projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d21182d) | patch
AMDGPU/GlobalISel: Legalize unpacked d16 image operations
authorMatt Arsenault <Matthew.Arsenault@amd.com>
Wed, 15 Jan 2020 19:23:20 +0000 (14:23 -0500)
committerMatt Arsenault <arsenm2@gmail.com>
Thu, 30 Jan 2020 13:36:11 +0000 (08:36 -0500)
commit872e899b7563a033228be935a04bfb853bd73acd
treef9f8c03ef45d10d81b12578065d4a6492cb74c4ftree | snapshot
parentd21182d692e8109fdc3af1eb52dd293fbb3e876fcommit | diff
AMDGPU/GlobalISel: Legalize unpacked d16 image operations

On targets that don't have the normal packed f16 layout, handle these
during legalization. Directly modify the register types. We can infer
this was a d16 load based on the mem operand size during selection.

A16 operands should possibly be handled here as well, but don't worry
about that yet.
llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp diff | blob | history
llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h diff | blob | history
llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.load.2d.d16.ll [new file with mode: 0644] blob
llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.store.2d.d16.ll [new file with mode: 0644] blob
Domain: System / Toolchain;