From patchwork Fri Jul 5 20:23:49 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: 50374 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a59:cc64:0:b0:482:c625:d099 with SMTP id k4csp4545423vqv; Fri, 5 Jul 2024 13:24:01 -0700 (PDT) X-Forwarded-Encrypted: i=2; AJvYcCUZLuhiySCKhM8ItigpMkbtqi2O5UfE5viiVrfwfzqLpqUU7pe089i8+rC5mrXwI7KtEoA4CgjP9ruZAC4GXjcbKktmhcHDVvjRLQ== X-Google-Smtp-Source: AGHT+IGoeCH+/d5xioCqsTa/rPKLCnYvoy2KdPK+f+uiqoL8qDnNF1HnvRexPTsXkCXkHrFWovbC X-Received: by 2002:ac2:5193:0:b0:52c:d76a:867f with SMTP id 2adb3069b0e04-52ea0554de8mr3997449e87.0.1720211041082; Fri, 05 Jul 2024 13:24:01 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1720211041; cv=none; d=google.com; s=arc-20160816; b=s7opzthEj8R3GVxaz3Al2BDIGmp03M78m1AoWjZxnLGu2LQ5DzmFYgplp6teE4VdDP z0MxOrm2/+Sdb2l2KRtehWWe137/t3Xj8AK12V3rNmLaJbQXIxSy/eiw8onAXfG/97rQ 2fVvLZf3i8bZ6tMuhqPFDI1lDdMB5F7s9QbcP2FZ8dC73XhDOimnXJO7Qnij5h//gTpB qDjVX823XBvn6OUcYys8SdOLziSI3S6qA4Qm7nTbWPoZliKrggAwwh2ticrweTRDMTUf AcvTrPA/D92tTGBD0YeEN6vDCuKmynCqWfCL/ks8zAMTgNMXhquKyeZwRlcMjW7cPIkS no0A== 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=QjwGkju7vXFFzBnl0xStxGfqLi30xwaGcGtX2CzY8Uc=; fh=YOA8vD9MJZuwZ71F/05pj6KdCjf6jQRmzLS+CATXUQk=; b=V7Db8aL97TYZKE9kwgA8oh4DnM5EAak0aMEorPs363KJWEZfs7TMOnECyAXKQbkP4v YWBywhGG75zJMYPbgQdwvceE8hKU+sfWWZVIeMPpVKxhUVscUqwi2WgMRlvgF+KKMoeQ RFnlkQpysZSdEqi1HoSS8WRBHd78cp/4MsQknNXFPcsxsp/5NzcKf3eXJSy1S55Anpfc bnKGlEw8s3Pbh/y2JOkMFYOpA7C54Ex9ZERTWo1qPaQgAIhzOWt+BMIKtIqWmuWO7cui y9Dbkg2RnWVxzsgzHEOG7pnamhVre2PPBrBP+KF2pyvAsq1ffsqN+Dbl0+jGsRCwiDfE OcEA==; 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-a77cb7c205esi116948566b.1000.2024.07.05.13.23.59; Fri, 05 Jul 2024 13:24:01 -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 A807E68DBAE; Fri, 5 Jul 2024 23:23:56 +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 8102E68DAA8 for ; Fri, 5 Jul 2024 23:23:50 +0300 (EEST) Received: from basile.remlab.net (localhost [IPv6:::1]) by ursule.remlab.net (Postfix) with ESMTP id 216CEC006E for ; Fri, 5 Jul 2024 23:23:50 +0300 (EEST) From: =?utf-8?q?R=C3=A9mi_Denis-Courmont?= To: ffmpeg-devel@ffmpeg.org Date: Fri, 5 Jul 2024 23:23:49 +0300 Message-ID: <20240705202349.51307-1-remi@remlab.net> X-Mailer: git-send-email 2.45.2 In-Reply-To: <20240705182816.27464-1-remi@remlab.net> References: <20240705182816.27464-1-remi@remlab.net> MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 2/2] lavc/h264dsp: R-V V 8-bit h264_biweight_pixels 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: NsPLML/CWAvS T-Head C908: h264_biweight2_8_c: 65.2 h264_biweight2_8_rvv_i32: 24.0 h264_biweight4_8_c: 135.2 h264_biweight4_8_rvv_i32: 48.0 h264_biweight8_8_c: 231.5 h264_biweight8_8_rvv_i32: 104.7 h264_biweight16_8_c: 454.0 h264_biweight16_8_rvv_i32: 93.7 SpacemiT X60: h264_biweight2_8_c: 57.7 h264_biweight2_8_rvv_i32: 16.7 h264_biweight4_8_c: 106.0 h264_biweight4_8_rvv_i32: 33.7 h264_biweight8_8_c: 205.7 h264_biweight8_8_rvv_i32: 77.7 h264_biweight16_8_c: 403.5 h264_biweight16_8_rvv_i32: 83.2 --- libavcodec/riscv/h264dsp_init.c | 14 ++++-- libavcodec/riscv/h264dsp_rvv.S | 78 +++++++++++++++++++++++++++++++++ 2 files changed, 88 insertions(+), 4 deletions(-) diff --git a/libavcodec/riscv/h264dsp_init.c b/libavcodec/riscv/h264dsp_init.c index e1b725dcbb..88afec8df0 100644 --- a/libavcodec/riscv/h264dsp_init.c +++ b/libavcodec/riscv/h264dsp_init.c @@ -28,7 +28,10 @@ #include "libavutil/riscv/cpu.h" #include "libavcodec/h264dsp.h" -extern const h264_weight_func ff_h264_weight_funcs_8_rvv[]; +extern const struct { + const h264_weight_func weight; + const h264_biweight_func biweight; +} ff_h264_weight_funcs_8_rvv[]; void ff_h264_v_loop_filter_luma_8_rvv(uint8_t *pix, ptrdiff_t stride, int alpha, int beta, int8_t *tc0); @@ -63,9 +66,12 @@ av_cold void ff_h264dsp_init_riscv(H264DSPContext *dsp, const int bit_depth, # if HAVE_RVV if (flags & AV_CPU_FLAG_RVV_I32) { if (bit_depth == 8 && ff_rv_vlen_least(128)) { - memcpy(dsp->weight_h264_pixels_tab, - ff_h264_weight_funcs_8_rvv, - sizeof (dsp->weight_h264_pixels_tab)); + for (int i = 0; i < 4; i++) { + dsp->weight_h264_pixels_tab[i] = + ff_h264_weight_funcs_8_rvv[i].weight; + dsp->biweight_h264_pixels_tab[i] = + ff_h264_weight_funcs_8_rvv[i].biweight; + } 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; diff --git a/libavcodec/riscv/h264dsp_rvv.S b/libavcodec/riscv/h264dsp_rvv.S index ab85bfbd69..6cbc699b21 100644 --- a/libavcodec/riscv/h264dsp_rvv.S +++ b/libavcodec/riscv/h264dsp_rvv.S @@ -48,6 +48,34 @@ func ff_h264_weight_pixels_simple_8_rvv, zve32x ret endfunc + .variant_cc ff_h264_biweight_pixels_simple_8_rvv +func ff_h264_biweight_pixels_simple_8_rvv, zve32x + csrwi vxrm, 0 + sll a7, a7, a3 + addi a4, a4, 1 +1: + vsetvli zero, t6, e32, m4, ta, ma + vle8.v v8, (a0) + addi a3, a3, -1 + vle8.v v12, (a1) + add a1, a1, a2 + vmv.v.x v16, a7 + vsetvli zero, zero, e16, m2, ta, ma + vzext.vf2 v24, v8 + vzext.vf2 v28, v12 + vwmaccsu.vx v16, a5, v24 + vwmaccsu.vx v16, a6, v28 + vnclip.wi v16, v16, 0 + vmax.vx v16, v16, zero + vsetvli zero, zero, e8, m1, ta, ma + vnclipu.wx v8, v16, a4 + vse8.v v8, (a0) + add a0, a0, a2 + bnez a3, 1b + + ret +endfunc + func ff_h264_weight_pixels_8_rvv, zve32x csrwi vxrm, 0 sll a5, a5, a3 @@ -78,6 +106,44 @@ func ff_h264_weight_pixels_8_rvv, zve32x ret endfunc + .variant_cc ff_h264_biweight_pixels_8_rvv +func ff_h264_biweight_pixels_8_rvv, zve32x + csrwi vxrm, 0 + sll a7, a7, a3 + addi a4, a4, 1 +1: + mv t0, a0 + mv t1, a1 + mv t5, t6 +2: + vsetvli t2, a3, e32, m8, ta, ma + vlse8.v v8, (t0), a2 + vlse8.v v12, (t1), a2 + addi t5, t5, -1 + vmv.v.x v16, a7 + vsetvli zero, zero, e16, m4, ta, ma + vzext.vf2 v24, v8 + vzext.vf2 v28, v12 + vwmaccsu.vx v16, a5, v24 + vwmaccsu.vx v16, a6, v28 + vnclip.wi v16, v16, 0 + vmax.vx v16, v16, zero + vsetvli zero, zero, e8, m2, ta, ma + vnclipu.wx v8, v16, a4 + vsse8.v v8, (t0), a2 + addi t0, t0, 1 + addi t1, t1, 1 + bnez t5, 2b + + mul t3, a2, t2 + sub a3, a3, t2 + add a0, a0, t3 + add a1, a1, t3 + bnez a3, 1b + + ret +endfunc + .irp w, 16, 8, 4, 2 func ff_h264_weight_pixels\w\()_8_rvv, zve32x li a6, \w @@ -87,6 +153,15 @@ func ff_h264_weight_pixels\w\()_8_rvv, zve32x j ff_h264_weight_pixels_8_rvv .endif endfunc + +func ff_h264_biweight_pixels\w\()_8_rvv, zve32x + li t6, \w + .if \w == 16 + j ff_h264_biweight_pixels_simple_8_rvv + .else + j ff_h264_biweight_pixels_8_rvv + .endif +endfunc .endr .global ff_h264_weight_funcs_8_rvv @@ -95,10 +170,13 @@ const ff_h264_weight_funcs_8_rvv .irp w, 16, 8, 4, 2 #if __riscv_xlen == 32 .word ff_h264_weight_pixels\w\()_8_rvv + .word ff_h264_biweight_pixels\w\()_8_rvv #elif __riscv_xlen == 64 .dword ff_h264_weight_pixels\w\()_8_rvv + .dword ff_h264_biweight_pixels\w\()_8_rvv #else .qword ff_h264_weight_pixels\w\()_8_rvv + .qword ff_h264_biweight_pixels\w\()_8_rvv #endif .endr endconst