[llvm] [AMDGPU] Allocate i1 argument to SGPRs (PR #72461)

[Matt Arsenault via llvm-commits]

================
@@ -513,6 +513,12 @@ bool SILowerI1Copies::lowerCopiesFromI1() {
       if (isLaneMaskReg(DstReg) || isVreg1(DstReg))
         continue;
 
+      // When the calling convention allocates i1 argument to SGPR,
+      // we may have a COPY with dst being an SGPR_32. This should
+      // not be lowered into V_CNDMASK_B32.
+      if (AMDGPU::SGPR_32RegClass.contains(DstReg))
----------------
arsenm wrote:

Ditto, this is wrong 

https://github.com/llvm/llvm-project/pull/72461