[PATCH] D158950: [X86][BF16] Lower FP_EXTEND for vector types under AVX512BF16

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/lib/Target/X86/X86ISelLowering.cpp:21196
+                         DAG.getNode(ISD::FP_EXTEND, DL, TmpVT, In));
+    }
+    MVT IntVT = SVT.changeTypeToInteger();
----------------
assert(VT.getVectorElementType() == MVT::f32 && "Unexpected fpext");


================
Comment at: llvm/lib/Target/X86/X86ISelLowering.cpp:21208
+                     DAG.getConstant(0, DL, IntVT));
+    In = DAG.getVectorShuffle(NVT, DL, In, DAG.getUNDEF(NVT), Mask);
+    return DAG.getBitcast(VT, In);
----------------
Would it be better to do a shl(zext(In,vXi32),8) here?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D158950/new/

https://reviews.llvm.org/D158950