From patchwork Sun Oct 2 11:54:59 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?q?R=C3=A9mi_Denis-Courmont?= X-Patchwork-Id: 38515 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a05:6a20:3b1c:b0:96:9ee8:5cfd with SMTP id c28csp744959pzh; Sun, 2 Oct 2022 04:55:18 -0700 (PDT) X-Google-Smtp-Source: AMsMyM43FxlmLsq2xmrvdOCW2nzIliztRp+v7rdCOEempK/0JKr1WNaixvxTjv0UQm3QN3G9x7Dp X-Received: by 2002:a17:907:960e:b0:782:68e3:620f with SMTP id gb14-20020a170907960e00b0078268e3620fmr11853094ejc.663.1664711718503; Sun, 02 Oct 2022 04:55:18 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1664711718; cv=none; d=google.com; s=arc-20160816; b=qtPwuf23aI9BrDuslsq3O+cIlG8jK1esu0j90gBMPDPINK7TBSqdMeHzaAB2rR4FUA K0N06L8Q5D/cUTyVpxf/KpUQq8j1iQqLC0Z21gOhIGsNFAhp/5Njdf6C84we4K0HD/UX Uzd6Jvk5fcQLBpPaPmmxT0d6j7BzsYKCe7oWrcI9jJBlfJcGkdZvMVvsJU/Fqe1qQ9G8 5VLbIlGUGHLrGL/dHkrdz8jRxYSsurSNYKNeHDHMw//vG98d2gNY+brA7xrCBfVT6yCP D/Kba9T3L0x/uWdYEW2R16HM4IV3YMPFRQ1H0VVg9BbNf8RVP58BRerTXKCiV+IVZeKt qtyw== 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=9ERhvO4smgUVW0lXRihyfemYiL4tqhh+3T9cK7JVZ3Q=; b=CG33Pzxi3P94BDRH7RYxRb7jse4c/K/VPoF0LOMsWjMp6c18wNIQ3Fh31DAu4y0yqG ZSFKJ0Zh9kYeE7PTrmVUrov1SZ/6iLSuOuNxESDsHuRoidNyXXQoneLRuLGbne33PkoW mBSWkg2zNom1YxQgdg6paD9YbzqZWZp69yl8zigOirHJZ3jPMKhs2dUr0U834vB7R2md /mZdclCfajq7kiW+JrYudbL8UPk0O7GuDXB3kA8Y+fMJAYB8zRlY8qjnwVmHUj5ybHRc N2MrOd8qWGDawfWcEFwPyd3a5DaooDJCwifJgHRoQxgmyaD1OjXu37lvCbqgv8wOaRdd xBKg== 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 f12-20020a50fc8c000000b0045889fa1b4csi4633824edq.484.2022.10.02.04.55.18; Sun, 02 Oct 2022 04:55:18 -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 58BCE68BB4C; Sun, 2 Oct 2022 14:55:09 +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 5B61868BAFE for ; Sun, 2 Oct 2022 14:55:02 +0300 (EEST) Received: from basile.remlab.net (localhost [IPv6:::1]) by ursule.remlab.net (Postfix) with ESMTP id E2F6CC0070 for ; Sun, 2 Oct 2022 14:55:01 +0300 (EEST) From: remi@remlab.net To: ffmpeg-devel@ffmpeg.org Date: Sun, 2 Oct 2022 14:54:59 +0300 Message-Id: <20221002115501.17996-2-remi@remlab.net> X-Mailer: git-send-email 2.37.2 In-Reply-To: <2650188.mvXUDI8C0e@basile.remlab.net> References: <2650188.mvXUDI8C0e@basile.remlab.net> MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 2/4] lavc/bswapdsp: RISC-V B bswap_buf 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: kvVOab3UfKSf From: Rémi Denis-Courmont Simply taking the Zbb REV8 instruction into use in a simple loop gives some significant savings: bswap_buf_c: 1081.0 bswap_buf_rvb_b: 771.0 But we can also use the 64-bit REV8 as a pseudo-SIMD instruction with just one additional shift, and one fewer load, effectively doubling the bandwidth. Consequently, this patch is useful even if the compile-time target has Zbb enabled for C code: bswap_buf_c: 1081.0 bswap_buf_rvb_b: 341.0 (this patch) On the other hand, this approach fails miserably for bswap16_buf as the ratio of shifts and stores becomes unfavorable compared to naïve C: bswap16_buf_c: 1542.0 bswap16_buf_rvb_b: 1803.7 Unrolling to process 128 bits (4 samples) at a time actually worsens performance ever so slightly: bswap_buf_c: 1081.0 bswap_buf_rvb_b: 408.5 --- libavcodec/bswapdsp.c | 4 +- libavcodec/bswapdsp.h | 1 + libavcodec/riscv/Makefile | 2 + libavcodec/riscv/bswapdsp_init.c | 38 ++++++++++++++++++ libavcodec/riscv/bswapdsp_rvb.S | 68 ++++++++++++++++++++++++++++++++ 5 files changed, 112 insertions(+), 1 deletion(-) create mode 100644 libavcodec/riscv/bswapdsp_init.c create mode 100644 libavcodec/riscv/bswapdsp_rvb.S diff --git a/libavcodec/bswapdsp.c b/libavcodec/bswapdsp.c index 4c4ea10acc..f0ea2b55c5 100644 --- a/libavcodec/bswapdsp.c +++ b/libavcodec/bswapdsp.c @@ -51,7 +51,9 @@ av_cold void ff_bswapdsp_init(BswapDSPContext *c) c->bswap_buf = bswap_buf; c->bswap16_buf = bswap16_buf; -#if ARCH_X86 +#if ARCH_RISCV + ff_bswapdsp_init_riscv(c); +#elif ARCH_X86 ff_bswapdsp_init_x86(c); #endif } diff --git a/libavcodec/bswapdsp.h b/libavcodec/bswapdsp.h index 4d19092254..6f4db66115 100644 --- a/libavcodec/bswapdsp.h +++ b/libavcodec/bswapdsp.h @@ -27,6 +27,7 @@ typedef struct BswapDSPContext { } BswapDSPContext; void ff_bswapdsp_init(BswapDSPContext *c); +void ff_bswapdsp_init_riscv(BswapDSPContext *c); void ff_bswapdsp_init_x86(BswapDSPContext *c); #endif /* AVCODEC_BSWAPDSP_H */ diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile index 0fb2c81c75..db4384bca7 100644 --- a/libavcodec/riscv/Makefile +++ b/libavcodec/riscv/Makefile @@ -3,6 +3,8 @@ RVV-OBJS-$(CONFIG_AAC_DECODER) += riscv/aacpsdsp_rvv.o OBJS-$(CONFIG_AUDIODSP) += riscv/audiodsp_init.o \ riscv/audiodsp_rvf.o RVV-OBJS-$(CONFIG_AUDIODSP) += riscv/audiodsp_rvv.o +OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_init.o \ + riscv/bswapdsp_rvb.o OBJS-$(CONFIG_FMTCONVERT) += riscv/fmtconvert_init.o RVV-OBJS-$(CONFIG_FMTCONVERT) += riscv/fmtconvert_rvv.o OBJS-$(CONFIG_IDCTDSP) += riscv/idctdsp_init.o diff --git a/libavcodec/riscv/bswapdsp_init.c b/libavcodec/riscv/bswapdsp_init.c new file mode 100644 index 0000000000..701dbeaaa6 --- /dev/null +++ b/libavcodec/riscv/bswapdsp_init.c @@ -0,0 +1,38 @@ +/* + * Copyright © 2022 Rémi Denis-Courmont. + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include + +#include "config.h" +#include "libavutil/attributes.h" +#include "libavutil/cpu.h" +#include "libavcodec/bswapdsp.h" + +void ff_bswap32_buf_rvb(uint32_t *dst, const uint32_t *src, int len); + +av_cold void ff_bswapdsp_init_riscv(BswapDSPContext *c) +{ +#if (__riscv_xlen >= 64) + int cpu_flags = av_get_cpu_flags(); + + if (cpu_flags & AV_CPU_FLAG_RVB_BASIC) + c->bswap_buf = ff_bswap32_buf_rvb; +#endif +} diff --git a/libavcodec/riscv/bswapdsp_rvb.S b/libavcodec/riscv/bswapdsp_rvb.S new file mode 100644 index 0000000000..91b47bf82d --- /dev/null +++ b/libavcodec/riscv/bswapdsp_rvb.S @@ -0,0 +1,68 @@ +/* + * Copyright © 2022 Rémi Denis-Courmont. + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" +#include "libavutil/riscv/asm.S" + +#if (__riscv_xlen >= 64) +func ff_bswap32_buf_rvb, zbb + andi t0, a1, 4 + beqz t0, 1f + /* Align a1 (input) to 64-bit */ + lwu t0, (a1) + addi a0, a0, 4 + rev8 t0, t0 + addi a2, a2, -1 + srli t0, t0, __riscv_xlen - 32 + addi a1, a1, 4 + sw t0, -4(a0) +1: + andi a3, a2, -2 + sh2add a2, a2, a0 + beqz a3, 3f + sh2add a3, a3, a0 +2: /* 2 elements (64 bits) at a time on a 64-bit boundary */ + ld t0, (a1) + addi a0, a0, 8 + rev8 t0, t0 +#if (__riscv_xlen == 64) + srli t2, t0, 32 + sw t0, -4(a0) +#else + srli t1, t0, __riscv_xlen - 64 + srli t2, t0, __riscv_xlen - 32 + sw t1, -4(a0) +#endif + addi a1, a1, 8 + sw t2, -8(a0) + bne a0, a3, 2b +3: + beq a0, a2, 5f +4: /* Process last element */ + lwu t0, (a1) + addi a0, a0, 4 + rev8 t0, t0 + addi a1, a1, 4 + srli t0, t0, __riscv_xlen - 32 + sw t0, -4(a0) +5: + ret +endfunc +#endif