projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e8adbae) | patch
[CodeGen][X86] Don't scalarize vector saturating add/sub
authorNikita Popov <nikita.ppv@gmail.com>
Sun, 10 Feb 2019 19:06:38 +0000 (19:06 +0000)
committerNikita Popov <nikita.ppv@gmail.com>
Sun, 10 Feb 2019 19:06:38 +0000 (19:06 +0000)
commita0e96bd56d99d6313b15af0900859a8cb3ad73df
tree61dc104e97810fed9563a231dfd8c1b6409c5b0ctree | snapshot
parente8adbae00a3e78795a653dadb58425cb89050e4ccommit | diff
[CodeGen][X86] Don't scalarize vector saturating add/sub

Now that we have vector support for [US](ADD|SUB)O we no longer
need to scalarize when expanding [US](ADD|SUB)SAT.

This matches what the cost model already does.

Differential Revision: https://reviews.llvm.org/D57348

llvm-svn: 353651
llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp diff | blob | history
llvm/test/CodeGen/X86/sadd_sat.ll diff | blob | history
llvm/test/CodeGen/X86/sadd_sat_vec.ll diff | blob | history
llvm/test/CodeGen/X86/ssub_sat.ll diff | blob | history
llvm/test/CodeGen/X86/ssub_sat_vec.ll diff | blob | history
Domain: System / Toolchain;