projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 437c286) | patch
Revert "[ARM] Add more pattern matching for f16 <-> f64 conversions"
authorBradley Smith <bradley.smith@arm.com>
Mon, 23 Mar 2015 16:52:52 +0000 (16:52 +0000)
committerBradley Smith <bradley.smith@arm.com>
Mon, 23 Mar 2015 16:52:52 +0000 (16:52 +0000)
commita75fecc3702562ef6c51c76242b9c31eed40b67a
treee2a17614af683aa06d59d709be6687b13959d4c5tree | snapshot
parent437c2863f0dc9795c7f2bb8c1b24baf8d96e8cf0commit | diff
Revert "[ARM] Add more pattern matching for f16 <-> f64 conversions"

This change is incorrect since it converts double rounding into single rounding,
which can produce different results. Instead this optimization will be done by
modifying Clang's codegen to not produce double rounding in the first place.

This reverts commit r232954.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@232962 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/ARM/ARMInstrVFP.td diff | blob | history
test/CodeGen/ARM/fp16-64.ll [deleted file] blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors