From patchwork Sat Nov 18 02:06:46 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Logan.Lyu" X-Patchwork-Id: 44703 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6a20:6a89:b0:181:818d:5e7f with SMTP id bi9csp293191pzb; Fri, 17 Nov 2023 18:07:12 -0800 (PST) X-Google-Smtp-Source: AGHT+IETbmECOQ3yaWqilRlgDdsYUBDczm2fQ8BnRnV4Gno5DT2s2BL5ApP3GFYmH4hRerZ0IJkE X-Received: by 2002:aa7:c513:0:b0:53f:6ed5:4dab with SMTP id o19-20020aa7c513000000b0053f6ed54dabmr664918edq.24.1700273232445; Fri, 17 Nov 2023 18:07:12 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1700273232; cv=none; d=google.com; s=arc-20160816; b=iPCs+Q5Gr/qyLRx0ZUgayvYlyG3w5gx/2/bCjbTHjpaqucX07wFFFvBcIJ6IBwlgnV ixKxbjbDzBMZN5YEJ/8RofpkQgxdWV8kP2ed5vsOSJv+kbLwYGpkJkCea1udqP6UvzH0 KcnLDkIF/TmokFEmIulgHL9dAId8iTTalruwQUbKGxTnsLrCiU1wDC6e6muzlEN2tf9a pua/ZiGKyeLUrQYkX01bOg9pwNqNJSXXTLQ8+bbd83TOApiRo65HMcMMf/miRaJXNUQ6 cut4TkdnZEutBO1n9Eq3dnFFJ7HNjJxOgXz7pYOiEKTFBI7lGouy58CPW6bCknRiMR6P BfBw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:cc:reply-to:list-subscribe:list-help:list-post :list-archive:list-unsubscribe:list-id:precedence:subject :organization:to:from:user-agent:mime-version:date:message-id :delivered-to; bh=f+5ILYXjPI3g/MB3fkSwDfFd/XyTsonVSyggfFn23Nw=; fh=4VBelKDE4DH3L7jF6H/1Jmu78FdN+YP76yfdJCQTJ30=; b=wlnzD2Z03wPXsbmdmWcs6FvRMviLtm/SU3AhoLrsO8GTTM7IfeS1L/ANIA+DnpjRHZ 0m3qLg8o9TFL1vGVyVqGLcikYJlneFY7ZqmnYPpYBG50nLXCb+RHfDbdPtRWtgFvPl3F B414EuJz1a37649aJ3ccEpGcHUSoyLEvRmUfoxW/8srNyx/oVKWspEoO5rVwRZbD2why CvsT7/gZl1CeGfQ1lgl1yhzxPdU7oOGBu79okOKZlCmcVmRWxWeJEgZA/+XHkN0qngOw zOJu3rUiSTQwFlaa6Cqfr1hqtI8dInsCDXfmmvOonD4TMDf+31YHBmuyfWb3vOFxputX YvKA== ARC-Authentication-Results: i=1; mx.google.com; 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 Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id x18-20020aa7dad2000000b0053b32146616si1609638eds.665.2023.11.17.18.07.12; Fri, 17 Nov 2023 18:07:12 -0800 (PST) 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; 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 Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 365E168CD3C; Sat, 18 Nov 2023 04:06:59 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from smtp-my3-01p8.yunyou.top (smtp-my3-01p8.yunyou.top [60.247.169.8]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id ADD1368CCFB for ; Sat, 18 Nov 2023 04:06:51 +0200 (EET) Received: from [192.168.15.105] (unknown [122.233.146.151]) by smtp-my-01.yunyou.top (WestCloudMail) with ESMTPA id 3CDDF141F26; Sat, 18 Nov 2023 10:06:46 +0800 (CST) Message-ID: <5969017d-81db-470d-91bd-37c5b02ae5c1@myais.com.cn> Date: Sat, 18 Nov 2023 10:06:46 +0800 MIME-Version: 1.0 User-Agent: Mozilla Thunderbird From: "Logan.Lyu" To: ffmpeg-devel@ffmpeg.org Organization: myais Subject: [FFmpeg-devel] [PATCH 2/6] lavc/aarch64: new optimization for 8-bit hevc_epel_bi_h 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: jdek@itanimul.li Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: B/JT41wPj4lp put_hevc_epel_bi_h4_8_c: 96.0 put_hevc_epel_bi_h4_8_neon: 36.3 put_hevc_epel_bi_h6_8_c: 288.3 put_hevc_epel_bi_h6_8_neon: 59.3 put_hevc_epel_bi_h8_8_c: 358.5 put_hevc_epel_bi_h8_8_neon: 61.5 put_hevc_epel_bi_h12_8_c: 759.8 put_hevc_epel_bi_h12_8_neon: 159.5 put_hevc_epel_bi_h16_8_c: 1307.0 put_hevc_epel_bi_h16_8_neon: 182.0 put_hevc_epel_bi_h24_8_c: 2778.3 put_hevc_epel_bi_h24_8_neon: 430.5 put_hevc_epel_bi_h32_8_c: 4952.3 put_hevc_epel_bi_h32_8_neon: 679.5 put_hevc_epel_bi_h48_8_c: 11803.3 put_hevc_epel_bi_h48_8_neon: 1443.5 put_hevc_epel_bi_h64_8_c: 20654.8 put_hevc_epel_bi_h64_8_neon: 2737.0 put_hevc_qpel_bi_h4_8_c: 140.0 put_hevc_qpel_bi_h4_8_neon: 111.5 put_hevc_qpel_bi_h6_8_c: 318.0 put_hevc_qpel_bi_h6_8_neon: 85.8 put_hevc_qpel_bi_h8_8_c: 536.5 put_hevc_qpel_bi_h8_8_neon: 95.3 put_hevc_qpel_bi_h12_8_c: 1188.5 put_hevc_qpel_bi_h12_8_neon: 291.3 put_hevc_qpel_bi_h16_8_c: 2064.3 put_hevc_qpel_bi_h16_8_neon: 365.3 put_hevc_qpel_bi_h24_8_c: 4757.5 put_hevc_qpel_bi_h24_8_neon: 1010.0 put_hevc_qpel_bi_h32_8_c: 8351.8 put_hevc_qpel_bi_h32_8_neon: 2917.8 put_hevc_qpel_bi_h48_8_c: 19299.8 put_hevc_qpel_bi_h48_8_neon: 2976.8 put_hevc_qpel_bi_h64_8_c: 34182.5 put_hevc_qpel_bi_h64_8_neon: 5236.3 Co-Authored-By: J. Dekker Signed-off-by: Logan Lyu --- libavcodec/aarch64/hevcdsp_epel_neon.S | 257 ++++++++++++++++++++++ libavcodec/aarch64/hevcdsp_init_aarch64.c | 5 + 2 files changed, 262 insertions(+) NEON8_FNASSIGN(c->put_hevc_epel_uni, 1, 0, epel_uni_v,); diff --git a/libavcodec/aarch64/hevcdsp_epel_neon.S b/libavcodec/aarch64/hevcdsp_epel_neon.S index 74165273d7..070c8c9990 100644 --- a/libavcodec/aarch64/hevcdsp_epel_neon.S +++ b/libavcodec/aarch64/hevcdsp_epel_neon.S @@ -423,6 +423,263 @@ function ff_hevc_put_hevc_pel_bi_pixels64_8_neon, export=1 ret endfunc + +function ff_hevc_put_hevc_epel_bi_h4_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ld1 {v4.8b}, [x2], x3 + ext v5.8b, v4.8b, v4.8b, #1 + ext v6.8b, v4.8b, v4.8b, #2 + ext v7.8b, v4.8b, v4.8b, #3 + calc_epelb v16, v4, v5, v6, v7 + ld1 {v20.4h}, [x4], x10 + sqadd v16.8h, v16.8h, v20.8h + sqrshrun v4.8b, v16.8h, #7 + st1 {v4.s}[0], [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h6_8_neon, export=1 + load_epel_filterb x6, x7 + sub w1, w1, #4 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ld1 {v24.16b}, [x2], x3 + ext v26.16b, v24.16b, v24.16b, #1 + ext v27.16b, v24.16b, v24.16b, #2 + ext v28.16b, v24.16b, v24.16b, #3 + calc_epelb v16, v24, v26, v27, v28 + ld1 {v20.8h}, [x4], x10 + sqadd v16.8h, v16.8h, v20.8h + sqrshrun v16.8b, v16.8h, #7 + st1 {v16.s}[0], [x0], #4 + st1 {v16.h}[2], [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h8_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ld1 {v24.16b}, [x2], x3 + ext v26.16b, v24.16b, v24.16b, #1 + ext v27.16b, v24.16b, v24.16b, #2 + ext v28.16b, v24.16b, v24.16b, #3 + calc_epelb v16, v24, v26 v27, v28 + ld1 {v20.8h}, [x4], x10 + sqadd v16.8h, v16.8h, v20.8h + sqrshrun v16.8b, v16.8h, #7 + st1 {v16.8b}, [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h12_8_neon, export=1 + load_epel_filterb x6, x7 + sub x1, x1, #8 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ld1 {v24.16b}, [x2], x3 + ext v26.16b, v24.16b, v24.16b, #1 + ext v27.16b, v24.16b, v24.16b, #2 + ext v28.16b, v24.16b, v24.16b, #3 + calc_epelb v16, v24, v26, v27, v28 + calc_epelb2 v17, v24, v26, v27, v28 + ld1 {v20.8h, v21.8h}, [x4], x10 + sqadd v18.8h, v16.8h, v20.8h + sqadd v19.8h, v17.8h, v21.8h + sqrshrun v20.8b, v18.8h, #7 + sqrshrun v21.8b, v19.8h, #7 + st1 {v20.8b}, [x0], #8 + st1 {v21.s}[0], [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h16_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ldr q24, [x2] + ldr s25, [x2, #16] + add x2, x2, x3 + ext v26.16b, v24.16b, v25.16b, #1 + ext v27.16b, v24.16b, v25.16b, #2 + ext v28.16b, v24.16b, v25.16b, #3 + calc_epelb v16, v24, v26, v27, v28 + calc_epelb2 v17, v24, v26, v27, v28 + ld1 {v24.8h, v25.8h}, [x4], x10 + sqadd v16.8h, v16.8h, v24.8h + sqadd v17.8h, v17.8h, v25.8h + sqrshrun v4.8b, v16.8h, #7 + sqrshrun2 v4.16b, v17.8h, #7 + st1 {v4.16b}, [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h24_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ld1 {v24.16b, v25.16b}, [x2], x3 + ext v26.16b, v24.16b, v25.16b, #1 + ext v27.16b, v24.16b, v25.16b, #2 + ext v28.16b, v24.16b, v25.16b, #3 + calc_epelb v16, v24, v26, v27, v28 + calc_epelb2 v17, v24, v26, v27, v28 + ext v26.16b, v25.16b, v25.16b, #1 + ext v27.16b, v25.16b, v25.16b, #2 + ext v28.16b, v25.16b, v25.16b, #3 + calc_epelb v18, v25, v26, v27, v28 + ld1 {v20.8h, v21.8h, v22.8h}, [x4], x10 + sqadd v16.8h, v16.8h, v20.8h + sqadd v17.8h, v17.8h, v21.8h + sqadd v18.8h, v18.8h, v22.8h + sqrshrun v4.8b, v16.8h, #7 + sqrshrun v5.8b, v17.8h, #7 + sqrshrun v6.8b, v18.8h, #7 + st1 {v4.8b, v5.8b, v6.8b}, [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h32_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 + mov x10, #(MAX_PB_SIZE * 2) +1: ldp q24, q25, [x2] + ldr s26, [x2, #32] + add x2, x2, x3 + ext v27.16b, v24.16b, v25.16b, #1 + ext v28.16b, v24.16b, v25.16b, #2 + ext v29.16b, v24.16b, v25.16b, #3 + calc_epelb v16, v24, v27, v28, v29 + calc_epelb2 v17, v24, v27, v28, v29 + ext v27.16b, v25.16b, v26.16b, #1 + ext v28.16b, v25.16b, v26.16b, #2 + ext v29.16b, v25.16b, v26.16b, #3 + calc_epelb v18, v25, v27, v28, v29 + calc_epelb2 v19, v25, v27, v28, v29 + ld1 {v24.8h, v25.8h, v26.8h, v27.8h}, [x4], x10 + sqadd v16.8h, v16.8h, v24.8h + sqadd v17.8h, v17.8h, v25.8h + sqadd v18.8h, v18.8h, v26.8h + sqadd v19.8h, v19.8h, v27.8h + sqrshrun v4.8b, v16.8h, #7 + sqrshrun2 v4.16b, v17.8h, #7 + sqrshrun v5.8b, v18.8h, #7 + sqrshrun2 v5.16b, v19.8h, #7 + st1 {v4.16b, v5.16b}, [x0], x1 + subs w5, w5, #1 + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h48_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 + mov x7, #24 + mov x10, #(MAX_PB_SIZE * 2 - 48) +1: ld1 {v24.16b, v25.16b, v26.16b}, [x2] + ldr s27, [x2, #48] + add x2, x2, x3 + ext v28.16b, v24.16b, v25.16b, #1 + ext v29.16b, v24.16b, v25.16b, #2 + ext v30.16b, v24.16b, v25.16b, #3 + calc_epelb v16, v24, v28, v29, v30 + calc_epelb2 v17, v24, v28, v29, v30 + ext v28.16b, v25.16b, v26.16b, #1 + ext v29.16b, v25.16b, v26.16b, #2 + ext v30.16b, v25.16b, v26.16b, #3 + calc_epelb v18, v25, v28, v29, v30 + calc_epelb2 v19, v25, v28, v29, v30 + ext v28.16b, v26.16b, v27.16b, #1 + ext v29.16b, v26.16b, v27.16b, #2 + ext v30.16b, v26.16b, v27.16b, #3 + calc_epelb v20, v26, v28, v29, v30 + calc_epelb2 v21, v26, v28, v29, v30 + + ld1 {v24.8h, v25.8h, v26.8h}, [x4], #48 + sqadd v16.8h, v16.8h, v24.8h + sqadd v17.8h, v17.8h, v25.8h + sqadd v18.8h, v18.8h, v26.8h + ld1 {v27.8h, v28.8h, v29.8h}, [x4], x10 + sqadd v19.8h, v19.8h, v27.8h + sqadd v20.8h, v20.8h, v28.8h + sqadd v21.8h, v21.8h, v29.8h + sqrshrun v4.8b, v16.8h, #7 + sqrshrun2 v4.16b, v17.8h, #7 + sqrshrun v5.8b, v18.8h, #7 + sqrshrun2 v5.16b, v19.8h, #7 + sqrshrun v6.8b, v20.8h, #7 + sqrshrun2 v6.16b, v21.8h, #7 + st1 {v4.16b, v5.16b, v6.16b}, [x0], x1 + subs w5, w5, #1 // height + b.ne 1b + ret +endfunc + +function ff_hevc_put_hevc_epel_bi_h64_8_neon, export=1 + load_epel_filterb x6, x7 + sub x2, x2, #1 +1: ld1 {v24.16b, v25.16b, v26.16b, v27.16b}, [x2] + ldr s28, [x2, #64] + add x2, x2, x3 + ext v29.16b, v24.16b, v25.16b, #1 + ext v30.16b, v24.16b, v25.16b, #2 + ext v31.16b, v24.16b, v25.16b, #3 + calc_epelb v16, v24, v29, v30, v31 + calc_epelb2 v17, v24, v29, v30, v31 + ext v29.16b, v25.16b, v26.16b, #1 + ext v30.16b, v25.16b, v26.16b, #2 + ext v31.16b, v25.16b, v26.16b, #3 + calc_epelb v18, v25, v29, v30, v31 + calc_epelb2 v19, v25, v29, v30, v31 + ld1 {v4.8h, v5.8h, v6.8h, v7.8h}, [x4], #64 + sqadd v16.8h, v16.8h, v4.8h + sqadd v17.8h, v17.8h, v5.8h + sqadd v18.8h, v18.8h, v6.8h + sqadd v19.8h, v19.8h, v7.8h + sqrshrun v16.8b, v16.8h, #7 + sqrshrun2 v16.16b, v17.8h, #7 + sqrshrun v17.8b, v18.8h, #7 + sqrshrun2 v17.16b, v19.8h, #7 + + ext v29.16b, v26.16b, v27.16b, #1 + ext v30.16b, v26.16b, v27.16b, #2 + ext v31.16b, v26.16b, v27.16b, #3 + calc_epelb v20, v26, v29, v30, v31 + calc_epelb2 v21, v26, v29, v30, v31 + ext v29.16b, v27.16b, v28.16b, #1 + ext v30.16b, v27.16b, v28.16b, #2 + ext v31.16b, v27.16b, v28.16b, #3 + calc_epelb v22, v27, v29, v30, v31 + calc_epelb2 v23, v27, v29, v30, v31 + ld1 {v4.8h, v5.8h, v6.8h, v7.8h}, [x4], #64 + sqadd v20.8h, v20.8h, v4.8h + sqadd v21.8h, v21.8h, v5.8h + sqadd v22.8h, v22.8h, v6.8h + sqadd v23.8h, v23.8h, v7.8h + sqrshrun v18.8b, v20.8h, #7 + sqrshrun2 v18.16b, v21.8h, #7 + sqrshrun v19.8b, v22.8h, #7 + sqrshrun2 v19.16b, v23.8h, #7 + st1 {v16.16b, v17.16b, v18.16b, v19.16b}, [x0], x1 + subs w5, w5, #1 + b.ne 1b + ret +endfunc + function ff_hevc_put_hevc_epel_v4_8_neon, export=1 load_epel_filterb x5, x4 sub x1, x1, x2 diff --git a/libavcodec/aarch64/hevcdsp_init_aarch64.c b/libavcodec/aarch64/hevcdsp_init_aarch64.c index cf171023e7..a16c1a83e0 100644 --- a/libavcodec/aarch64/hevcdsp_init_aarch64.c +++ b/libavcodec/aarch64/hevcdsp_init_aarch64.c @@ -160,6 +160,10 @@ NEON8_FNPROTO(pel_bi_pixels, (uint8_t *dst, ptrdiff_t dststride, const uint8_t *_src, ptrdiff_t _srcstride, const int16_t *src2, int height, intptr_t mx, intptr_t my, int width),); +NEON8_FNPROTO(epel_bi_h, (uint8_t *dst, ptrdiff_t dststride, + const uint8_t *src, ptrdiff_t srcstride, const int16_t *src2, + int height, intptr_t mx, intptr_t my, int width),); + NEON8_FNPROTO(epel_v, (uint8_t *dst, ptrdiff_t dststride, const uint8_t *_src, ptrdiff_t _srcstride, const int16_t *src2, int height, intptr_t mx, intptr_t my, int width),); @@ -329,6 +333,7 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth) NEON8_FNASSIGN(c->put_hevc_qpel, 0, 0, pel_pixels,); NEON8_FNASSIGN(c->put_hevc_qpel, 1, 0, qpel_v,); NEON8_FNASSIGN(c->put_hevc_epel_bi, 0, 0, pel_bi_pixels,); + NEON8_FNASSIGN(c->put_hevc_epel_bi, 0, 1, epel_bi_h,); NEON8_FNASSIGN(c->put_hevc_qpel_bi, 0, 0, pel_bi_pixels,); NEON8_FNASSIGN(c->put_hevc_epel_uni, 0, 0, pel_uni_pixels,);