[llvm-branch-commits] [llvm] [AArch64] Fold zero-high vector inserts in MI peephole optimisation (PR #182835)

[David Green via llvm-branch-commits]

================
@@ -0,0 +1,13 @@
+; RUN: llc -mtriple=aarch64-none-linux-gnu -start-before=aarch64-isel %s -o /dev/null
+; Regression test for AArch64 compile-time regression, referring to PR #166962.
+
+define fastcc <2 x i64> @_ZN10tensorflow12_GLOBAL__N_125ComputeXWeightsAndIndicesERKNS_17ImageResizerStateEbPNSt3__u6vectorINS0_17WeightsAndIndicesENS4_9allocatorIS6_EEEE(<2 x i64> %0) {
----------------
davemgreen wrote:

This test might be better placed in the other PR, as AFAIU it was cause by the other part.

Also:
 The test name could be a bit shorted.
 You can remove Function Attrs and declare @llvm.smin.v2i64 and usually the attributes.
 And remove fastcc usually too.

https://github.com/llvm/llvm-project/pull/182835