Files
llvm-project/llvm/test/CodeGen/ARM
Charlie Turner 7968b981bf [ARM] Don't pessimize i32 vselect.
The underlying issues surrounding codegen for 32-bit vselects have been resolved. The pessimistic costs for 64-bit vselects remain due to the bad
scalarization that is still happening there.

I tested this on A57 in T32, A32 and A64 modes. I saw no regressions, and some improvements.

From my benchmarks, I saw these improvements in A57 (T32)
spec.cpu2000.ref.177_mesa 5.95%
lnt.SingleSource/Benchmarks/Shootout/strcat 12.93%
lnt.MultiSource/Benchmarks/MiBench/telecomm-CRC32/telecomm-CRC32 11.89%

I also measured A57 A32, A53 T32 and A9 T32 and found no performance regressions. I see much bigger wins in third-party benchmarks with this change

Differential Revision: http://reviews.llvm.org/D14743

llvm-svn: 253349
2015-11-17 17:25:15 +00:00
..
2015-10-09 03:19:03 +00:00
2014-04-03 17:35:22 +00:00
2014-09-02 22:45:13 +00:00
2014-02-16 07:31:05 +00:00
2014-08-26 12:47:26 +00:00
2014-01-29 11:50:56 +00:00
2014-01-29 11:50:56 +00:00
2015-11-17 00:51:23 +00:00
2015-10-26 20:49:49 +00:00
2014-04-25 17:51:25 +00:00
2014-11-17 14:08:57 +00:00
2014-04-03 17:35:22 +00:00
2014-04-03 17:35:22 +00:00
2014-12-04 19:34:50 +00:00
2014-04-03 17:35:22 +00:00
2014-04-03 17:35:22 +00:00
2015-01-19 15:16:06 +00:00
2014-04-03 17:35:22 +00:00
2014-04-23 01:09:29 +00:00
2014-01-24 17:20:08 +00:00
2015-08-13 17:28:16 +00:00