Re: [FFmpeg-devel] [PATCH 1/2] lavu/bswap: remove some inline assembler

["Rémi Denis-Courmont" <remi@remlab.net>]

Le tiistaina 11. kesäkuuta 2024, 19.04.17 EEST Michael Niedermayer a écrit :
> then simply remove avr32 with that explanation (no C11 compiler, and any
> other reason)

No. Måns and my optimisation arguments stand, even if it is purely 
hypothetical in the case of AVR32 (for which there is no working compiler). It 
is a *general* argument.

Removing the AVR32 support is not the point of *this* patch, so you are asking 
me to misrepresent what the patch does and why. As for SH4, James already 
addressed that.

> but if a commit message says the code is removed because that "allows for
> better optimisations" then yes i ask for benchmarks

"Allows for better optimisations" means exactly that: enable compilers to 
*potentially* optimise better. I never claimed that it actually improved 
performance in any given particular case.

Nevertheless it will make performance worse in one and only one case: a 
defective/half-baked compiler: missing the byte-swap instruction (if it 
exists) and/or a proper scheduling model, for the target. In other words, you 
are essentially arguing that FFmpeg should be optimised for bad C compilers 
instead of good ones.

-- 
雷米‧德尼-库尔蒙
http://www.remlab.net/



_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".