From patchwork Thu Dec 1 16:57:44 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: James Darnley X-Patchwork-Id: 1641 Delivered-To: ffmpegpatchwork@gmail.com Received: by 10.103.65.86 with SMTP id o83csp1042624vsa; Thu, 1 Dec 2016 09:00:01 -0800 (PST) X-Received: by 10.28.161.67 with SMTP id k64mr35713121wme.69.1480611601818; Thu, 01 Dec 2016 09:00:01 -0800 (PST) Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id m1si960673wjt.59.2016.12.01.09.00.01; Thu, 01 Dec 2016 09:00:01 -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=@ob-encoder-com.20150623.gappssmtp.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 69DE8689EFC; Thu, 1 Dec 2016 18:59:50 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-wj0-f195.google.com (mail-wj0-f195.google.com [209.85.210.195]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id C495B689E39 for ; Thu, 1 Dec 2016 18:59:43 +0200 (EET) Received: by mail-wj0-f195.google.com with SMTP id kp2so27132769wjc.0 for ; Thu, 01 Dec 2016 08:59:51 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=ob-encoder-com.20150623.gappssmtp.com; s=20150623; h=sender:from:to:subject:date:message-id:in-reply-to:references; bh=9AGRKZY0soG6LcvU9j99iUux0dg2GbG2WOZhkru25sA=; b=0AcVcmdnxiyGd19lPxCtnWGRYTpciqJirhRdJM2a1rp1sRn8TXV3Z9OIqvjXgBAtol QXvMCwQ6zNlqbNO5ADoA+yujaaOPaSX6TKrYA9I1NCwqMq8x0FRYc8tT+HJsKiOB0O6r VVFSFV1dV6AfeJQgepK3Du8v7i0JdJRIQ18IDLkhLTMqQe4LR9VYDmyBVOwfL8VXB2D5 RXO1Etl+MZ19g5d8bPrvkE0JVlWehLQoawiSxZgViYhgXx61GCTG6xzJsmSJafLsVU6c sQqP4fanuZp5j7KDae4ENVT56HTL5aRQlxKnxlJ/fiqaYrKgvaW6AMjgSTvcdi+a9Ma7 AwYA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:sender:from:to:subject:date:message-id :in-reply-to:references; bh=9AGRKZY0soG6LcvU9j99iUux0dg2GbG2WOZhkru25sA=; b=aOeZDvtZJuHBQVB0MyTRqFSXC+49dv8ybr6GpH62kN45IFro26b2jXCKs+JvW7nHsH +TV49dCUnkAAKWPmW+DCLvIdd+npYhQynAvoZRBWAbh+cx0axGK/F0LmXTK7i1j1nJpB YFqM/vHjh0HAXEo66+CC7r8vVE1Aq8Lfn+FLJYZlZrH2hcg/c8HDHRBVMRuz+tS1h84F K4dNfJa8iBViaUbEfcAyb+DQL+siQz5h5iuI0zTyt49SDTInWXEr0XzUODLG8NGKboVO h7mNlNUpPfFyiMbqXPe9v6X1cw2sfqWAUSap9rpBjDovTWWlojGLQhEwC02/e3UkBHtl j3vA== X-Gm-Message-State: AKaTC03wZM/iJKNQWGX/1CtpRBBywY3qH6YgES9eIcGb30BsylFguov3nYkH69SDeXxJbw== X-Received: by 10.194.72.100 with SMTP id c4mr33623150wjv.188.1480611590624; Thu, 01 Dec 2016 08:59:50 -0800 (PST) Received: from Ifrit.systemlords.lan (d51a44418.access.telenet.be. [81.164.68.24]) by smtp.gmail.com with ESMTPSA id kp5sm1206172wjb.8.2016.12.01.08.59.50 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Thu, 01 Dec 2016 08:59:50 -0800 (PST) From: James Darnley To: FFmpeg development discussions and patches Date: Thu, 1 Dec 2016 17:57:44 +0100 Message-Id: <20161201165749.10368-2-jdarnley@obe.tv> X-Mailer: git-send-email 2.10.2 In-Reply-To: <20161201165749.10368-1-jdarnley@obe.tv> References: <20161201165749.10368-1-jdarnley@obe.tv> Subject: [FFmpeg-devel] [PATCH 1/6] avcodec/h264: mmx2, sse2, avx 10-bit h chroma deblock/loop filter 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 MIME-Version: 1.0 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" Yorkfield: - mmx2: 2.44x faster (278 vs. 114 cycles) - sse2: 3.35x faster (278 vs. 83 cycles) Skylake: - mmx2: 1.69x faster (169 vs. 100 cycles) - sse2: 2.34x faster (169 vs. 72 cycles) - avx: 2.32x faster (169 vs. 73 cycles) --- libavcodec/x86/h264_deblock_10bit.asm | 118 ++++++++++++++++++++++++++++++++++ libavcodec/x86/h264dsp_init.c | 9 +++ 2 files changed, 127 insertions(+) diff --git a/libavcodec/x86/h264_deblock_10bit.asm b/libavcodec/x86/h264_deblock_10bit.asm index ebf8a3f..e91b1c6 100644 --- a/libavcodec/x86/h264_deblock_10bit.asm +++ b/libavcodec/x86/h264_deblock_10bit.asm @@ -843,6 +843,88 @@ DEBLOCK_LUMA_INTRA mova [r0+2*r1], m2 %endmacro +; expands to [base],...,[base+7*stride] +%define PASS8ROWS(base, base3, stride, stride3) \ + [base], [base+stride], [base+stride*2], [base3], \ + [base3+stride], [base3+stride*2], [base3+stride3], [base3+stride*4] + +; in: 8 rows of 4 words in %4..%11 +; out: 4 rows of 8 words in m0..m3 +%macro TRANSPOSE4x8W_LOAD 8 + movq m0, %1 + movq m2, %2 + movq m1, %3 + movq m3, %4 + + punpcklwd m0, m2 + punpcklwd m1, m3 + punpckhdq m2, m0, m1 + punpckldq m0, m1 + + movq m4, %5 + movq m6, %6 + movq m5, %7 + movq m3, %8 + + punpcklwd m4, m6 + punpcklwd m5, m3 + punpckhdq m6, m4, m5 + punpckldq m4, m5 + + punpckhqdq m1, m0, m4 + punpcklqdq m0, m4 + punpckhqdq m3, m2, m6 + punpcklqdq m2, m6 +%endmacro + +; in: 4 rows of 8 words in m0..m3 +; out: 8 rows of 4 words in %1..%8 +%macro TRANSPOSE8x4W_STORE 8 + TRANSPOSE4x4W 0, 1, 2, 3, 4 + movq %1, m0 + movhps %2, m0 + movq %3, m1 + movhps %4, m1 + movq %5, m2 + movhps %6, m2 + movq %7, m3 + movhps %8, m3 +%endmacro + +; %1 = base + 3*stride +; %2 = 3*stride (unused on mmx) +; %3, %4 = place to store p1 and q1 values +%macro CHROMA_H_LOAD 4 + %if mmsize == 8 + movq m0, [pix_q - 4] + movq m1, [pix_q + stride_q - 4] + movq m2, [pix_q + 2*stride_q - 4] + movq m3, [%1 - 4] + TRANSPOSE4x4W 0, 1, 2, 3, 4 + %else + TRANSPOSE4x8W_LOAD PASS8ROWS(pix_q-4, %1-4, stride_q, %2) + %endif + mova %3, m0 + mova %4, m3 +%endmacro + +; %1 = base + 3*stride +; %2 = 3*stride (unused on mmx) +; %3, %4 = place to load p1 and q1 values +%macro CHROMA_H_STORE 4 + mova m0, %3 + mova m3, %4 + %if mmsize == 8 + TRANSPOSE4x4W 0, 1, 2, 3, 4 + movq [pix_q - 4], m0 + movq [pix_q + stride_q - 4], m1 + movq [pix_q + 2*stride_q - 4], m2 + movq [%1 - 4], m3 + %else + TRANSPOSE8x4W_STORE PASS8ROWS(pix_q-4, %1-4, stride_q, %2) + %endif +%endmacro + %macro CHROMA_V_LOAD_TC 2 movd %1, [%2] punpcklbw %1, %1 @@ -914,6 +996,42 @@ cglobal deblock_v_chroma_intra_10, 4,6-(mmsize/16),8*(mmsize/16) %else RET %endif + +;----------------------------------------------------------------------------- +; void ff_deblock_h_chroma_10(uint16_t *pix, int stride, int alpha, int beta, +; int8_t *tc0) +;----------------------------------------------------------------------------- +cglobal deblock_h_chroma_10, 5, 7, 8, 2*mmsize, pix_, stride_, alpha_, beta_, tc0_ + shl alpha_d, 2 + shl beta_d, 2 + mov r5, pix_q + lea r6, [3*stride_q] + add r5, r6 +%if mmsize == 8 + mov r6b, 2 + .loop: +%endif + + CHROMA_H_LOAD r5, r6, [rsp], [rsp + mmsize] + LOAD_AB m4, m5, alpha_d, beta_d + LOAD_MASK m0, m1, m2, m3, m4, m5, m7, m6, m4 + pxor m4, m4 + CHROMA_V_LOAD_TC m6, tc0_q + psubw m6, [pw_3] + pmaxsw m6, m4 + pand m7, m6 + DEBLOCK_P0_Q0 m1, m2, m0, m3, m7, m5, m6 + CHROMA_H_STORE r5, r6, [rsp], [rsp + mmsize] + +%if mmsize == 8 + lea pix_q, [pix_q + 4*stride_q] + lea r5, [r5 + 4*stride_q] + add tc0_q, 2 + dec r6b + jg .loop +%endif +RET + %endmacro %if ARCH_X86_64 == 0 diff --git a/libavcodec/x86/h264dsp_init.c b/libavcodec/x86/h264dsp_init.c index c6c643a..c568762 100644 --- a/libavcodec/x86/h264dsp_init.c +++ b/libavcodec/x86/h264dsp_init.c @@ -310,6 +310,9 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, #if ARCH_X86_32 c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_mmxext; c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmxext; + if (chroma_format_idc <= 1) { + c->h264_h_loop_filter_chroma = ff_deblock_h_chroma_10_mmxext; + } c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_mmxext; c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_mmxext; c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_10_mmxext; @@ -343,6 +346,9 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_sse2; c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_sse2; + if (chroma_format_idc <= 1) { + c->h264_h_loop_filter_chroma = ff_deblock_h_chroma_10_sse2; + } #if HAVE_ALIGNED_STACK c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_sse2; c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_sse2; @@ -378,6 +384,9 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_avx; c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_avx; + if (chroma_format_idc <= 1) { + c->h264_h_loop_filter_chroma = ff_deblock_h_chroma_10_avx; + } #if HAVE_ALIGNED_STACK c->h264_v_loop_filter_luma = ff_deblock_v_luma_10_avx; c->h264_h_loop_filter_luma = ff_deblock_h_luma_10_avx;