Don't lower log1p(x) to log(1 + x).
authorJohannes Reifferscheid <jreiffers@google.com>
Thu, 11 Aug 2022 13:56:07 +0000 (15:56 +0200)
committerJohannes Reifferscheid <jreiffers@google.com>
Tue, 16 Aug 2022 04:58:00 +0000 (06:58 +0200)
commit375a5cb648835db0b1eacfc921cbb04844b8b3b4
tree34ceca893af40ac67b38ff1c1a7b09b5c277e27f
parent556efdba8569f748fd805e30beeae8afa2d6a343
Don't lower log1p(x) to log(1 + x).

The latter has accuracy issues around 0. The lowering in MathToLLVM is kept for now.

Reviewed By: bkramer

Differential Revision: https://reviews.llvm.org/D131676
mlir/include/mlir/Conversion/MathToLibm/MathToLibm.h
mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
mlir/lib/Conversion/MathToLibm/MathToLibm.cpp
mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
mlir/test/Conversion/MathToLibm/convert-to-libm.mlir