projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 64ad6ea) | patch
[AArch64] Use fneg instead of fsub -0.0, X Cin IR expansion of __builtin_neon_vfmsh_f16.
authorCraig Topper <craig.topper@sifive.com>
Tue, 4 Apr 2023 16:00:45 +0000 (09:00 -0700)
committerCraig Topper <craig.topper@sifive.com>
Tue, 4 Apr 2023 16:01:24 +0000 (09:01 -0700)
commit0109f8d1e3bf64e4b23db6e2f284185207e46541
tree1507a1f12273ab1d0fd695ee9bed5bcb74638faatree | snapshot
parent64ad6ead1f5839dc6920295d00ec8b16ea118f4acommit | diff
[AArch64] Use fneg instead of fsub -0.0, X Cin IR expansion of __builtin_neon_vfmsh_f16.

Addresses the FIXME and removes the only in tree use of
llvm::ConstantFP::getZeroValueForNegation for an FP type.

Reviewed By: dmgreen, SjoerdMeijer

Differential Revision: https://reviews.llvm.org/D147497
clang/lib/CodeGen/CGBuiltin.cpp diff | blob | history
clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c diff | blob | history
clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c diff | blob | history
Domain: System / Toolchain;