[AMDGPU] Inline two helpers in SIInsertWaitcnts. NFC.

[jayfoad]

No description provided.

[jayfoad]

@RyanRio

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Jay Foad (jayfoad)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/174557.diff

1 Files Affected:

• (modified) llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp (+9-22)

diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index 10138f59ab1b4..8207bd46ea9f7 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -662,8 +662,6 @@ class WaitcntBrackets {
   void simplifyWaitcnt(const AMDGPU::Waitcnt &CheckWait,
                        AMDGPU::Waitcnt &UpdateWait) const;
   void simplifyWaitcnt(InstCounterType T, unsigned &Count) const;
-  bool hasRedundantXCntWithKmCnt(const AMDGPU::Waitcnt &Wait) const;
-  bool canOptimizeXCntWithLoadCnt(const AMDGPU::Waitcnt &Wait) const;
   void simplifyXcnt(const AMDGPU::Waitcnt &CheckWait,
                     AMDGPU::Waitcnt &UpdateWait) const;
 
@@ -1354,23 +1352,6 @@ void WaitcntBrackets::applyWaitcnt(InstCounterType T, unsigned Count) {
   }
 }
 
-bool WaitcntBrackets::hasRedundantXCntWithKmCnt(
-    const AMDGPU::Waitcnt &Wait) const {
-  // Wait on XCNT is redundant if we are already waiting for a load to complete.
-  // SMEM can return out of order, so only omit XCNT wait if we are waiting till
-  // zero.
-  return Wait.KmCnt == 0 && hasPendingEvent(SMEM_GROUP);
-}
-
-bool WaitcntBrackets::canOptimizeXCntWithLoadCnt(
-    const AMDGPU::Waitcnt &Wait) const {
-  // If we have pending store we cannot optimize XCnt because we do not wait for
-  // stores. VMEM loads retun in order, so if we only have loads XCnt is
-  // decremented to the same number as LOADCnt.
-  return Wait.LoadCnt != ~0u && hasPendingEvent(VMEM_GROUP) &&
-         !hasPendingEvent(STORE_CNT);
-}
-
 void WaitcntBrackets::simplifyXcnt(const AMDGPU::Waitcnt &CheckWait,
                                    AMDGPU::Waitcnt &UpdateWait) const {
   // Try to simplify xcnt further by checking for joint kmcnt and loadcnt
@@ -1378,10 +1359,16 @@ void WaitcntBrackets::simplifyXcnt(const AMDGPU::Waitcnt &CheckWait,
   // be pending SMEM and VMEM events active at the same time.
   // In such cases, only clear one active event at a time.
   // TODO: Revisit xcnt optimizations for gfx1250.
-  if (hasRedundantXCntWithKmCnt(CheckWait))
+  // Wait on XCNT is redundant if we are already waiting for a load to complete.
+  // SMEM can return out of order, so only omit XCNT wait if we are waiting till
+  // zero.
+  if (CheckWait.KmCnt == 0 && hasPendingEvent(SMEM_GROUP))
     UpdateWait.XCnt = ~0u;
-  if (canOptimizeXCntWithLoadCnt(CheckWait) &&
-      CheckWait.XCnt >= CheckWait.LoadCnt)
+  // If we have pending store we cannot optimize XCnt because we do not wait for
+  // stores. VMEM loads retun in order, so if we only have loads XCnt is
+  // decremented to the same number as LOADCnt.
+  if (CheckWait.LoadCnt != ~0u && hasPendingEvent(VMEM_GROUP) &&
+      !hasPendingEvent(STORE_CNT) && CheckWait.XCnt >= CheckWait.LoadCnt)
     UpdateWait.XCnt = ~0u;
   simplifyWaitcnt(X_CNT, UpdateWait.XCnt);
 }

[jayfoad]

The patch was created by an LLM.

[github-actions]

🪟 Windows x64 Test Results

• 129123 tests passed
• 2842 tests skipped
• 1 test failed

Failed Tests

(click on a test name to see its output)

LLVM

LLVM.CodeGen/RISCV/riscv-macho.ll

Exit Code: 3221225477

Command Output (stdout):
--
# RUN: at line 1
c:\_work\llvm-project\llvm-project\build\bin\llc.exe -mtriple=riscv32-apple-macho C:\_work\llvm-project\llvm-project\llvm\test\CodeGen\RISCV\riscv-macho.ll -o - | c:\_work\llvm-project\llvm-project\build\bin\filecheck.exe C:\_work\llvm-project\llvm-project\llvm\test\CodeGen\RISCV\riscv-macho.ll
# executed command: 'c:\_work\llvm-project\llvm-project\build\bin\llc.exe' -mtriple=riscv32-apple-macho 'C:\_work\llvm-project\llvm-project\llvm\test\CodeGen\RISCV\riscv-macho.ll' -o -
# note: command had no output on stdout or stderr
# executed command: 'c:\_work\llvm-project\llvm-project\build\bin\filecheck.exe' 'C:\_work\llvm-project\llvm-project\llvm\test\CodeGen\RISCV\riscv-macho.ll'
# note: command had no output on stdout or stderr
# RUN: at line 2
c:\_work\llvm-project\llvm-project\build\bin\llc.exe -mtriple=riscv32-apple-macho -filetype=obj C:\_work\llvm-project\llvm-project\llvm\test\CodeGen\RISCV\riscv-macho.ll -o C:\_work\llvm-project\llvm-project\build\test\CodeGen\RISCV\Output\riscv-macho.ll.tmp.o
# executed command: 'c:\_work\llvm-project\llvm-project\build\bin\llc.exe' -mtriple=riscv32-apple-macho -filetype=obj 'C:\_work\llvm-project\llvm-project\llvm\test\CodeGen\RISCV\riscv-macho.ll' -o 'C:\_work\llvm-project\llvm-project\build\test\CodeGen\RISCV\Output\riscv-macho.ll.tmp.o'
# .---command stderr------------
# | PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace and instructions to reproduce the bug.
# | Stack dump:
# | 0.	Program arguments: c:\\_work\\llvm-project\\llvm-project\\build\\bin\\llc.exe -mtriple=riscv32-apple-macho -filetype=obj C:\\_work\\llvm-project\\llvm-project\\llvm\\test\\CodeGen\\RISCV\\riscv-macho.ll -o C:\\_work\\llvm-project\\llvm-project\\build\\test\\CodeGen\\RISCV\\Output\\riscv-macho.ll.tmp.o
# | Exception Code: 0xC0000005
# |  #0 0x00007ffbfc852004 (C:\Windows\SYSTEM32\ntdll.dll+0x12004)
# |  #1 0x00007ffbfc851dda (C:\Windows\SYSTEM32\ntdll.dll+0x11dda)
# |  #2 0x00007ffbec47ff29 (C:\Windows\System32\ucrtbase.dll+0xff29)
# |  #3 0x00007ff700d71429 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x6f1429)
# |  #4 0x00007ff7034cfb73 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x2e4fb73)
# |  #5 0x00007ff702509ee8 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x1e89ee8)
# |  #6 0x00007ff700eef105 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x86f105)
# |  #7 0x00007ff700b419a1 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x4c19a1)
# |  #8 0x00007ff700b39fb1 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x4b9fb1)
# |  #9 0x00007ff700687c0c (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x7c0c)
# | #10 0x00007ff700684e86 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x4e86)
# | #11 0x00007ff704c9a1f4 (c:\_work\llvm-project\llvm-project\build\bin\llc.exe+0x461a1f4)
# | #12 0x00007ffbf0a74cb0 (C:\Windows\System32\KERNEL32.DLL+0x14cb0)
# | #13 0x00007ffbfc8bedcb (C:\Windows\SYSTEM32\ntdll.dll+0x7edcb)
# `-----------------------------
# error: command failed with exit status: 0xc0000005

--

If these failures are unrelated to your changes (for example tests are broken or flaky at HEAD), please open an issue at https://github.com/llvm/llvm-project/issues and add the infrastructure label.

[nhaehnle]

LGTM