From: Niklas Haas <ffmpeg@haasn.xyz> To: ffmpeg-devel@ffmpeg.org Cc: Niklas Haas <git@haasn.dev> Subject: Re: [FFmpeg-devel] [PATCH v4 13/13] avfilter/x86/f_ebur128: implement AVX true peak calculation Date: Tue, 17 Jun 2025 15:48:40 +0200 Message-ID: <20250617154840.GB2472956@haasn.xyz> (raw) In-Reply-To: <20250617120609.1987435-13-ffmpeg@haasn.xyz> On Tue, 17 Jun 2025 14:06:06 +0200 Niklas Haas <ffmpeg@haasn.xyz> wrote: > From: Niklas Haas <git@haasn.dev> > > Stereo only, for simplicity. Slightly faster than the C code. I will drop this and the prior commit and resubmit them in a refactored form that also speeds up the sample peak calculation. In the meantime, I would appreciate a review of the rest. > --- > libavfilter/f_ebur128.c | 8 +++----- > libavfilter/f_ebur128.h | 2 +- > libavfilter/x86/f_ebur128.asm | 25 +++++++++++++++++++++++++ > libavfilter/x86/f_ebur128_init.c | 9 +++++++-- > 4 files changed, 36 insertions(+), 8 deletions(-) > > diff --git a/libavfilter/f_ebur128.c b/libavfilter/f_ebur128.c > index 0adc89c823..c64f6ed032 100644 > --- a/libavfilter/f_ebur128.c > +++ b/libavfilter/f_ebur128.c > @@ -502,6 +502,9 @@ static int config_audio_output(AVFilterLink *outlink) > return AVERROR(ENOMEM); > } > > +#if ARCH_X86 > + ff_ebur128_init_x86(&ebur128->dsp, nb_channels); > +#endif > return 0; > } > > @@ -581,11 +584,6 @@ static av_cold int init(AVFilterContext *ctx) > > ebur128->dsp.filter_channels = ff_ebur128_filter_channels_c; > ebur128->dsp.true_peak = ff_ebur128_true_peak_c; > - > -#if ARCH_X86 > - ff_ebur128_init_x86(&ebur128->dsp); > -#endif > - > return 0; > } > > diff --git a/libavfilter/f_ebur128.h b/libavfilter/f_ebur128.h > index 8aab7838a0..5fb9d4a8d5 100644 > --- a/libavfilter/f_ebur128.h > +++ b/libavfilter/f_ebur128.h > @@ -53,7 +53,7 @@ static_assert(offsetof(EBUR128DSPContext, pre) == 0, "struct l > static_assert(offsetof(EBUR128DSPContext, rlb) == 5 * sizeof(double), "struct layout mismatch"); > static_assert(offsetof(EBUR128DSPContext, y) == 10 * sizeof(double), "struct layout mismatch"); > > -void ff_ebur128_init_x86(EBUR128DSPContext *dsp); > +void ff_ebur128_init_x86(EBUR128DSPContext *dsp, int nb_channels); > > void ff_ebur128_filter_channels_c(const EBUR128DSPContext *, const double *, > double *, double *, double *, double *, int); > diff --git a/libavfilter/x86/f_ebur128.asm b/libavfilter/x86/f_ebur128.asm > index d9cc8d9361..53dd3f858d 100644 > --- a/libavfilter/x86/f_ebur128.asm > +++ b/libavfilter/x86/f_ebur128.asm > @@ -39,6 +39,10 @@ struc DSP > .z resq 1 > endstruc > > +SECTION_RODATA > + > +abs_mask: dq 0x7FFFFFFFFFFFFFFF > + > SECTION .text > > %macro MOVNQ 3 ; num, dst, src > @@ -139,3 +143,24 @@ cglobal ebur128_filter_channels, 7, 9, 14, dsp, samples, cache400, cache3000, su > test channelsd, channelsd > jnz .loop > RET > + > +cglobal ebur128_true_peak_stereo, 5, 6, 4, tp, tppf, channels, samples, nb_samples > + vpbroadcastq m4, [abs_mask] > + pxor m0, m0 ; maxpeak > + movupd m1, [tpq] ; tp > + pxor m2, m2 ; tppf > +.inner: > + movupd m3, [samplesq] > + add samplesq, 16 > + pand m3, m4 > + maxpd m1, m3 > + maxpd m2, m3 > + dec nb_samplesd > + jg .inner > + movupd [tpq], m1 > + movupd [tppfq], m2 > + maxpd m0, m1 > + shufpd m1, m0, m0, 1 > + maxpd m0, m1 > + movq rax, m0 > + RET > diff --git a/libavfilter/x86/f_ebur128_init.c b/libavfilter/x86/f_ebur128_init.c > index 8f38aee967..527e5e4dbc 100644 > --- a/libavfilter/x86/f_ebur128_init.c > +++ b/libavfilter/x86/f_ebur128_init.c > @@ -26,10 +26,15 @@ > void ff_ebur128_filter_channels_avx(const EBUR128DSPContext *, const double *, > double *, double *, double *, double *, int); > > -av_cold void ff_ebur128_init_x86(EBUR128DSPContext *dsp) > +double ff_ebur128_true_peak_stereo_avx(double *, double *, int, const double *, int); > + > +av_cold void ff_ebur128_init_x86(EBUR128DSPContext *dsp, int nb_channels) > { > int cpu_flags = av_get_cpu_flags(); > > - if (ARCH_X86_64 && EXTERNAL_AVX(cpu_flags)) > + if (ARCH_X86_64 && EXTERNAL_AVX(cpu_flags)) { > dsp->filter_channels = ff_ebur128_filter_channels_avx; > + if (nb_channels == 2) > + dsp->true_peak = ff_ebur128_true_peak_stereo_avx; > + } > } > -- > 2.49.0 > _______________________________________________ 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".
prev parent reply other threads:[~2025-06-17 13:48 UTC|newest] Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top 2025-06-17 12:05 [FFmpeg-devel] [PATCH v4 01/13] avfilter/f_ebur128: use transformed direct form II Niklas Haas 2025-06-17 12:05 ` [FFmpeg-devel] [PATCH v4 02/13] avfilter/f_ebur128: simplify sample cache array Niklas Haas 2025-06-17 12:05 ` [FFmpeg-devel] [PATCH v4 03/13] avfilter/f_ebur128: use structs for biquad weights Niklas Haas 2025-06-17 12:05 ` [FFmpeg-devel] [PATCH v4 04/13] avfilter/f_ebur128: use a single packed array for the integrator cache Niklas Haas 2025-06-17 12:05 ` [FFmpeg-devel] [PATCH v4 05/13] avfilter/f_ebur128: move weights and cache to EBUR128DSPContext Niklas Haas 2025-06-17 12:05 ` [FFmpeg-devel] [PATCH v4 06/13] avfilter/f_ebur128: split off C implementation to separate function Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 07/13] avfilter/x86/f_ebur128: add x86 AVX implementation Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 08/13] avfilter/f_ebur128: remove pointless macro Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 09/13] avfilter/f_ebur128: move true peak calculation out of main loop Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 10/13] avfilter/f_ebur128: lift sample " Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 11/13] avfilter/f_ebur128: move variable declarations to usage site Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 12/13] avfilter/f_ebur128: move true peak calculation to DSP function Niklas Haas 2025-06-17 12:06 ` [FFmpeg-devel] [PATCH v4 13/13] avfilter/x86/f_ebur128: implement AVX true peak calculation Niklas Haas 2025-06-17 13:48 ` Niklas Haas [this message]
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20250617154840.GB2472956@haasn.xyz \ --to=ffmpeg@haasn.xyz \ --cc=ffmpeg-devel@ffmpeg.org \ --cc=git@haasn.dev \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel This inbox may be cloned and mirrored by anyone: git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \ ffmpegdev@gitmailbox.com public-inbox-index ffmpegdev Example config snippet for mirrors. AGPL code for this site: git clone https://public-inbox.org/public-inbox.git