projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 7fa75ce) | patch
For ARM subtargets with useNEONForSinglePrecisionFP, double count uses
authorAndrew Trick <atrick@apple.com>
Wed, 19 Jan 2011 02:35:27 +0000 (02:35 +0000)
committerAndrew Trick <atrick@apple.com>
Wed, 19 Jan 2011 02:35:27 +0000 (02:35 +0000)
commit32cec0a75678401ce079bb31fa748ae0d6613e2d
tree28cf916f9f739aa306c25a50286a73ceff1301aftree | snapshot
parent7fa75ce11d8f491d9f68c3cabe60f6fc70787748commit | diff
For ARM subtargets with useNEONForSinglePrecisionFP, double count uses
of the floating point types less than 64-bits. It's somewhat of a temporary
hack but forces more accurate modeling of register pressure and results
in fewer spills.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@123811 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/ARM/ARMISelLowering.cpp diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors