[Mlir-commits] [mlir] [mlir][vector] Fix mask neutral value for masked fmaximum/fminimum reductions (PR #222275)

[Andrzej Warzyński]

================
@@ -0,0 +1,80 @@
+// RUN: mlir-opt %s -test-lower-to-llvm  | \
+// RUN: mlir-runner -e entry -entry-point-result=void  \
+// RUN:   -shared-libs=%mlir_c_runner_utils | \
+// RUN: FileCheck %s
+
+func.func @entry() {
----------------
banach-space wrote:

To me this is mixing too many extreme cases in one function and I am struggling to see the impact of `fastmath<ninf> ` on the result. Note that currently this is modifying the input vector (finite vs infinite values), combining op (max vs min) and mask type (ALL_FALSE vs partially TRUE). 

I suggest splitting this into multiple functions.

https://github.com/llvm/llvm-project/pull/222275