[PATCH] D58563: [AArch64] Improve FP16 vector convert from short instructions

[Abderrazek Zaafrani via Phabricator via llvm-commits]

az created this revision.
az added a reviewer: SjoerdMeijer.
Herald added subscribers: llvm-commits, hiraditya, kristof.beyls, javed.absar.
Herald added a project: LLVM.

We currently generate inefficient code for vector convert from short int to half instructions: promote to int first, then convert, and then demote to half. With this patch and when the fp16 feature is on, we generate straight fp16 vector conversion.


Repository:
  rL LLVM

https://reviews.llvm.org/D58563

Files:
  llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
  llvm/test/CodeGen/AArch64/fp16-v4-instructions.ll
  llvm/test/CodeGen/AArch64/fp16-v8-instructions.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D58563.188008.patch
Type: text/x-patch
Size: 6906 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190223/235612b4/attachment.bin>