From patchwork Sun Jun 30 09:05:27 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: =?utf-8?q?R=C3=A9mi_Denis-Courmont?= X-Patchwork-Id: 50218 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a59:cc64:0:b0:482:c625:d099 with SMTP id k4csp1176479vqv; Sun, 30 Jun 2024 02:05:45 -0700 (PDT) X-Forwarded-Encrypted: i=2; AJvYcCWNUmaU5k98QVWBT4hEfVR7y8BwIphwWLb49bvEO39KErSbVVNLGEwRgoGmteTOqBcBsorhLTXiB+VPgrv1Wnpp0uApGBfNXxNGyQ== X-Google-Smtp-Source: AGHT+IF6r9cFUzZfiHewOsNFsF18QZ6lvJSmIb9Vz7grdyvzS5SCk/ivo/UWHYbORU+e2F/ZbauB X-Received: by 2002:a17:906:d930:b0:a6f:5723:fb11 with SMTP id a640c23a62f3a-a7514495948mr152825066b.58.1719738345245; Sun, 30 Jun 2024 02:05:45 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1719738345; cv=none; d=google.com; s=arc-20160816; b=pqaZGEaOye3sZSAts4Zjp6gq7kIwiv3NIKczm98xBYnZhSiQJ8Yn8xGq9oeSJk3IgP 3ut/zGIszDMxfPBuPs5fJY/lZiHambEv1nsfE0ndeg4b/2z20RXNcidgV3rVg9rkaiQ0 OKlWaOMcDO8ft0JMtvrwU4BYB4iGxi+DcnAEqvo/WICMj4HuKg+UmhndxGt6e0V4Dr30 ZbVPl5o30McKHbNO2jRAJV3UkF2+rMULYFMWqAQdhu0Ik3Li4hnE6WsqvpxbwRkJD7XI SUz+2FVJCz34UkQI7xAOvcGiTx0z7IaZbQa/i1tryVV8OrA8/JAQ5l8kSbAhacsMmITj 20xQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding: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:delivered-to; bh=lfz4eNKiiNjvjKAQf+Otl53f7wPjuKF507Cl5/PXtX8=; fh=YOA8vD9MJZuwZ71F/05pj6KdCjf6jQRmzLS+CATXUQk=; b=0BPEPNsRYj0slE829LuE3kJT7LqzMnYVK/5yJ/Fk2a9z49gb95TWUN8fUqcuzTFTsz E3JN8bphMKrkinPp+fpiOjQrhQwBp8DljYAzJmydXz58gPU+VlC6+DF1kXFFebofskP2 tdxdc/bc/l9Vp/SI/UkOhUhR16VmIp7hgyCGdcglh2Hxst2p4OVyTbQRLkzf6b0EulIC OtX/eCQN81ZMofMyc3MNYsjUi7AqezAMjx1UVAjF6xmhdSkKC9728Apg3Vl/dgJ75lNL MYW7eE+WbmeGMAGf5vtDLMtNxbnSkHW3qjr0kMOpZonAvcAkfNsfgaSpQLIEX9jJ/oV+ mcEQ==; dara=google.com 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 a640c23a62f3a-a72ab04a157si247337566b.288.2024.06.30.02.05.44; Sun, 30 Jun 2024 02:05:45 -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; 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 BE13D68D75E; Sun, 30 Jun 2024 12:05:36 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from ursule.remlab.net (vps-a2bccee9.vps.ovh.net [51.75.19.47]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 9F7A868D688 for ; Sun, 30 Jun 2024 12:05:28 +0300 (EEST) Received: from basile.remlab.net (localhost [IPv6:::1]) by ursule.remlab.net (Postfix) with ESMTP id 3733FC0145 for ; Sun, 30 Jun 2024 12:05:28 +0300 (EEST) From: =?utf-8?q?R=C3=A9mi_Denis-Courmont?= To: ffmpeg-devel@ffmpeg.org Date: Sun, 30 Jun 2024 12:05:27 +0300 Message-ID: <20240630090527.65998-2-remi@remlab.net> X-Mailer: git-send-email 2.45.2 In-Reply-To: <20240630090527.65998-1-remi@remlab.net> References: <20240630090527.65998-1-remi@remlab.net> MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 2/2] lavc/h264dsp: R-V V 8-bit MBAFF loop filter 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 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: gYWMzB9doKK+ Performance is (unfortunately) the same as with non-MBAFF, since the hardware under test does not short-circuit vector tail calculations. (IMO, a generic solution or work-around should be agreed on, rather than bespoke approaches all over the place.) --- libavcodec/riscv/h264dsp_init.c | 4 ++++ libavcodec/riscv/h264dsp_rvv.S | 19 ++++++++++++++++--- 2 files changed, 20 insertions(+), 3 deletions(-) diff --git a/libavcodec/riscv/h264dsp_init.c b/libavcodec/riscv/h264dsp_init.c index 0d4d541992..ab412a9924 100644 --- a/libavcodec/riscv/h264dsp_init.c +++ b/libavcodec/riscv/h264dsp_init.c @@ -31,6 +31,8 @@ void ff_h264_v_loop_filter_luma_8_rvv(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0); void ff_h264_h_loop_filter_luma_8_rvv(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0); +void ff_h264_h_loop_filter_luma_mbaff_8_rvv(uint8_t *pix, ptrdiff_t stride, + int alpha, int beta, int8_t *tc0); extern int ff_startcode_find_candidate_rvb(const uint8_t *, int); extern int ff_startcode_find_candidate_rvv(const uint8_t *, int); @@ -48,6 +50,8 @@ av_cold void ff_h264dsp_init_riscv(H264DSPContext *dsp, const int bit_depth, if (bit_depth == 8 && ff_rv_vlen_least(128)) { dsp->h264_v_loop_filter_luma = ff_h264_v_loop_filter_luma_8_rvv; dsp->h264_h_loop_filter_luma = ff_h264_h_loop_filter_luma_8_rvv; + dsp->h264_h_loop_filter_luma_mbaff = + ff_h264_h_loop_filter_luma_mbaff_8_rvv; } dsp->startcode_find_candidate = ff_startcode_find_candidate_rvv; } diff --git a/libavcodec/riscv/h264dsp_rvv.S b/libavcodec/riscv/h264dsp_rvv.S index ea9dfb1a7e..c5d21ba607 100644 --- a/libavcodec/riscv/h264dsp_rvv.S +++ b/libavcodec/riscv/h264dsp_rvv.S @@ -26,6 +26,15 @@ #include "libavutil/riscv/asm.S" + .macro .mbaff + .irp type,,_mbaff + .ifb \type + .equ IS_MBAFF, 0 + .else + .equ IS_MBAFF, 1 + .endif + .endm + .variant_cc ff_h264_loop_filter_luma_8_rvv func ff_h264_loop_filter_luma_8_rvv, zve32x # p2: v8, p1: v9, p0: v10, q0: v11, q1: v12, q2: v13 @@ -33,7 +42,7 @@ func ff_h264_loop_filter_luma_8_rvv, zve32x csrwi vxrm, 0 vid.v v0 vaaddu.vv v14, v10, v11 # (p0 + q0 + 1) / 2 - vsrl.vi v0, v0, 2 # v0[i] = i / inner_iters + vsrl.vx v0, v0, a4 # v0[i] = i / inner_iters vwsubu.vv v16, v9, v12 vrgather.vv v6, v4, v0 # tc_orig vwaddu.vv v18, v8, v14 @@ -116,6 +125,7 @@ func ff_h264_v_loop_filter_luma_8_rvv, zve32x add t6, t5, a1 vle8.v v12, (t5) vle8.v v13, (t6) + li a4, 2 # log2(inner_iters) jal t0, ff_h264_loop_filter_luma_8_rvv vse8.v v9, (t2) vse8.v v10, (t3) @@ -124,13 +134,16 @@ func ff_h264_v_loop_filter_luma_8_rvv, zve32x ret endfunc -func ff_h264_h_loop_filter_luma_8_rvv, zve32x +.mbaff +func ff_h264_h_loop_filter_luma\type\()_8_rvv, zve32x vsetivli zero, 4, e8, mf4, ta, ma vle8.v v4, (a4) addi a0, a0, -3 - vsetivli zero, 16, e8, m1, ta, ma + vsetivli zero, 16 >> IS_MBAFF, e8, m1, ta, ma vlsseg6e8.v v8, (a0), a1 + li a4, 2 >> IS_MBAFF # log2(inner_iters) jal t0, ff_h264_loop_filter_luma_8_rvv vssseg6e8.v v8, (a0), a1 ret endfunc +.endr