From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100]) by master.gitmailbox.com (Postfix) with ESMTP id E5ED746A39 for ; Sun, 2 Jul 2023 10:43:16 +0000 (UTC) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 4DAB068C210; Sun, 2 Jul 2023 13:43:13 +0300 (EEST) Received: from mail-wr1-f54.google.com (mail-wr1-f54.google.com [209.85.221.54]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 1B5A268BEFE for ; Sun, 2 Jul 2023 13:43:06 +0300 (EEST) Received: by mail-wr1-f54.google.com with SMTP id ffacd0b85a97d-3110ab7110aso4016695f8f.3 for ; Sun, 02 Jul 2023 03:43:06 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=kynesim-co-uk.20221208.gappssmtp.com; s=20221208; t=1688294586; x=1690886586; h=content-transfer-encoding:mime-version:user-agent:in-reply-to :references:message-id:date:subject:cc:to:from:from:to:cc:subject :date:message-id:reply-to; bh=XG+fvElL0ytvwrPRiI7Q0P//ayXydF518mhwVRhdaAA=; b=TXkhOzIQwLvEPTuijUo0YvWhxTTnTG5wz2Chr/L9eTkDct2lQQfEIUQCi/to1sVr9J /EBN9RGsWmlXcoFcqgV/ZZCBaJFHikneIHVnFrwFJmfYkWgAc+h/oFET5/BTn+Jn3dLb qlQGkxpwWrmndZLbXyJaii7mvi97TB7HRwVSmnOUBzy5MrURl2rdO5GP8CAD23k7HjPb sTHtUbDjw9ZarQfAJ3k9WUeeVt8mqdct0nVSQZtMMvPikdaxseQP7GMHNszBxttURoiT Tna1Ru6O1sTySuRjV28mHVZH8ZK0ab1LzAV1dt5UwxQOBU+9yAlGjB8L1r0FFoYnayKc iysg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1688294586; x=1690886586; h=content-transfer-encoding:mime-version:user-agent:in-reply-to :references:message-id:date:subject:cc:to:from:x-gm-message-state :from:to:cc:subject:date:message-id:reply-to; bh=XG+fvElL0ytvwrPRiI7Q0P//ayXydF518mhwVRhdaAA=; b=XE7bu4rCcAMADNIT5UekQDcTZSu/PqAWuE3Ax5QOwOKVPjon6aZroOXvzWAoWuY283 4zyY30NRzUUZFl/MQdtVkFzQzWcdyZdwYTwRNSKPx+BcMMiB2UZumzoWIX2HeZEXBpqH L9At/pOBaHxffHUW9gh6br2n/vdCUV+6N32eEDVv86hhR17T64JbWA1rXRUycUG8MRB2 ddWIbZZVHnA90M6TnxwfkBs1l6cub1XTaUYMRx8nRSqMV52MBAeKFIRdkL7vd+vgQKLd NgTfdcWIF9ZBPXkTIPzTtnNfT3PQ2puLZKRCk+Q4kwzmMGiLRtPdKCM4J/gjnxjVQ5J8 64nQ== X-Gm-Message-State: ABy/qLY5WkeAao3h2bZFtGAKS6Imx1PFFECoHXXg0m1FvABIVISEBgod 7oSO9Adt3etkKQLXfAYk3TJgLqWLw2gGqPMTUcc= X-Google-Smtp-Source: APBJJlHNBZ441GXhU57NdPP2UiPerpgtMFyj7iUOeJcFIfpuIlDWcqoHSocnY+GDbCzqW8X1cfl74w== X-Received: by 2002:adf:f3c3:0:b0:314:1261:4fb2 with SMTP id g3-20020adff3c3000000b0031412614fb2mr6269628wrp.31.1688294586151; Sun, 02 Jul 2023 03:43:06 -0700 (PDT) Received: from CTHALPA.outer.uphall.net (cpc1-cmbg20-2-0-cust759.5-4.cable.virginm.net. [86.21.218.248]) by smtp.gmail.com with ESMTPSA id g11-20020a7bc4cb000000b003fbab76165asm13236944wmk.48.2023.07.02.03.43.05 (version=TLS1 cipher=ECDHE-ECDSA-AES128-SHA bits=128/128); Sun, 02 Jul 2023 03:43:05 -0700 (PDT) From: John Cox To: FFmpeg development discussions and patches Date: Sun, 02 Jul 2023 11:43:04 +0100 Message-ID: <4dk2aitjmvbkevaogr2knejgvt293ksajc@4ax.com> References: <20230629175729.224383-1-jc@kynesim.co.uk> <20230629175729.224383-5-jc@kynesim.co.uk> In-Reply-To: User-Agent: ForteAgent/8.00.32.1272 MIME-Version: 1.0 Subject: Re: [FFmpeg-devel] [PATCH 04/15] avfilter/vf_bwdif: Add neon for filter_intra X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Cc: thomas.mundt@hr.de, FFmpeg development discussions and patches Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" Archived-At: List-Archive: List-Post: On Sun, 2 Jul 2023 00:37:35 +0300 (EEST), you wrote: >On Thu, 29 Jun 2023, John Cox wrote: > >> Signed-off-by: John Cox >> --- >> libavfilter/aarch64/vf_bwdif_init_aarch64.c | 17 +++++++ >> libavfilter/aarch64/vf_bwdif_neon.S | 53 +++++++++++++++++++++ >> 2 files changed, 70 insertions(+) >> >> diff --git a/libavfilter/aarch64/vf_bwdif_init_aarch64.c b/libavfilter/aarch64/vf_bwdif_init_aarch64.c >> index 86d53b2ca1..3ffaa07ab3 100644 >> --- a/libavfilter/aarch64/vf_bwdif_init_aarch64.c >> +++ b/libavfilter/aarch64/vf_bwdif_init_aarch64.c >> @@ -24,6 +24,22 @@ >> #include "libavfilter/bwdif.h" >> #include "libavutil/aarch64/cpu.h" >> >> +void ff_bwdif_filter_intra_neon(void *dst1, void *cur1, int w, int prefs, int mrefs, >> + int prefs3, int mrefs3, int parity, int clip_max); >> + >> + >> +static void filter_intra_helper(void *dst1, void *cur1, int w, int prefs, int mrefs, >> + int prefs3, int mrefs3, int parity, int clip_max) >> +{ >> + const int w0 = clip_max != 255 ? 0 : w & ~15; >> + >> + ff_bwdif_filter_intra_neon(dst1, cur1, w0, prefs, mrefs, prefs3, mrefs3, parity, clip_max); >> + >> + if (w0 < w) >> + ff_bwdif_filter_intra_c((char *)dst1 + w0, (char *)cur1 + w0, >> + w - w0, prefs, mrefs, prefs3, mrefs3, parity, clip_max); >> +} >> + >> void >> ff_bwdif_init_aarch64(BWDIFContext *s, int bit_depth) >> { >> @@ -35,5 +51,6 @@ ff_bwdif_init_aarch64(BWDIFContext *s, int bit_depth) >> if (!have_neon(cpu_flags)) >> return; >> >> + s->filter_intra = filter_intra_helper; >> } >> >> diff --git a/libavfilter/aarch64/vf_bwdif_neon.S b/libavfilter/aarch64/vf_bwdif_neon.S >> index a8f0ed525a..b863b3447d 100644 >> --- a/libavfilter/aarch64/vf_bwdif_neon.S >> +++ b/libavfilter/aarch64/vf_bwdif_neon.S >> @@ -69,3 +69,56 @@ coeffs: >> .hword 5570, 3801, 1016, -3801 // hf[0] = v0.h[2], -hf[1] = v0.h[5] >> .hword 5077, 981 // sp[0] = v0.h[6] >> >> +// ============================================================================ >> +// >> +// void ff_bwdif_filter_intra_neon( >> +// void *dst1, // x0 >> +// void *cur1, // x1 >> +// int w, // w2 >> +// int prefs, // w3 >> +// int mrefs, // w4 >> +// int prefs3, // w5 >> +// int mrefs3, // w6 >> +// int parity, // w7 unused >> +// int clip_max) // [sp, #0] unused > >This bit is great to have > >> + >> +function ff_bwdif_filter_intra_neon, export=1 >> + cmp w2, #0 >> + ble 99f >> + >> + ldr q0, coeffs >> + >> +// for (x = 0; x < w; x++) { >> +10: >> + >> +// interpol = (coef_sp[0] * (cur[mrefs] + cur[prefs]) - coef_sp[1] * (cur[mrefs3] + cur[prefs3])) >> 13; > >I guess the style with intermixed C code is a bit uncommon in our >assembly, but as long as it doesn't affect the overall code style I guess >we can keep it. I needed it to track where I was whilst writing the code & if I ever need to change it I'll be lost without it - so I, at least, rate it as valuable and in line3 where I am very tight on registers it was invaluable for keeping track of what referred to what. >> + ldr q31, [x1, w4, SXTW] >> + ldr q30, [x1, w3, SXTW] >> + ldr q29, [x1, w6, SXTW] >> + ldr q28, [x1, w5, SXTW] > >Don't use shouty uppercase SXTW here Will change. >> + >> + uaddl v20.8h, v31.8b, v30.8b >> + uaddl2 v21.8h, v31.16b, v30.16b >> + >> + UMULL4K v2, v3, v4, v5, v20, v21, v0.h[6] >> + >> + uaddl v20.8h, v29.8b, v28.8b >> + uaddl2 v21.8h, v29.16b, v28.16b >> + >> + UMLSL4K v2, v3, v4, v5, v20, v21, v0.h[7] >> + >> +// dst[0] = av_clip(interpol, 0, clip_max); >> + SQSHRUNN v2, v2, v3, v4, v5, 13 >> + str q2, [x0], #16 >> + >> +// dst++; >> +// cur++; >> +// } >> + >> + subs w2, w2, #16 >> + add x1, x1, #16 > >For in-order cores, it might be good to update these variables sometime >sooner, e.g. before the str instruction. But such scheduling breaks your >mapping between neat C code and assembly. I take your point but there is at least 1 instruction between set and use which is normally enough. I did my benching on a Pi4 and found, to my surprise, that in most cases reordering instructions to interleavse operations made life worse and seeing as Pi4 is my target platform I stopped trying to do that and stuck with code that was easier to read! (Also filter_intra seems to be much more memory b/w limited than code limited on a Pi4.) JC >// 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". _______________________________________________ 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".