On Thu, Nov 21, 2013 at 11:37 AM, Jakub Jelinek wrote:
> On Thu, Nov 21, 2013 at 07:43:35AM +1000, Richard Henderson wrote:
>> On 11/20/2013 07:44 PM, Jakub Jelinek wrote:
>> > On Wed, Nov 20, 2013 at 10:31:38AM +0100, Richard Biener wrote:
>> >> Aww ;) Nice improvement. Generally when I see thi
On Thu, Nov 21, 2013 at 12:37:01PM +0100, Richard Biener wrote:
> Oh, indeed. Bah. That case makes the whole stuff quadratic, too ;)
True, O(nelts^2), but largest nelts we have right now is 64 (V64QImode
on -mavx512f).
> For
>
> typedef int vLARGEsi __attribute__((vector_size(1024*1024)));
Th
On Thu, 21 Nov 2013, Jakub Jelinek wrote:
> On Thu, Nov 21, 2013 at 12:18:45PM +0100, Richard Biener wrote:
> > > Bootstrap/regtest pending, ok at least for this for the start and can be
> > > improved later on?
> >
> > Ok, this should catch most of the vectorizer cases.
> >
> > Zero could also
On Thu, Nov 21, 2013 at 12:18:45PM +0100, Richard Biener wrote:
> > Bootstrap/regtest pending, ok at least for this for the start and can be
> > improved later on?
>
> Ok, this should catch most of the vectorizer cases.
>
> Zero could also be handled for PLUS_EXPR, likewise one for MULT_EXPR.
> I
On Thu, 21 Nov 2013, Jakub Jelinek wrote:
> On Thu, Nov 21, 2013 at 07:43:35AM +1000, Richard Henderson wrote:
> > On 11/20/2013 07:44 PM, Jakub Jelinek wrote:
> > > On Wed, Nov 20, 2013 at 10:31:38AM +0100, Richard Biener wrote:
> > >> Aww ;) Nice improvement. Generally when I see this I always
On Thu, Nov 21, 2013 at 07:43:35AM +1000, Richard Henderson wrote:
> On 11/20/2013 07:44 PM, Jakub Jelinek wrote:
> > On Wed, Nov 20, 2013 at 10:31:38AM +0100, Richard Biener wrote:
> >> Aww ;) Nice improvement. Generally when I see this I always wonder
> >> whether we want to do this kind of stu
On 11/20/2013 07:44 PM, Jakub Jelinek wrote:
> On Wed, Nov 20, 2013 at 10:31:38AM +0100, Richard Biener wrote:
>> Aww ;) Nice improvement. Generally when I see this I always wonder
>> whether we want to do this kind of stuff pre RTL expansion.
>> 1st to not rely on being able to TER, 2nd to final
On Wed, 20 Nov 2013, Jakub Jelinek wrote:
> On Wed, Nov 20, 2013 at 10:31:38AM +0100, Richard Biener wrote:
> > Aww ;) Nice improvement. Generally when I see this I always wonder
> > whether we want to do this kind of stuff pre RTL expansion.
> > 1st to not rely on being able to TER, 2nd to fina
On Wed, Nov 20, 2013 at 10:31:38AM +0100, Richard Biener wrote:
> Aww ;) Nice improvement. Generally when I see this I always wonder
> whether we want to do this kind of stuff pre RTL expansion.
> 1st to not rely on being able to TER, 2nd to finally eventually
> get rid of TER.
>
> These patches
On Wed, 20 Nov 2013, Jakub Jelinek wrote:
> Hi!
>
> I've noticed we generate terrible code for the testcase below.
> E.g. with -mavx2 it is:
> leal6(%rdi), %edx
> leal12(%rdi), %ecx
> leal18(%rdi), %esi
> leal3(%rdi), %eax
> movl%edx, -2
Hi!
I've noticed we generate terrible code for the testcase below.
E.g. with -mavx2 it is:
leal6(%rdi), %edx
leal12(%rdi), %ecx
leal18(%rdi), %esi
leal3(%rdi), %eax
movl%edx, -20(%rsp)
movl%ecx, -24(%rsp)
leal9(%rd
11 matches
Mail list logo
