[Mlir-commits] [mlir] [mlir][NVVM] Lower vector<1xf32> nvvm.ex2 through the scalar intrinsic (PR #231517)

[Bastian Hagedorn]

bastianhagedorn wrote:

Gentle ping @grypp @durga4github. This now follows grypp's suggestion: `vector<1xf32>` stays accepted and is lowered through the scalar f32 intrinsic, using bitcasts as Durga suggested. Premerge is green. Could one of you take another look and approve?

https://github.com/llvm/llvm-project/pull/231517