projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 94ca041) | patch
[TargetLowering] Use ComputeMaxSignificantBits instead of ComputeNumSignBits in expan...
authorCraig Topper <craig.topper@sifive.com>
Mon, 5 Sep 2022 05:20:52 +0000 (22:20 -0700)
committerCraig Topper <craig.topper@sifive.com>
Mon, 5 Sep 2022 05:35:16 +0000 (22:35 -0700)
commite529c0a2a03fb4eb0ddffafe0ddc7a02059f74cc
treef285ceb7005b1d0a615d08fa34a8d77967b1b33ftree | snapshot
parent94ca04190573908a811248559668a9b42b537ec1commit | diff
[TargetLowering] Use ComputeMaxSignificantBits instead of ComputeNumSignBits in expandMUL_LOHI. NFC

The way ComputeNumSignBits was being used was only correct if
OuterBitSize is exactly 2x InnerBitSize. Which is always true,
but not obviously so. Comparing ComputeMaxSignificantBits to
InnerBitSize feels more correct.
llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp diff | blob | history
Domain: System / Toolchain;