[FFmpeg-cvslog] r16823 - trunk/libavcodec/arm/dsputil_arm_s.S

[Måns Rullgård]

Michael Niedermayer <michaelni at gmx.at> writes:

> On Tue, Jan 27, 2009 at 07:40:29PM +0000, M?ns Rullg?rd wrote:
>> Michael Niedermayer <michaelni at gmx.at> writes:
>> 
>> > On Tue, Jan 27, 2009 at 05:25:23PM +0000, M?ns Rullg?rd wrote:
>> >> Reimar D?ffinger <Reimar.Doeffinger at stud.uni-karlsruhe.de> writes:
>> >> 
>> >> > On Tue, Jan 27, 2009 at 05:06:51PM +0100, mru wrote:
>> >> >> Author: mru
>> >> >> Date: Tue Jan 27 17:06:51 2009
>> >> >> New Revision: 16823
>> >> >> 
>> >> >> Log:
>> >> >> ARM: reorder some instructions in put_pixels*_arm
>> >> >
>> >> > I think a log message is supposed to contain the why (faster, more
>> >> > readable, ...?)
>> >> 
>> >> Such changes are always for speed.
>> >
>> > and the cpu and speed difference ...
>> >
>> > I think i remember you complaining that i forgot the cpu name, now you
>> > do too :)
>> 
>> ARM of course.
>
> so i should say i benchmarked the x86 asm on x86 next time?

x86 CPUs vary wildly in terms of what code is fast.  What's
lightning-fast on one can be slow on another.  ARM is much more
predictable.  If a trivial change makes code faster on one core,
it is virtually always faster or unchanged on others.

Why this sudden inquisition anyway?

-- 
M?ns Rullg?rd
mans at mansr.com