From mboxrd@z Thu Jan 1 00:00:00 1970
Return-Path: <ffmpeg-devel-bounces@ffmpeg.org>
Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100])
by master.gitmailbox.com (Postfix) with ESMTPS id D91104B524
for <ffmpegdev@gitmailbox.com>; Fri, 25 Apr 2025 08:25:40 +0000 (UTC)
Received: from [127.0.1.1] (localhost [127.0.0.1])
by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 410B6687DB9;
Fri, 25 Apr 2025 11:25:36 +0300 (EEST)
Received: from mail-lf1-f45.google.com (mail-lf1-f45.google.com
[209.85.167.45])
by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 09E58687C94
for <ffmpeg-devel@ffmpeg.org>; Fri, 25 Apr 2025 11:25:29 +0300 (EEST)
Received: by mail-lf1-f45.google.com with SMTP id
2adb3069b0e04-5499614d3d2so2037666e87.3
for <ffmpeg-devel@ffmpeg.org>; Fri, 25 Apr 2025 01:25:29 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;
d=martin-st.20230601.gappssmtp.com; s=20230601; t=1745569528; x=1746174328;
darn=ffmpeg.org;
h=mime-version:references:message-id:in-reply-to:subject:cc:to:from
:date:from:to:cc:subject:date:message-id:reply-to;
bh=MlM/KdB54k0c5AS5/69PSCGHoCPmw3fL7tgs1Sbb3vI=;
b=1L3Y4RW4lX4haQY3RhXrnKwBNAM2OKutmAGqmojNqeTsE1CCyNKhX69kgDLi8f8XHO
x6NdBsclh06e8uTSL4QDCmJwBPGgg+mgz0RIhdZVi7DFwABVMcq22W5P2lek3wFBdT+P
2b1jj+ENVSAY0eZiaix60KdcdYRd8AELgmNAs72GjJ3EnVIC2RJeFf4CiAq2vgdSddaL
hcsLCC1ViVxLTH82Wj4eFEJgjM9MyEiDqBMyeX3BYyTFbspRH6OSjwttsTyQ1btb+kJz
HqZXSzky/3rmQiMzL/0eNECpZikU076CQGr4RERM/CDypjTcEZlA14xCvr6lsBDDyOcF
8gyw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;
d=1e100.net; s=20230601; t=1745569528; x=1746174328;
h=mime-version:references:message-id:in-reply-to:subject:cc:to:from
:date:x-gm-message-state:from:to:cc:subject:date:message-id:reply-to;
bh=MlM/KdB54k0c5AS5/69PSCGHoCPmw3fL7tgs1Sbb3vI=;
b=J6bf0WmxFc3aUm/8s5pMobU71uZ9xz3g4KNV1CKnfqjlgGBtnvSBe0diGe91as009j
XG33Hl3MLWFGjpFJt6np4Z+UkEabk91W4xUnMSTkUkbCfUuvqJVYgv1R/HArOaWbcr5s
wq3vm0uniuGJykRciK5GJL9g5hLiKiR/EEMiz6zE5PvmBU6asqhmGhMGcJz3TOPD16lY
yBBSBta6DSiClduIvSj7La6f0AP2+v+NQwJ02mO6NtuGsFc3fvFQ88bTpM+31QeZyAV0
ufcekbOc5IxT7JQ1LaFcCc5Dj/WFKoHvzYtV4BaonBtgGoicTCEp3HgcCRCr9Jplwxto
8f7A==
X-Gm-Message-State: AOJu0YyHkrpZd9epKAVzpzJ+4qAYUCgB3oayquHocXiLMh1aCMJ0upE1
CaufJ+DKYBacF5GgVjOA12+FaqROLGMplIBSLuyfNjpfbafJ7hbgfDMUN0hLPLXc7ZgBjcg/NxS
Zfw==
X-Gm-Gg: ASbGnctootBTB0pMPgOlki/aY6zkoYJxU1SzVOJETdsDDsSGr/c8xrOvvV9LDOD/yr8
cKdfm5Kmz+P7cN2VO03KBqJ6d8MVA+yRncApUsWSTzuFacQQgRCvsWaGu38dhyhr5fMHZByLi7B
iZlmsdNtIk4QbLA5oytQQQ0oAuV9KjZjBk+hyqSLZqMjoYYEjjt9BBm2c2nCXxIhJzpIRC333al
zd/aey42vmU5rPMZfmFLhjfaI/9ptzaLVgjozc8Sys7x5St9txS/7EHl6+e1TS1wWWsa56XJJbg
FXguQVdQutrZqLg8Li6m7MnNrXOzpa5WP51bQoxc4BRighYdggh1OjC38NZb0LNJyagcXV+gkLh
1PaTudh++fqEx2uiMcfwe8FzHlRmzyaIOK1+z
X-Google-Smtp-Source: AGHT+IHzZ8NDH7wN9rw2cbPsY4JiGdNOMxFDaS8dQz0kqQxMmcLaNVYRyUnQm4vy1Zkj/PSS3mR47Q==
X-Received: by 2002:a05:6512:1321:b0:54d:67d7:c530 with SMTP id
2adb3069b0e04-54e8cc10e76mr410508e87.56.1745569527906;
Fri, 25 Apr 2025 01:25:27 -0700 (PDT)
Received: from tunnel335574-pt.tunnel.tserv24.sto1.ipv6.he.net
(tunnel335574-pt.tunnel.tserv24.sto1.ipv6.he.net. [2001:470:27:11::2])
by smtp.gmail.com with ESMTPSA id
2adb3069b0e04-54e7ccaa959sm520226e87.200.2025.04.25.01.25.27
(version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256);
Fri, 25 Apr 2025 01:25:27 -0700 (PDT)
Date: Fri, 25 Apr 2025 11:25:25 +0300 (EEST)
From: =?ISO-8859-15?Q?Martin_Storsj=F6?= <martin@martin.st>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
In-Reply-To: <tencent_B6A71838849627CD75A00E4DBC4FBA556B07@qq.com>
Message-ID: <c405429-f2b4-3764-4e-fdef3aa91795@martin.st>
References: <tencent_B6A71838849627CD75A00E4DBC4FBA556B07@qq.com>
MIME-Version: 1.0
Subject: Re: [FFmpeg-devel] [PATCH] aarch64/h26x: optimize sao_band_filter
X-BeenThere: ffmpeg-devel@ffmpeg.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: FFmpeg development discussions and patches <ffmpeg-devel.ffmpeg.org>
List-Unsubscribe: <https://ffmpeg.org/mailman/options/ffmpeg-devel>,
<mailto:ffmpeg-devel-request@ffmpeg.org?subject=unsubscribe>
List-Archive: <https://ffmpeg.org/pipermail/ffmpeg-devel>
List-Post: <mailto:ffmpeg-devel@ffmpeg.org>
List-Help: <mailto:ffmpeg-devel-request@ffmpeg.org?subject=help>
List-Subscribe: <https://ffmpeg.org/mailman/listinfo/ffmpeg-devel>,
<mailto:ffmpeg-devel-request@ffmpeg.org?subject=subscribe>
Reply-To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Cc: Zhao Zhili <zhilizhao@tencent.com>
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset="us-ascii"; Format="flowed"
Errors-To: ffmpeg-devel-bounces@ffmpeg.org
Sender: "ffmpeg-devel" <ffmpeg-devel-bounces@ffmpeg.org>
Archived-At: <https://master.gitmailbox.com/ffmpegdev/c405429-f2b4-3764-4e-fdef3aa91795@martin.st/>
List-Archive: <https://master.gitmailbox.com/ffmpegdev/>
List-Post: <mailto:ffmpegdev@gitmailbox.com>
On Tue, 15 Apr 2025, Zhao Zhili wrote:
> From: Zhao Zhili <zhilizhao@tencent.com>
>
> int8_t[] is enough for offset_table of 8 bit streams.
>
> On rpi5:
> Before After
> hevc_sao_band_8_8_c: 252.3 ( 1.00x) 252.3 ( 1.00x)
> hevc_sao_band_8_8_neon: 95.8 ( 2.63x) 61.0 ( 4.14x)
> hevc_sao_band_16_8_c: 875.2 ( 1.00x) 864.9 ( 1.00x)
> hevc_sao_band_16_8_neon: 317.5 ( 2.76x) 150.0 ( 5.76x)
> hevc_sao_band_32_8_c: 3853.5 ( 1.00x) 3871.6 ( 1.00x)
> hevc_sao_band_32_8_neon: 1222.3 ( 3.15x) 550.6 ( 7.03x)
> hevc_sao_band_48_8_c: 8203.6 ( 1.00x) 8182.6 ( 1.00x)
> hevc_sao_band_48_8_neon: 2685.7 ( 3.05x) 1185.8 ( 6.90x)
> hevc_sao_band_64_8_c: 14023.0 ( 1.00x) 14038.9 ( 1.00x)
> hevc_sao_band_64_8_neon: 4783.2 ( 2.93x) 2078.4 ( 6.75x)
> ---
> libavcodec/aarch64/h26x/dsp.h | 4 +
> libavcodec/aarch64/h26x/sao_neon.S | 93 ++++++++++++++---------
> libavcodec/aarch64/hevcdsp_init_aarch64.c | 4 +-
> libavcodec/aarch64/vvc/dsp_init.c | 5 +-
> 4 files changed, 65 insertions(+), 41 deletions(-)
>
> diff --git a/libavcodec/aarch64/h26x/dsp.h b/libavcodec/aarch64/h26x/dsp.h
> index 0fefb4d70f..6ea6a8d36a 100644
> --- a/libavcodec/aarch64/h26x/dsp.h
> +++ b/libavcodec/aarch64/h26x/dsp.h
> @@ -28,6 +28,10 @@ void ff_h26x_sao_band_filter_8x8_8_neon(uint8_t *_dst, const uint8_t *_src,
> ptrdiff_t stride_dst, ptrdiff_t stride_src,
> const int16_t *sao_offset_val, int sao_left_class,
> int width, int height);
> +void ff_h26x_sao_band_filter_16x16_8_neon(uint8_t *_dst, const uint8_t *_src,
> + ptrdiff_t stride_dst, ptrdiff_t stride_src,
> + const int16_t *sao_offset_val, int sao_left_class,
> + int width, int height);
> void ff_hevc_sao_edge_filter_16x16_8_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride_dst,
> const int16_t *sao_offset_val, int eo, int width, int height);
> void ff_hevc_sao_edge_filter_8x8_8_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride_dst,
> diff --git a/libavcodec/aarch64/h26x/sao_neon.S b/libavcodec/aarch64/h26x/sao_neon.S
> index c43820135e..60c026fe95 100644
> --- a/libavcodec/aarch64/h26x/sao_neon.S
> +++ b/libavcodec/aarch64/h26x/sao_neon.S
> @@ -35,48 +35,67 @@
> // int16_t *sao_offset_val, int sao_left_class,
> // int width, int height)
> function ff_h26x_sao_band_filter_8x8_8_neon, export=1
> - stp xzr, xzr, [sp, #-64]!
> + stp xzr, xzr, [sp, #-32]!
> stp xzr, xzr, [sp, #16]
> - stp xzr, xzr, [sp, #32]
> - stp xzr, xzr, [sp, #48]
> mov w8, #4
> -0: ldrsh x9, [x4, x8, lsl #1] // sao_offset_val[k+1]
> - subs w8, w8, #1
> - add w10, w8, w5 // k + sao_left_class
> +0:
> + ldrsh x9, [x4, x8, lsl #1] // sao_offset_val[k+1]
> + subs w8, w8, #1
> + add w10, w8, w5 // k + sao_left_class
> and w10, w10, #0x1F
> - strh w9, [sp, x10, lsl #1]
> + strb w9, [sp, x10]
> bne 0b
> - add w6, w6, #7
> - bic w6, w6, #7
> - ld1 {v16.16b-v19.16b}, [sp], #64
> - sub x2, x2, x6
> - sub x3, x3, x6
> - movi v20.8h, #1
> -1: mov w8, w6 // beginning of line
> -2: // Simple layout for accessing 16bit values
> - // with 8bit LUT.
> - //
> - // 00 01 02 03 04 05 06 07
> - // +----------------------------------->
> - // |xDE#xAD|xCA#xFE|xBE#xEF|xFE#xED|....
> - // +----------------------------------->
> - // i-0 i-1 i-2 i-3
> - ld1 {v2.8b}, [x1], #8 // dst[x] = av_clip_pixel(src[x] + offset_table[src[x] >> shift]);
> - subs w8, w8, #8
> - uxtl v0.8h, v2.8b // load src[x]
> - ushr v2.8h, v0.8h, #3 // >> BIT_DEPTH - 3
> - shl v1.8h, v2.8h, #1 // low (x2, accessing short)
> - add v3.8h, v1.8h, v20.8h // +1 access upper short
> - sli v1.8h, v3.8h, #8 // shift insert index to upper byte
> - tbx v2.16b, {v16.16b-v19.16b}, v1.16b // table
> - add v1.8h, v0.8h, v2.8h // src[x] + table
> - sqxtun v4.8b, v1.8h // clip + narrow
> - st1 {v4.8b}, [x0], #8 // store
> - // done 8 pixels
> + ldp q16, q17, [sp], #32
> +1:
> + ld1 {v2.8b}, [x1], x3
> + subs w7, w7, #1
> + uxtl v0.8h, v2.8b
> + ushr v3.8b, v2.8b, #3 // >> BIT_DEPTH - 3
Nitpick: The comment on this line seems to be misaligned with the other
comments below - please check.
> + tbx v3.8b, {v16.16b-v17.16b}, v3.8b
Is there any specific reason for preferring tbx over tbl here? (I know the
existing code used tbx.) Without having studied cycle tables, I would
expect tbl to maybe be slightly simpler, but perhaps there's no difference
(or tbx is faster)?
Other than these comments, this patch looks good to me, thanks - feel free
to push.
// Martin
_______________________________________________
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".