HomePhabricator

AMDGPU: Emit llvm.fshr for __builtin_amdgcn_alignbit

Authored by arsenm on Mar 19 2020, 6:40 PM.

Description

AMDGPU: Emit llvm.fshr for __builtin_amdgcn_alignbit

These are equivalent. The generic rotate builtins do not directly map
to the fshr intrinsic.

Details

Committed
arsenmMar 23 2020, 1:51 PM
Parents
rGb20a1d840f0c: GVNSink: Allow handling addrspacecast
Branches
Unknown
Tags
Unknown