[AMDGPU] Enhance s_waitcnt insertion before barrier for gfx12 (#90595)
authorDavid Stuttard <david.stuttard@amd.com>
Wed, 1 May 2024 10:37:13 +0000 (11:37 +0100)
committerTom Stellard <tstellar@redhat.com>
Thu, 9 May 2024 03:08:59 +0000 (20:08 -0700)
commit58e44d3c6f67d5402ec38913d4262b94e73ac123
treeaf1eff96ee2b6c344c087c480b04869619c3d1e8
parentd1d7131906d9c912fad01078483ba0248d16feb2
[AMDGPU] Enhance s_waitcnt insertion before barrier for gfx12 (#90595)

Code to determine if a waitcnt is required before a barrier instruction
only
considered S_BARRIER.
gfx12 adds barrier_signal/wait so need to enhance the existing code to
look for
a barrier start (which is just an S_BARRIER for earlier architectures).
llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
llvm/lib/Target/AMDGPU/SIInstrInfo.h
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.s.barrier.ll
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.s.barrier.wait.ll