This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [tree-ssa] xfail i386-ssetypes-[13].c


> On Wed, Feb 04, 2004 at 01:31:09PM +0100, Jan Hubicka wrote:
> > 	* i386.md (sse/sse2 packed logical expanders): Do not accept memory
> > 	operands.
> 
> I'm not thrilled about this at all.

Me neither, but it makes it work so we don't have to xfail this testcase
and forget about the bug forever :)

I always worried about fragility of the code, this is why i made the
testcases at first place.  However it appears to mostly work - I
teststed quite large testsuite with counters for reformating penalties
and found none.
> 
> > > I have no idea how to fix this except to avoid using logical:TI here at all.
> > > Alternatives are unspec:V2DF or (possibly) logical:V2DF.
> 
> I really think this might be a better idea.

I don't quite like the unspec way, because we sometimes auto generate
the logicals and thus it make sense to optimize them.
If you think that the logical:V2DF is the way to go, I will implement
it.

Honza
> 
> In any case, I'm not going to address this right away.
> 
> 
> r~


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]