b96b37dc8fb417dc4123246d7bfd910789de810a - llvm

commit	b96b37dc8fb417dc4123246d7bfd910789de810a	[log] [tgz]
author	Nikita Popov <nikita.ppv@gmail.com>	Tue Jan 15 18:43:41 2019 +0000
committer	Nikita Popov <nikita.ppv@gmail.com>	Tue Jan 15 18:43:41 2019 +0000
tree	2cbeb95db17139bd86f0289c9a442376200f29ca
parent	bb3d0e960a93c1b140b7322a306bb5b908a3650e [diff]

Reapply "[CodeGen][X86] Expand USUBSAT to UMAX+SUB, also for vectors"

Related to https://bugs.llvm.org/show_bug.cgi?id=40123.

Rather than scalarizing, expand a vector USUBSAT into UMAX+SUB,
which produces much better code for X86.

Reapplying with updated SLPVectorizer tests.

Differential Revision: https://reviews.llvm.org/D56636

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@351219 91177308-0d34-0410-b5e6-96231b3b80d8

lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp[diff]
lib/CodeGen/SelectionDAG/TargetLowering.cpp[diff]
lib/Target/X86/X86TargetTransformInfo.cpp[diff]
test/Analysis/CostModel/X86/arith-usat.ll[diff]
test/CodeGen/X86/usub_sat.ll[diff]
test/CodeGen/X86/usub_sat_vec.ll[diff]
test/Transforms/SLPVectorizer/X86/arith-sub-usat.ll[diff]

7 files changed

tree: 2cbeb95db17139bd86f0289c9a442376200f29ca