[llvm] [AMDGPU] Optimize DPP for fmin/fmax functions (PR #195282)

Shilei Tian via llvm-commits llvm-commits at lists.llvm.org
Fri May 1 11:19:20 PDT 2026


================
@@ -489,6 +490,44 @@ static bool isIdentityValue(unsigned OrigMIOp, MachineOperand *OldOpnd) {
     if (OldOpnd->getImm() == 1)
       return true;
     break;
+  case AMDGPU::V_MIN_F32_e32:
+  case AMDGPU::V_MIN_F32_e64:
+    if (static_cast<uint32_t>(OldOpnd->getImm()) == /*+inf=*/0x7F800000)
----------------
shiltian wrote:

Well, that is not sufficient, since it doesn't have -inf, but only +inf.

https://github.com/llvm/llvm-project/pull/195282


More information about the llvm-commits mailing list