projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 0bd8ea3) | patch
ARM64: fix fmsub patterns which assumed accum operand was first
authorTim Northover <tnorthover@apple.com>
Tue, 8 Apr 2014 12:23:51 +0000 (12:23 +0000)
committerTim Northover <tnorthover@apple.com>
Tue, 8 Apr 2014 12:23:51 +0000 (12:23 +0000)
commit362090adf501e0aa4f39c0bf21002c26cd2eb6d7
tree83d6eccf6ac5ee0426a839f5b9b2e27cb473d180tree | snapshot
parent0bd8ea3823bd842a1852db97bad6c7d1cbebdc47commit | diff
ARM64: fix fmsub patterns which assumed accum operand was first

Confusingly, the NEON fmla instructions put the accumulator first but the
scalar versions put it at the end (like the fma lib function & LLVM's
intrinsic).

This should fix PR19345, assuming there's only one issue.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205758 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/ARM64/ARM64InstrInfo.td diff | blob | history
test/CodeGen/ARM64/fmadd.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors