Re: ARM ldm/stm peepholes

[Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>]

Hi Bernd,

On Tue, 2010-04-20 at 13:41 +0200, Bernd Schmidt wrote:

> The patch may need more tuning to decide when to use these instructions;
> I'll need suggestions as I don't think I have enough information
> available to make these decisions.  For now, this uses the same
> heuristics as previously, but it may trigger in slightly more
> situations.  I'm open to suggestions.

Apologies for the delayed response, been away from my desk this week and
will be away for more today.

It would be useful to be able to control ldm /stm generation on a
per-core basis i.e. on some cores it might be useful to have ldm's and
stm's as a size optimization rather than as a performance optimization.
For instance the A8 and A9 have a good ldm and stm setup but there might
be circumstances where this might not be the right thing to do.

An instance of where you don't want to use ldm / stm is this patch
suggested here.
  http://gcc.gnu.org/ml/gcc-patches/2009-11/msg01553.html . 

Having a hook that acted as a gate for matching this peephole would be
useful. Then we could change this hook to work on a per-core basis or
put in any heuristics we want to change or tweak later.


> I've tested this quite often with arm-none-linux-gnueabi QEMU testsuite
> runs.  Yesterday, after fixing an aliasing issue, I managed to get a
> successful SPEC2000 run on a Cortex-A9 board (with a 4.4-based compiler;
> overall performance minimally higher than before); since then I've only
> made cleanups and added comments.  Once approved, I'll rerun all tests
> before committing.

Out of curiosity, was the aliasing issue PR42509 on a 4.4 branch ? If so
do you have a testcase that necessitates this fix to be backported to
the 4.4 branch upstream as well ?

cheers
Ramana

> 
> 
> Bernd