[llvm-branch-commits] [llvm] [AMDGPU] Add wave reduce intrinsics for float types - 2 (PR #161815)

[Juan Manuel Martinez Caamaño via llvm-branch-commits]

================
@@ -5330,11 +5330,13 @@ static uint32_t getIdentityValueFor32BitWaveReduction(unsigned Opc) {
   case AMDGPU::S_MAX_U32:
     return std::numeric_limits<uint32_t>::min();
   case AMDGPU::S_MAX_I32:
+  case AMDGPU::V_SUB_F32_e64: // +0.0
----------------
jmmartinez wrote:

This doesn't seem right.

Isn't `0b10000000000000000000000000000000` the opposite, -0.0 ?

I'd feel reassured if you used a bitcast instead of a comment (comments tend to diverge from the code eventually): `__builtin_bit_cast(+0.0f, uint32_t)`.

https://github.com/llvm/llvm-project/pull/161815