[Mlir-commits] [mlir] [mlir][ArmSVE] Add `arm_sve.psel` operation (PR #95764)

[Andrzej Warzyński]

================
@@ -140,6 +140,28 @@ using ConvertFromSvboolOpLowering =
 using ZipX2OpLowering = OneToOneConvertToLLVMPattern<ZipX2Op, ZipX2IntrOp>;
 using ZipX4OpLowering = OneToOneConvertToLLVMPattern<ZipX4Op, ZipX4IntrOp>;
 
+/// Lower `arm_sve.psel` to LLVM intrinsics. This is almost a 1-to-1 conversion
+/// but first input (P1) and result predicates need conversion to/from svbool.
----------------
banach-space wrote:

How come `p2` doesn't require converting?

https://github.com/llvm/llvm-project/pull/95764