From patchwork Wed Nov 22 11:12:04 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Shengbin Meng X-Patchwork-Id: 6263 Delivered-To: ffmpegpatchwork@gmail.com Received: by 10.2.161.94 with SMTP id m30csp31972jah; Wed, 22 Nov 2017 03:18:39 -0800 (PST) X-Google-Smtp-Source: AGs4zMYJehuyjPMX7q7D/u39m2euP2Mu1fF/1UKiSNGu+/KwmQijFV87SIvgrLo031I16HtIwSfq X-Received: by 10.28.58.211 with SMTP id h202mr3581470wma.56.1511349519241; Wed, 22 Nov 2017 03:18:39 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1511349519; cv=none; d=google.com; s=arc-20160816; b=irflJG5rDdlZy40WrDvm95cg1BjDlQ/GFh/iPzKaEuBBh0PUYILm428R63fH22nNUL vhx8U8Nkor7WVriI5IlHo2/HlLv2zRHD6AQnYt7+rPLWLop5PyAVJnkELelQI1rEz3xm HM50Fw/0B528dX48ZRvnuJNkY5pjr9JpZQnQ+1mDJgOwkcwy9DUV6m4rooxIdAlSO6tk fcd236Xuye5cIfqqJ6YycNHIKsXCzPDCpnOvDWfb7K+2joMvZEDXnI6hQhtyUQlI4Z5S NlYHT12zOr96O/bONCaoq5L5YPTFQXA6lXHqFfWu60F28hnPH9JciwE+uNbQr3fL90Vy 0GKg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:mime-version:cc:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:references:in-reply-to:message-id:date :to:from:dkim-signature:delivered-to:arc-authentication-results; bh=hsPLf5S+kOfFII5z2Df2azWnJ5STqGd+8s5h/1j3W+w=; b=vIXJXBi9wgqIxOyqI3LEbLviCkP9GFDLwhgy0wwKKa9vr787fjHZuoXiRW4YnAhzZL VQDzomx2vCEPNeNOK6q6Tz1SIQ5V/xUcc/mrWLFSq+Ub1vnIF7boj42h/j862nOtA2QS C/AIwJrdRlkqiGi5volWAebt8ZQLfdRorjz4pCsQeHEewfzTwM2r+mBkIFVoxvDwsIp9 n5vVurT0ckt8drA9yzcOTMYJxWtx2FqcXfRoHc0tDXaWTiRbdt+Mh7qHRQPhGk910NfT hY/99uuh3onqxDiyvEvlSZul7tYzgJ+nHyg45ePElYPIcqjBN15cMjz5j3jhlarf4GyG /JFg== ARC-Authentication-Results: i=1; mx.google.com; dkim=neutral (body hash did not verify) header.i=@gmail.com header.s=20161025 header.b=ct30gpAh; 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=NONE sp=NONE dis=NONE) header.from=gmail.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id u129si3141479wmd.183.2017.11.22.03.18.38; Wed, 22 Nov 2017 03:18:39 -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; dkim=neutral (body hash did not verify) header.i=@gmail.com header.s=20161025 header.b=ct30gpAh; 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=NONE sp=NONE dis=NONE) header.from=gmail.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 69D3668A119; Wed, 22 Nov 2017 13:18:37 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-pl0-f65.google.com (mail-pl0-f65.google.com [209.85.160.65]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 998EC68A131 for ; Wed, 22 Nov 2017 13:18:30 +0200 (EET) Received: by mail-pl0-f65.google.com with SMTP id 62so552546plc.2 for ; Wed, 22 Nov 2017 03:18:30 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=SEeWa/Iuxr/H3/jeBxmsdFnBzQR5n/PQGE7P+xbyf6g=; b=ct30gpAhzfQMSizUWwqetAihHiIjyNkljSY9KftU31w1N45pairWXG3WtQzoa/SKop 1SDFf0dN8Bezfi515ItHS37F3v7F2MOsl8j1S/ePHIGO8952IxLjNIckRbdKDhh2wN7N 8tqhd5t+YbF5aKiy8D7nN0qWN6lKZ8TJzUGl06aMszqHMEyW/V8QCfmNdQvtH7md3tKB 2fhp8dpz6jKFa559Bv5z1YpDpz8p5jocDesOlMn2s04L+stgW5+YNbNqzfsXULgLAFRc NcYmpvDiHdrRpthg22LvrLGtrwnyRiq8KThp7zuhmoGhqHN66JmnNej86jx2tglviK3g +E0w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=SEeWa/Iuxr/H3/jeBxmsdFnBzQR5n/PQGE7P+xbyf6g=; b=qTqgy3UFXiCc8B1n3pPXIHRZtr+N2P81A8kBKd1ltDiSps8JcIg09o2atcake/IEpv bqM2CVfwgH7OYadSk70eXxQtAZam9f1c8DEgEaX9RhPl6ItwTO/qM14pzc+lMHtAX5Rl tHddMdl1rWmcOdRM8tuJCtHY0zYQ0DnjURdFi7rnH5i9DU/FuuBhPNGTtuxnPCrEJIie M66qYAPF3NGzGZrkrRYfr1Ab6vXTftgrhA/QOXJPXtJYJ4YR+TuAulnivWnaZ0SdFBr9 S4ug3IZDPZpxVIxrL2TmriAgVdcXmzHbfIGWuuPdk5j5HZ7tZO/3tGlzlyvD6x4HzUr3 e1rg== X-Gm-Message-State: AJaThX7p0ZqDZnBFWyUvkKf5DCTD7JE0W3aU2ggLxcwOAVnftsyEn8x5 4U4y+3R6S62dnQ9D6bjG4b11+A== X-Received: by 10.84.211.105 with SMTP id b96mr20269713pli.86.1511349140109; Wed, 22 Nov 2017 03:12:20 -0800 (PST) Received: from Shengbins-Mac-mini.corp.bytedance.com ([103.48.142.82]) by smtp.googlemail.com with ESMTPSA id 69sm29516227pft.11.2017.11.22.03.12.18 (version=TLS1_2 cipher=ECDHE-RSA-AES128-SHA bits=128/128); Wed, 22 Nov 2017 03:12:19 -0800 (PST) From: Shengbin Meng To: ffmpeg-devel@ffmpeg.org Date: Wed, 22 Nov 2017 19:12:04 +0800 Message-Id: <20171122111206.17214-5-shengbinmeng@gmail.com> X-Mailer: git-send-email 2.13.6 (Apple Git-96) In-Reply-To: <20171122111206.17214-1-shengbinmeng@gmail.com> References: <20171122111206.17214-1-shengbinmeng@gmail.com> Subject: [FFmpeg-devel] [PATCH 4/6] avcodec/hevcdsp: Use pre-load (pld) to optimize data loading X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.20 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: Meng Wang MIME-Version: 1.0 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" From: Meng Wang Signed-off-by: Meng Wang --- libavcodec/arm/hevcdsp_epel_neon.S | 10 ++++++++++ libavcodec/arm/hevcdsp_qpel_neon.S | 24 ++++++++++++++++++++---- 2 files changed, 30 insertions(+), 4 deletions(-) diff --git a/libavcodec/arm/hevcdsp_epel_neon.S b/libavcodec/arm/hevcdsp_epel_neon.S index d0d93e8033..03e6504481 100644 --- a/libavcodec/arm/hevcdsp_epel_neon.S +++ b/libavcodec/arm/hevcdsp_epel_neon.S @@ -306,6 +306,7 @@ cmp r5, #2 beq 2f 8: subs r4, #1 + pld [r2] \filter vst1.16 {q7}, [r0], r1 regshuffle_d4 @@ -320,6 +321,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vst1.16 d14, [r0], r1 regshuffle_d4 @@ -357,6 +359,7 @@ cmp r5, #2 beq 2f 8: subs r4, #1 + pld [r2] \filter vqrshrun.s16 d0, q7, #6 vst1.8 d0, [r0], r1 @@ -372,6 +375,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vqrshrun.s16 d0, q7, #6 vst1.32 d0[0], [r0], r1 @@ -396,6 +400,7 @@ cmp r5, #2 beq 2f 8: subs r4, #1 + pld [r2] \filter vld1.16 {q0}, [r8], r9 vqadd.s16 q0, q7 @@ -415,6 +420,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vld1.16 d0, [r8], r9 vqadd.s16 d0, d14 @@ -465,6 +471,7 @@ cmp r5, #2 beq 2f 8: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmovl.s16 q13, d15 @@ -490,6 +497,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmul.s32 q14, q12, q6 @@ -535,6 +543,7 @@ cmp r5, #2 beq 2f 8: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmovl.s16 q13, d15 @@ -569,6 +578,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 vmul.s32 q14, q12, q6 diff --git a/libavcodec/arm/hevcdsp_qpel_neon.S b/libavcodec/arm/hevcdsp_qpel_neon.S index 71ecc00b6e..b507fbc13b 100644 --- a/libavcodec/arm/hevcdsp_qpel_neon.S +++ b/libavcodec/arm/hevcdsp_qpel_neon.S @@ -231,6 +231,7 @@ cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] \filter vst1.16 {q7}, [r0], r1 regshuffle_d8 @@ -245,6 +246,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vst1.16 d14, [r0], r1 regshuffle_d8 @@ -273,6 +275,7 @@ cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] \filter vqrshrun.s16 d0, q7, #6 vst1.8 d0, [r0], r1 @@ -288,6 +291,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vqrshrun.s16 d0, q7, #6 vst1.32 d0[0], [r0], r1 @@ -301,6 +305,7 @@ cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] \filter vld1.16 {q0}, [r8], r9 vqadd.s16 q0, q7 @@ -320,6 +325,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vld1.16 d0, [r8], r9 vqadd.s16 d0, d14 @@ -358,6 +364,7 @@ cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmovl.s16 q13, d15 @@ -383,6 +390,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmul.s32 q14, q12, q6 @@ -412,6 +420,7 @@ cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmovl.s16 q13, d15 @@ -446,6 +455,7 @@ mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] \filter vmovl.s16 q12, d14 vmul.s32 q14, q12, q6 @@ -1524,8 +1534,9 @@ function ff_hevc_put_qpel_bi_uw_pixels_neon_8, export=1 cmp r5, #4 beq 4f 8: subs r4, #1 - vshll.u8 q7 , d8, #6 // src[x] << 6 and move long to 8x16bit + pld [r2] vld1.16 {q0}, [r8], r9 // load 8x16bit src2 + vshll.u8 q7 , d8, #6 // src[x] << 6 and move long to 8x16bit vqadd.s16 q0, q7 // ((src << 6) + src2) on 8x16bit operation vqrshrun.s16 d0, q0, #7 // (((src << 6) + src2) + offset) >> 7 narrow to 8x8bit vst1.8 d0, [r0], r1 @@ -1535,15 +1546,16 @@ function ff_hevc_put_qpel_bi_uw_pixels_neon_8, export=1 beq 99f mov r4, r12 add r6, #8 - mov r0, r6 add r10, #16 - mov r8, r10 add r7, #8 + mov r0, r6 + mov r8, r10 mov r2, r7 b 0b 4: subs r4, #1 - vshll.u8 q7 , d8, #6 // src[x] << 6 and move long to 8x16bit + pld [r2] vld1.16 d0, [r8], r9 + vshll.u8 q7 , d8, #6 // src[x] << 6 and move long to 8x16bit vqadd.s16 d0, d14 vqrshrun.s16 d0, q0, #7 vst1.32 d0[0], [r0], r1 @@ -1578,6 +1590,7 @@ function ff_hevc_put_qpel_wt_pixels_neon_8, export=1 cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] vshll.u8 q7 , d16, #6 // src[x] << 6 and move long to 8x16bit vmovl.u16 q12, d14 // extending unsigned 4x16bit data to 4x32 bit vmovl.u16 q13, d15 @@ -1602,6 +1615,7 @@ function ff_hevc_put_qpel_wt_pixels_neon_8, export=1 mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] vshll.u8 q7 , d16, #6 // src[x] << 6 and move long to 8x16bit vmovl.u16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmul.u32 q14, q12, q6 @@ -1632,6 +1646,7 @@ function ff_hevc_put_qpel_wt_pixels_neon_8, export=1 cmp r5, #4 beq 4f 8: subs r4, #1 + pld [r2] vshll.u8 q7, d16, #6 // src[x] << 6 and move long to 8x16bit vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmovl.s16 q13, d15 @@ -1665,6 +1680,7 @@ function ff_hevc_put_qpel_wt_pixels_neon_8, export=1 mov r2, r7 b 0b 4: subs r4, #1 + pld [r2] vshll.u8 q7, d16, #6 // src[x] << 6 and move long to 8x16bit vmovl.s16 q12, d14 // extending signed 4x16bit data to 4x32 bit vmul.s32 q14, q12, q6 // src * w1