Re: [FFmpeg-devel] [PATCH 2/4] lavc/mpegvideo: use H263DSP dequant function - Rémi Denis-Courmont

From: "Rémi Denis-Courmont" <remi@remlab.net>
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH 2/4] lavc/mpegvideo: use H263DSP dequant function
Date: Fri, 14 Jun 2024 22:03:00 +0300
Message-ID: <4869369.p5KnOXz9Un@basile.remlab.net> (raw)
In-Reply-To: <27edc4bb-b668-4344-b0cd-45fa11a7712e@gmail.com>

Le perjantaina 14. kesäkuuta 2024, 18.41.48 EEST James Almer a écrit :
> On 6/14/2024 12:11 PM, Rémi Denis-Courmont wrote:
> > Le perjantaina 14. kesäkuuta 2024, 17.45.50 EEST Rémi Denis-Courmont a 
écrit :
> >> And x86 only has MMX.
> > 
> > And the x86 code uses inline assembler. That's not viable on any ISA with
> > sane conventions such as Arm or RISC-V. The compiler will reject our
> > vector clobbers, unless the Vector extension is included in the
> > compilation target. But then that breaks run-time detection, and will be
> > rejected by all Linux distros.
> 
> I was thinking on removing that inline version and probably replacing it
> with sse2, using the dsp prototypes you introduce in this set.

I saw that, but we can't have the cake and eat it. Those DSP callbacks can't 
exist without adding a layer of indirection.

-- 
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".