maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
263,272 Commits 23 Branches 0 Tags 2.2 GiB
Commit Graph

6 Commits

Author SHA1 Message Date
Craig Topper 464b8cb244 [X86] Don't base domain decisions on VEXTRACTF128/VINSERTF128 if only AVX1 is available. 
Seems the execution dependency pass likes to use FP instructions when most of the consuming code is integer if a vextractf128 instruction produced the register. Without AVX2 we don't have the corresponding integer instruction available.

This patch suppresses the domain on these instructions to GenericDomain if AVX2 is not supported so that they are ignored by domain fixing. If AVX2 is supported we'll report the correct domain and allow them to switch between integer and fp.

Overall I think this produces better results in the modified test cases.

llvm-svn: 294824
 2017-02-11 05:32:57 +00:00
Simon Pilgrim 9a36318c54 EOL fixes 
llvm-svn: 280624
 2016-09-04 13:30:46 +00:00
Simon Pilgrim 7b09af193a [X86][SSE] Improve awareness of fptrunc implicit zeroing of upper 64-bits of xmm result 
Add patterns to avoid inserting unnecessary zeroing shuffles when lowering fptrunc to (v)cvtpd2ps

Differential Revision: https://reviews.llvm.org/D23797

llvm-svn: 280214
 2016-08-31 10:35:13 +00:00
Simon Pilgrim 9eb978b47b [X86][SSE] Demonstrate inability to recognise that (v)cvtpd2ps implicitly zeroes the upper half of the xmm 
llvm-svn: 279508
 2016-08-23 10:35:24 +00:00
Simon Pilgrim 8275583be6 Regenerate test 
llvm-svn: 279386
 2016-08-20 21:37:30 +00:00
Simon Pilgrim a2c8da9e06 [X86][AVX] Added vector float truncation (double2float) tests 
llvm-svn: 265222
 2016-04-02 14:09:17 +00:00