From patchwork Tue Sep 20 11:01:56 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Hubert Mazur X-Patchwork-Id: 38071 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6a20:3b1c:b0:96:9ee8:5cfd with SMTP id c28csp1849191pzh; Tue, 20 Sep 2022 04:02:30 -0700 (PDT) X-Google-Smtp-Source: AMsMyM6KsHN4wZYXIAXzvgJ6lpmVT90JOAKJI3xlIhe50gDNfFkSDzojlxKWezQRMH9H/uAZ26Kp X-Received: by 2002:a05:6402:42d4:b0:443:4b8:8cba with SMTP id i20-20020a05640242d400b0044304b88cbamr19572269edc.3.1663671750468; Tue, 20 Sep 2022 04:02:30 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1663671750; cv=none; d=google.com; s=arc-20160816; b=PBnUgCukULFN4SiwSE4bOVgVwpLpqjAx61/1pX78+Yu4fddKliHiNFplRQkjCRCdtB So3AO9xKjFIFTp2AtTlrxz4UTexc12WAEuM46uH43Q0TXMvmFwLsdTrx0Eztbss1Rire GXefisdmeZh7DXEjQIwE96eoQjLfBZhW6pPECIIVB4GEhg3OvrAyL4i/NF4GAJW4Ce4m Xw4h0Yy31zmMrBwpuo+CnfrnmqeuVm78S9LpOBvah3yjjrU0Thm3iLIpK1cBgx9bRGBi L6dLvtzzavQupMbgLWm0QX1L1WPvB+gbK4T78eQT5nasBoyhcLacINFXj1wrykU6NsDy 8FzQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:cc:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:mime-version:references:in-reply-to :message-id:date:to:from:dkim-signature:delivered-to; bh=3BtYT9XSSl7OKvLdM1A7Q6lvSW0hf+Thq0NuFFqF7cc=; b=npWPmDP5iopmFiVNAL8hYTeB4hZ9qNZI5jYbO8q4x2No+3EGDFdM/P0cNg6+M9NOJo 4baHwqKdfefeEAV0inDmw8TBoEGox8nvrfN8fNR6fNGD0Urvx1yl7P8DfJ99FY+zyOed BB7JnUAA0c2ML/MC4RFww3Ik7QbKMO7IFt2T9/Tg+2PUWsHTOJF+e0S+7kO1hExiGAf1 Y0U63D/2rGj8NSKThCiPSgjNl2b3sOyDxYpBU7+i99BKJa3LcrKVMXMq5/W3caxqKmSI QWVyZ5ua4Qu5kwhlcXkcZjElLmlo6ruXBUI/HmuY3L7R3vxDCKvFQIj8vmW5mCKrKhWi yeiQ== ARC-Authentication-Results: i=1; mx.google.com; dkim=neutral (body hash did not verify) header.i=@semihalf.com header.s=google header.b=jWDWwQ9v; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=QUARANTINE sp=QUARANTINE dis=NONE) header.from=semihalf.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id p11-20020a170906604b00b0077ce17dac95si869171ejj.120.2022.09.20.04.02.30; Tue, 20 Sep 2022 04:02:30 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@semihalf.com header.s=google header.b=jWDWwQ9v; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=QUARANTINE sp=QUARANTINE dis=NONE) header.from=semihalf.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 346C068B9B4; Tue, 20 Sep 2022 14:02:21 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-lf1-f49.google.com (mail-lf1-f49.google.com [209.85.167.49]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id DDB5068B872 for ; Tue, 20 Sep 2022 14:02:14 +0300 (EEST) Received: by mail-lf1-f49.google.com with SMTP id o2so3099800lfc.10 for ; Tue, 20 Sep 2022 04:02:14 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=semihalf.com; s=google; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date; bh=dOJb3hvskaeaLNDoOXi+r9/dUBBXUOE1rbu4SRnrx5s=; b=jWDWwQ9v44xQiorCajFmjB0F6kdIECVaX2llcg8u2Onofs3dbp8p5GDXPWbdbKvLwD rrRKMpt28L/4gGdjWOFfVNvtNQ+L+yNzEHoJ7m5AmS5D6SiA7OCrFbhlap+6XNvNQVvS Vf3HvFP0bwjPWR0TxiB+zgWBgxVbTnn5p+UKVR4s8v7+SbpBTsDO8Y5b2cpqUDGhXx34 eK2bm/nlzbNdkv2s6EvzRxmqZpNOBc34CrsE80vkvu+ImVRALlRzP0IW1rPOH4Jq45Gj u5bgp05AT5JxlOf5/mWk8cUfDQ+bHU20BgffTFZSN352ydE7UOPC4uCxZk5RkIG5+W94 OFkA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-message-state:from:to:cc :subject:date; bh=dOJb3hvskaeaLNDoOXi+r9/dUBBXUOE1rbu4SRnrx5s=; b=xwiZ048mlJmgkyZfyC2cDgpzdlajf1kmUip77y3j8Un5QRf79VqqJ67CQnW94GZQ6J s3T+nXBlagnBMys6Y+J/QYNuWDZEklvYU5WpR/nH4WnWkwRHBm521g4hS2T+cj81H2Tw QrD2R8OID15plnFf0PZlw61cLjMPngZUH/k5IEHprEw4Mvm2ZaWuQVR2WAEzHnFmzR1D n6AGNr3+8PqkgRc6saOE6yFKw+SQ85hfBgpAANSmvWnqvOPyzSVXV9zcFYDL64FyHujP PaJM2yQKbnDqeNJW/wDrKttyTViYzWghBlhnN5Snz7tUNMcjjDG9JFxQ0aG7X2vk70Db 5qdA== X-Gm-Message-State: ACrzQf0wDe4a9jYTeZx3NKD5ADfZ9eeib1KAD9p3309PVyz3UACDVCli +UZtZG2ryfheAfL8NXNE93nEBbLq8mZ7wA== X-Received: by 2002:a05:6512:118b:b0:492:e3c4:a164 with SMTP id g11-20020a056512118b00b00492e3c4a164mr8615295lfr.598.1663671733807; Tue, 20 Sep 2022 04:02:13 -0700 (PDT) Received: from hum-HP-ProBook-440-G7.office.semihalf.net ([83.142.187.84]) by smtp.gmail.com with ESMTPSA id d3-20020a19f243000000b00492f5ad0ae7sm265220lfk.43.2022.09.20.04.02.11 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Tue, 20 Sep 2022 04:02:13 -0700 (PDT) From: Hubert Mazur To: ffmpeg-devel@ffmpeg.org Date: Tue, 20 Sep 2022 13:01:56 +0200 Message-Id: <20220920110158.15384-2-hum@semihalf.com> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20220920110158.15384-1-hum@semihalf.com> References: <20220920110158.15384-1-hum@semihalf.com> MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 1/3] lavc/aarch64: Add neon implementation for pix_median_abs16 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: gjb@semihalf.com, upstream@semihalf.com, jswinney@amazon.com, Hubert Mazur , martin@martin.st, mw@semihalf.com, spop@amazon.com Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: BgX9ViV9Mh4X Provide optimized implementation for pix_median_abs16 function. Performance comparison tests are shown below. - median_sad_0_c: 720.5 - median_sad_0_neon: 127.2 Benchmarks and tests run with checkasm tool on AWS Graviton 3. Signed-off-by: Hubert Mazur --- libavcodec/aarch64/me_cmp_init_aarch64.c | 4 ++ libavcodec/aarch64/me_cmp_neon.S | 78 ++++++++++++++++++++++++ libavcodec/me_cmp.c | 5 +- 3 files changed, 85 insertions(+), 2 deletions(-) diff --git a/libavcodec/aarch64/me_cmp_init_aarch64.c b/libavcodec/aarch64/me_cmp_init_aarch64.c index ade3e9a4c1..fb51a833be 100644 --- a/libavcodec/aarch64/me_cmp_init_aarch64.c +++ b/libavcodec/aarch64/me_cmp_init_aarch64.c @@ -53,6 +53,8 @@ int nsse16_neon(int multiplier, const uint8_t *s, const uint8_t *s2, ptrdiff_t stride, int h); int nsse16_neon_wrapper(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h); +int pix_median_abs16_neon(MpegEncContext *v, const uint8_t *pix1, const uint8_t *pix2, + ptrdiff_t stride, int h); av_cold void ff_me_cmp_init_aarch64(MECmpContext *c, AVCodecContext *avctx) { @@ -78,6 +80,8 @@ av_cold void ff_me_cmp_init_aarch64(MECmpContext *c, AVCodecContext *avctx) c->vsse[4] = vsse_intra16_neon; c->nsse[0] = nsse16_neon_wrapper; + + c->median_sad[0] = pix_median_abs16_neon; } } diff --git a/libavcodec/aarch64/me_cmp_neon.S b/libavcodec/aarch64/me_cmp_neon.S index f8998749a5..1e08d04869 100644 --- a/libavcodec/aarch64/me_cmp_neon.S +++ b/libavcodec/aarch64/me_cmp_neon.S @@ -969,3 +969,81 @@ function nsse16_neon, export=1 ret endfunc + +function pix_median_abs16_neon, export=1 + // x0 unused + // x1 uint8_t *pix1 + // x2 uint8_t *pix2 + // x3 ptrdiff_t stride + // w4 int h + + ld1 {v2.16b}, [x1], x3 + ld1 {v3.16b}, [x2], x3 + movi v31.8h, #0 + movi v16.8h, #0 + ext v0.16b, v2.16b, v2.16b, #1 + ext v1.16b, v3.16b, v3.16b, #1 + usubl v28.8h, v2.8b, v3.8b + usubl2 v27.8h, v2.16b, v3.16b + usubl v26.8h, v0.8b, v1.8b + usubl2 v25.8h, v0.16b, v1.16b + sub w4, w4, #1 // we need to make h-1 iterations + saba v31.8h, v26.8h, v28.8h + saba v16.8h, v25.8h, v27.8h + mov h18, v28.h[0] + cmp w4, #1 + sqabs h18, h18 + movi v0.8h, #0 + + b.lt 2f +1: + + ld1 {v6.16b}, [x1], x3 // pix1 vector for V(j-1) + ld1 {v7.16b}, [x2], x3 // pix2 vector for V(j-1) + subs w4, w4, #1 + ext v4.16b, v6.16b, v6.16b, #1 // pix1 vector for V(j) + ext v5.16b, v7.16b, v7.16b, #1 // pix2 vector for V(j) + + // protected registers: v30, v29, v28, v27, v26, v25, v24, v23 + // scratch registers: v22, v21, v20, v19, v17 + + // To find median of three values, calculate sum of them + // and subtract max and min value from it. + usubl v30.8h, v6.8b, v7.8b // V(j-1) + usubl2 v29.8h, v6.16b, v7.16b // V(j-1) + usubl v24.8h, v4.8b, v5.8b // V(j) + usubl2 v23.8h, v4.16b, v5.16b // V(j) + saba v0.8h, v30.8h, v28.8h + add v22.8h, v26.8h, v30.8h + smin v20.8h, v26.8h, v30.8h + add v21.8h, v25.8h, v29.8h + smax v19.8h, v26.8h, v30.8h + sub v22.8h, v22.8h, v28.8h + sub v21.8h, v21.8h, v27.8h + smin v17.8h, v19.8h, v22.8h + smin v22.8h, v25.8h, v29.8h + mov v28.16b, v30.16b + smax v20.8h, v20.8h, v17.8h // median values lower half + smax v19.8h, v25.8h, v29.8h + saba v31.8h, v24.8h, v20.8h + mov v27.16b, v29.16b + smin v19.8h, v19.8h, v21.8h + mov v26.16b, v24.16b + smax v17.8h, v22.8h, v19.8h // median values upper half + mov v25.16b, v23.16b + saba v16.8h, v23.8h, v17.8h + + b.ne 1b + +2: + mov h17, v0.h[0] + ins v16.h[7], wzr + add d18, d18, d17 + add v31.8h, v31.8h, v16.8h + uaddlv s17, v31.8h + add d18, d18, d17 + fmov w0, s18 + + ret + +endfunc diff --git a/libavcodec/me_cmp.c b/libavcodec/me_cmp.c index 4242fbc6e4..e2f9f84b05 100644 --- a/libavcodec/me_cmp.c +++ b/libavcodec/me_cmp.c @@ -1048,6 +1048,9 @@ av_cold void ff_me_cmp_init(MECmpContext *c, AVCodecContext *avctx) ff_dsputil_init_dwt(c); #endif + c->median_sad[0] = pix_median_abs16_c; + c->median_sad[1] = pix_median_abs8_c; + #if ARCH_AARCH64 ff_me_cmp_init_aarch64(c, avctx); #elif ARCH_ALPHA @@ -1062,6 +1065,4 @@ av_cold void ff_me_cmp_init(MECmpContext *c, AVCodecContext *avctx) ff_me_cmp_init_mips(c, avctx); #endif - c->median_sad[0] = pix_median_abs16_c; - c->median_sad[1] = pix_median_abs8_c; }