[PATCH] D35700: DAGCombiner: Extend reduceBuildVecToTrunc to handle non-zero offset

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: include/llvm/Target/TargetLowering.h:2782
+      int Stride, int Offset, EVT SrcVT, EVT BuildVecVT) const {
+    return false;
+  }
----------------
It would possibly make sense to change this to
```
isDesirableToCombineBuildVectorToShuffleTruncate(ArrayRef<int> ShuffleMask, EVT SrcVT, EVT TruncVT)
```
checking with a general shuffle mask - I'd expect any non-lane crossing shuffle mask to be valid on AVX512.


https://reviews.llvm.org/D35700