AMDGPU: Hack out noinline on functions using LDS globals
authorMatt Arsenault <Matthew.Arsenault@amd.com>
Thu, 2 Apr 2020 15:52:23 +0000 (11:52 -0400)
committerMatt Arsenault <arsenm2@gmail.com>
Thu, 2 Apr 2020 18:12:07 +0000 (14:12 -0400)
commitc3d3c22a5834ac8afe9618ef9eed5dac370c080d
tree6e0c3fd1b95e087c6196e25ff96341e3b99b6f7e
parentfb80b6b2d58c476747a3206bd4371b787108591b
AMDGPU: Hack out noinline on functions using LDS globals

This is a workaround for clang adding noinline to all functions at
-O0. Previously, we would just add alwaysinline, and the verifier
would complain about having both noinline and alwaysinline. We
currently can't truly codegen this case as a freestanding function, so
override the user forcing noinline.
llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
llvm/test/CodeGen/AMDGPU/force-alwaysinline-lds-global-address.ll