[Pixman] ARM iwmmxt patches

[Matt Turner]

On Wed, Aug 31, 2011 at 8:12 AM, Soeren Sandmann <sandmann at cs.au.dk> wrote:
> Matt Turner <mattst88 at gmail.com> writes:
>
>> I've been trying to figure out if the ARM iwmmxt inline assembly makes
>> any difference at all. I think the conclusion is that it does not.
>> Updated code is here:
>> http://cgit.freedesktop.org/~mattst88/pixman/log/?h=iwmmxt-optimizations
>
> You mean that inline assembly in mmx_fill() doesn't make a difference?

That, and mmx_blt().

>> See
>> http://people.freedesktop.org/~mattst88/pixman-iwmmxt-benchdata.txt
>
> The lowlevel-blt benchmark doesn't hit the fill and blt routines at all,
> so this data doesn't support the conclusion that inline assembly in
> mmx_fill() and mmx_blt() makes no difference.

Well, that explains it.

>> Never does using inline assembly seem to make any sort of meaningful
>> difference over simply compiling pixman-mmx.c for ARM/iwmmxt. I tried
>> checking the alignment in the 'wip' commit in the blt function to
>> avoid a lot of unnecessary walign instructions, but as you can see
>> from the benchmark results, it doesn't help anything.
>
> The cairo-trace tests are better benchmarks to use in general because
> they reflect real-world use. lowlevel-blt-bench really should only be
> used for the case where you are optimizing a specific compositing
> routine.

OK, I'll run cairo-trace to determine the effect of the inline
assembly. I think the addition of inline assembly could go in as
follow-on patches though, right?

Matt