[PATCH] D15675: AVX512: VPMOVB2M intrinsic implementation.

[Elena Demikhovsky via llvm-commits]

delena added a comment.

Please take the shifts patch separately.


================
Comment at: lib/Target/X86/X86ISelLowering.cpp:13226
@@ +13225,3 @@
+    // Use AND with 1 ( memory access) and TESTM.
+    if (InVT == MVT::v64i8) {
+      SDValue OneV = DAG.getBitcast(MVT::v16i32, DAG.getConstant(1, DL, InVT));
----------------
32i8?


Repository:
  rL LLVM

http://reviews.llvm.org/D15675