diff options
| author | Jay Foad <jay.foad@amd.com> | 2023-11-15 10:58:00 +0000 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-11-15 10:58:00 +0000 |
| commit | 1e8c17e9c770d801ef5e9929eb330e30750e38f5 (patch) | |
| tree | dbed03b2ab208af551d5a7c23163f65d96d22c2a /flang/lib/Frontend/CompilerInvocation.cpp | |
| parent | 767b34297d372f8fcf854cf1b77eb685e8ea8e33 (diff) | |
| download | llvm-1e8c17e9c770d801ef5e9929eb330e30750e38f5.zip llvm-1e8c17e9c770d801ef5e9929eb330e30750e38f5.tar.gz llvm-1e8c17e9c770d801ef5e9929eb330e30750e38f5.tar.bz2 | |
[AMDGPU] Allow folding to FMAMK with SGPR and immediate operand on GFX10+ (#72258)
Allow foldImmediate to create instructions like:
v_fmamk_f32 v0, s0, 0x42000000, v0
This instruction has two "scalar values": s0 and 0x42000000. On GFX10+
this is allowed. This fold was originally implemented before the
compiler supported GFX10, when all ASICs were limited to one scalar
value.
Diffstat (limited to 'flang/lib/Frontend/CompilerInvocation.cpp')
0 files changed, 0 insertions, 0 deletions