Re: [AArch64] Fix possible wrong code generation when comparing DImode values.

2014-02-27 Thread Marcus Shawcroft
On 24 February 2014 18:17, James Greenhalgh wrote: > gcc/ > > 2014-02-24 James Greenhalgh > > * config/aarch64/aarch64-simd.md > (aarch64_cmdi): Always split. > (*aarch64_cmdi): New. > (aarch64_cmtstdi): Always split. > (*aarch64_cmtstdi): New. OK for s

Re: [AArch64] Fix possible wrong code generation when comparing DImode values.

2014-02-24 Thread James Greenhalgh
Hi, Back in May of last year I committed this patch to fix a nasty bug where we would accidentally trash the CCmode registers. http://gcc.gnu.org/ml/gcc-patches/2013-05/msg01337.html Shorty thereafter, Richard Earnshaw contacted me with a suggestion for a better fix. In short, in addition to spl

Re: [AArch64] Fix possible wrong code generation when comparing DImode values.

2013-05-23 Thread Richard Earnshaw
On 23/05/13 09:17, James Greenhalgh wrote: Hi, With the aarch64_cmdi patterns a bug was introduced. While the unsplit versions of these patterns, which operate in the SIMD register set, do not clobber CC_REGNUM, the split versions, which operate in the general purpose register set, do clobber C

[AArch64] Fix possible wrong code generation when comparing DImode values.

2013-05-23 Thread James Greenhalgh
Hi, With the aarch64_cmdi patterns a bug was introduced. While the unsplit versions of these patterns, which operate in the SIMD register set, do not clobber CC_REGNUM, the split versions, which operate in the general purpose register set, do clobber CC_REGNUM. This causes a problem if schedulin