[PATCH] D6970: R600/SI: Try to form (f64 [s|u]int_to_fp i1)
Matt Arsenault via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Sat Nov 12 10:44:14 PST 2022
arsenm updated this revision to Diff 474949.
arsenm retitled this revision from "R600/SI: Try to form (f64 [s|u]int_to_fp i1)
" to "R600/SI: Try to form (f64 [s|u]int_to_fp i1)".
arsenm edited the summary of this revision.
arsenm added a reviewer: AMDGPU.
arsenm added a comment.
Herald added subscribers: kosarev, foad, ecnelises, kerbowa, hiraditya, wdng, jvesely.
Herald added projects: LLVM, All.
Rebase 7 years into the future.
This optimization doesn't make sense if v_cvt_f64_i32 is quarter rate, which it is on some subtargets at least. I'm not sure the scheduler model is 100% accurate
for every subtarget, and it seems to always say it's quarter rate
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D6970/new/
https://reviews.llvm.org/D6970
Files:
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
llvm/lib/Target/AMDGPU/SIISelLowering.cpp
llvm/test/CodeGen/AMDGPU/sint_to_fp.f64.ll
llvm/test/CodeGen/AMDGPU/uint_to_fp.f64.ll
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D6970.474949.patch
Type: text/x-patch
Size: 13879 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221112/0a4e1cde/attachment.bin>
More information about the llvm-commits
mailing list