScoreToWait, getWaitCountMax(Context->getLimits(), T) - 1);
+      unsigned NeededWait =
+          std::min(Counters[T].getWait(ScoreToWait),
+                   getWaitCountMax(Context->getLimits(), T) - 1);
       addWait(Wait, T, NeededWait);
     }
   }

``````````

</details>


https://github.com/llvm/llvm-project/pull/193374