[llvm] ISel/AArch64/SVE: custom lower vector ISD::[L]LRINT (PR #89035)

[Ramkumar Ramachandra via llvm-commits]

artagnon wrote:

Hi,

I looked at the code of FP_TO_SINT before deciding that it was impossible to custom-lower on neon. I've now updated the patch to use FP_TO_SINT, but there are no test updates. Since the added patch adds an extra SelectionDAG overhead, without any benefit, I'd be tempted to go with my original version: however, if you think the code for lowerVectorXRINT should be simpler, let me know.

https://github.com/llvm/llvm-project/pull/89035