[AMDGPU] Turn off clang-format in moveToVALU

[piotrAMD]

No description provided.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Piotr Sobczak (piotrAMD)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/75188.diff

1 Files Affected:

• (modified) llvm/lib/Target/AMDGPU/SIInstrInfo.cpp (+5)

diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index dda54a0929c194..42eb4fa55ed525 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -5157,6 +5157,9 @@ bool SIInstrInfo::verifyInstruction(const MachineInstr &MI,
   return true;
 }
 
+// It is more readable to list mapped opcodes on the same line.
+// clang-format off
+
 unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) const {
   switch (MI.getOpcode()) {
   default: return AMDGPU::INSTRUCTION_LIST_END;
@@ -5296,6 +5299,8 @@ unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) const {
       "Unexpected scalar opcode without corresponding vector one!");
 }
 
+// clang-format on
+
 void SIInstrInfo::insertScratchExecCopy(MachineFunction &MF,
                                         MachineBasicBlock &MBB,
                                         MachineBasicBlock::iterator MBBI,

[jayfoad]

Personally I have a slight preference for clang-formatting this function. But this seems fine too.

[piotrAMD]

I was in two minds about this one. The advantage of not doing any reformatting now is that maintaining the status quo preserves history (git blame etc.). So this just "documents" the existing convention.