[PATCH] D56451: [AMDGPU] Expand fdot2 intrinsic if unsupported

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec created this revision.
rampitec added reviewers: b-sumner, msearles, arsenm.
Herald added subscribers: t-tye, tpr, dstuttard, yaxunl, nhaehnle, wdng, jvesely, kzhuravl.

This makes intrinsic useful in the library even if there is no
HW support. The answer may slightly differ from v_dot2_f32_f16
because of the intermediate rounding.


https://reviews.llvm.org/D56451

Files:
  lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
  test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D56451.180723.patch
Type: text/x-patch
Size: 5757 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190108/1fc8045e/attachment.bin>