[PATCH] D38220: [X86] Finishing broadcast[f|i]32x2 intrinsics lowering to IR.

michael zuckerman via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Sep 25 05:04:58 PDT 2017


m_zuckerman added inline comments.


================
Comment at: lib/IR/AutoUpgrade.cpp:250
       Name == "avx2.vbroadcasti128" || // Added in 3.7
+      Name.startswith("avx512.mask.broadcastf32x2.") || // Added in 6.0
       Name.startswith("avx512.mask.broadcastf32x4.") || // Added in 6.0
----------------
can you use the "avx512.mask.broadcastf" and "avx512.mask.broadcasti" instead of adding a new line? 


================
Comment at: lib/IR/AutoUpgrade.cpp:253
       Name.startswith("avx512.mask.broadcastf64x2.") || // Added in 6.0
+      Name.startswith("avx512.mask.broadcasti32x2.") || // Added in 6.0
       Name.startswith("avx512.mask.broadcasti32x4.") || // Added in 6.0
----------------
same as above 


https://reviews.llvm.org/D38220





More information about the llvm-commits mailing list