On Tue, 13 Sep 2011 21:44:54 +0100
Ramana Radhakrishnan wrote:
> On 9 September 2011 13:56, Richard Sandiford
> wrote:
> > Ping for this patch:
>
> I do have a one nit on the ml bit though I must say I'm not an ML
> expert which is why I resisted for a while. The one comment that I
> have and I
On 14 September 2011 08:47, Richard Sandiford
wrote:
> Ramana Radhakrishnan writes:
>> I do have a one nit on the ml bit though I must say I'm not an ML
>> expert which is why I resisted for a while. The one comment that I
>> have and I should have realized earlier was that the file had been
>> p
Ramana Radhakrishnan writes:
> I do have a one nit on the ml bit though I must say I'm not an ML
> expert which is why I resisted for a while. The one comment that I
> have and I should have realized earlier was that the file had been
> parameterized by the core in quite a few places and I would l
On 9 September 2011 13:56, Richard Sandiford
wrote:
> Ping for this patch:
>> This is the NEON part of the patch to handle address register writeback
>> in the Cortex A8 and A9 schedulers. Although I can find no documentation
>> to say exactly how this is handled by the pipelines, a latency of 1
Ping for this patch:
http://gcc.gnu.org/ml/gcc-patches/2011-08/msg01488.html
which models address register writeback in the Cortex A8 and A9 NEON
schedulers. (Ramana has already approved the core equivalent, thanks.)
Richard
> Tested on arm-linux-gnueabi. OK to install?
> gcc/
>* config/arm/arm-protos.h (arm_writeback_dep): Declare.
> (arm_writeback_only_dep): Likewise.
> * config/arm/arm.c (arm_writeback_dep): New function.
> (arm_writeback_only_dep_1, arm_writeback_only_dep): Likewise.
>
This is the NEON part of the patch to handle address register writeback
in the Cortex A8 and A9 schedulers. Although I can find no documentation
to say exactly how this is handled by the pipelines, a latency of 1
does seem to work well in practice, and is much easier to implement.
Tested in the s
...well, for core instructions at least. NEON is a separate patch.
I haven't measured any benefit or regression from this change on its own.
It makes a difference with the new auto-inc-dec pass though.
I diffed the "before" and "after" assembly code for libav to get a sense
for whether the patch
