From patchwork Thu Jul 12 12:45:26 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Shiyou Yin X-Patchwork-Id: 9699 Delivered-To: ffmpegpatchwork@gmail.com Received: by 2002:a02:104:0:0:0:0:0 with SMTP id c4-v6csp1618924jad; Thu, 12 Jul 2018 05:45:47 -0700 (PDT) X-Google-Smtp-Source: AAOMgpcah2LOAZfQSL5Z+gfWYa3gOIcceTU2372catQWA6onFvm/exTbKLRhgcCisZZ+8jpjW1Pu X-Received: by 2002:adf:af27:: with SMTP id z36-v6mr1587946wrc.59.1531399547010; Thu, 12 Jul 2018 05:45:47 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1531399546; cv=none; d=google.com; s=arc-20160816; b=D9pI7BAn7xSuRaOnpEs/R5s27gkJXN7uOuJKhn9V81ErfAWr6U4E4WO9ecEyIZ0RsP trTg80262LAbedVPud9Gk3cee3DvVjBqzefZ+IvrxBvgEK/3DwlgdSwBaztt+s8BfO/e m/wWR7glLx9A9P4skFW5Y3KbIxbn7LoTJcI2L20HpVBMJhr3Z29b3rRX7ejkNS+GThT6 62w6ds1ln8+lxdZV9ORe4bvcnWcauL69MjlHZJfl6+sjNVHD03pmKAOteBrWNPczPLIV k4JmcXtkl9gxmK4J36CRj9dw/WdVZRSvp88VjWVKQPYTqU2O/GnJ19d2Ps9SGJ8V3Rcj 4Tkg== 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:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:message-id:date:to:from:delivered-to :arc-authentication-results; bh=teOIlEbSUD5L8IZxNhPprE9ocUv4znIoy6LzlwGrJXE=; b=u4vdtUTWSJsTabMJ3Wolrom+MXtvgJ86z9CpTv4VvAEF/d1jPnRN7FhZ9Asu0zUzPs d05yFb8BE3H8nShBT9uJUjnS9nV1dBtOUb/3+DEzFNIG/iN/5xO5f70o1zvv88R0dNvI eU23XTnGmNhoH7IiTPYkx494gOj9RCKeKnNviRfV+RXT7eRg1aKimdvYOPz08/6QWyZZ wVt+ieBKAaizZIQRUzVLm1Hlvv1Q44ViIvFBJ3C0OCzq7phzT4oN1Eua0632GG679Qxf TVg03VilWm5sn5baHMdti6lgpX6orSMEP/Kdb+bkgukmNeiOZsOlgJEU1JWG+6grPVUH r01A== 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 c4-v6si19653443wri.363.2018.07.12.05.45.46; Thu, 12 Jul 2018 05:45:46 -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 4573768A6E8; Thu, 12 Jul 2018 15:45:23 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail.loongson.cn (mail.loongson.cn [114.242.206.163]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 18F1268A6E1 for ; Thu, 12 Jul 2018 15:45:19 +0300 (EEST) Received: from localhost (unknown [202.141.160.96]) by mail (Coremail) with SMTP id QMiowPAxmOFETEdbI5uTAA--.51136S3; Thu, 12 Jul 2018 20:40:36 +0800 (CST) From: Shiyou Yin To: ffmpeg-devel@ffmpeg.org Date: Thu, 12 Jul 2018 20:45:26 +0800 Message-Id: <1531399526-15262-1-git-send-email-yinshiyou-hf@loongson.cn> X-Mailer: git-send-email 2.1.0 X-CM-TRANSID: QMiowPAxmOFETEdbI5uTAA--.51136S3 X-Coremail-Antispam: 1UD129KBjvJXoW3uw43GF4rur1UZF1DKF45Wrg_yoWDtFyUpa 48ArnYv3W2yF45Xa17Jr1UCry3Zr4DtFsakFWqgF10v3s5tFn7XrWxGrykGrWYka4avF4D Zr1DAa4Sywn5GF7anT9S1TB71UUUUUUqnTZGkaVYY2UrUUUUjbIjqfuFe4nvWSU5nxnvy2 9KBjDU0xBIdaVrnRJUUUkIb4IE77IF4wAFF20E14v26r1j6r4UM7CY07I20VC2zVCF04k2 6cxKx2IYs7xG6rWj6s0DM7CIcVAFz4kK6r1j6r18M28lY4IEw2IIxxk0rwA2F7IY1VAKz4 vEj48ve4kI8wA2z4x0Y4vE2Ix0cI8IcVAFwI0_Gr0_Xr1l84ACjcxK6xIIjxv20xvEc7Cj xVAFwI0_Gr0_Cr1l84ACjcxK6I8E87Iv67AKxVWxJr0_GcWl84ACjcxK6I8E87Iv6xkF7I 0E14v26rxl6s0DM2AIxVAIcxkEcVAq07x20xvEncxIr21l5I8CrVACY4xI64kE6c02F40E x7xfMcIj6xIIjxv20xvE14v26r1q6rW5McIj6I8E87Iv67AKxVW8JVWxJwAm72CE4IkC6x 0Yz7v_Jr0_Gr1lF7xvr2IYc2Ij64vIr41lF7xvr2IYc2Ij64vIr40E4x8a64kEw24lc2xS Y4AK67AK6ryUMxAIw28IcxkI7VAKI48JMxC20s026xCaFVCjc4AY6r1j6r4UMI8I3I0E5I 8CrVAFwI0_Jr0_Jr4lx2IqxVCjr7xvwVAFwI0_JrI_JrWlx4CE17CEb7AF67AKxVWUJVWU XwCI42IY6xIIjxv20xvE14v26r1I6r4UMIIF0xvE2Ix0cI8IcVCY1x0267AKxVWUJVW8Jw CI42IY6xAIw20EY4v20xvaj40_Gr0_Zr1lIxAIcVC2z280aVAFwI0_Gr0_Cr1lIxAIcVC2 z280aVCY1x0267AKxVW8JVW8JrUvcSsGvfC2KfnxnUUI43ZEXa7IU05rc3UUUUU== X-CM-SenderInfo: p1lq2x5l1r3gtki6z05rqj20fqof0/ Subject: [FFmpeg-devel] [PATCH 8/9] avcodec/mips: [loongson] reoptimize put&add pixels with mmi v2. 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" Reoptimized following functions with mmi: 1. ff_put_pixels4_8_mmi. 2. ff_put_pixels8_8_mmi. 3. ff_put_pixels16_8_mmi. 4. ff_add_pixels_clamped_mmi. This optimization improved mpeg4 decoding performance about 6fps(from 158fps to 164fps, tested on loongson 3A3000). Change-Id: I6baff96587c5441cd8b74c9362dc01ca1d20c1a0 Signed-off-by: Shiyou Yin --- libavcodec/mips/hpeldsp_mmi.c | 106 +++++++++++++++++++++--------------------- libavcodec/mips/idctdsp_mmi.c | 62 +++++++++++------------- 2 files changed, 80 insertions(+), 88 deletions(-) diff --git a/libavcodec/mips/hpeldsp_mmi.c b/libavcodec/mips/hpeldsp_mmi.c index 1bba70e..a0597c1 100644 --- a/libavcodec/mips/hpeldsp_mmi.c +++ b/libavcodec/mips/hpeldsp_mmi.c @@ -29,36 +29,35 @@ void ff_put_pixels4_8_mmi(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) { - double ftmp[2]; - mips_reg addr[2]; + double ftmp[4]; DECLARE_VAR_LOW32; - DECLARE_VAR_ADDRT; __asm__ volatile ( - PTR_ADDU "%[addr1], %[line_size], %[line_size] \n\t" "1: \n\t" - PTR_ADDU "%[addr0], %[pixels], %[line_size] \n\t" MMI_ULWC1(%[ftmp0], %[pixels], 0x00) - MMI_ULWC1(%[ftmp1], %[addr0], 0x00) - MMI_SWC1(%[ftmp0], %[block], 0x00) - MMI_SWXC1(%[ftmp1], %[block], %[line_size], 0x00) - PTR_ADDU "%[pixels], %[pixels], %[addr1] \n\t" - PTR_ADDU "%[block], %[block], %[addr1] \n\t" + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_ULWC1(%[ftmp1], %[pixels], 0x00) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_ULWC1(%[ftmp2], %[pixels], 0x00) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_ULWC1(%[ftmp3], %[pixels], 0x00) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + + PTR_ADDI "%[h], %[h], -0x04 \n\t" - PTR_ADDU "%[addr0], %[pixels], %[line_size] \n\t" - MMI_ULWC1(%[ftmp0], %[pixels], 0x00) - MMI_ULWC1(%[ftmp1], %[addr0], 0x00) MMI_SWC1(%[ftmp0], %[block], 0x00) - MMI_SWXC1(%[ftmp1], %[block], %[line_size], 0x00) - PTR_ADDU "%[pixels], %[pixels], %[addr1] \n\t" - PTR_ADDU "%[block], %[block], %[addr1] \n\t" + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SWC1(%[ftmp1], %[block], 0x00) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SWC1(%[ftmp2], %[block], 0x00) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SWC1(%[ftmp3], %[block], 0x00) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" - PTR_ADDI "%[h], %[h], -0x04 \n\t" "bnez %[h], 1b \n\t" : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), + [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), RESTRICT_ASM_LOW32 - RESTRICT_ASM_ADDRT - [addr0]"=&r"(addr[0]), [addr1]"=&r"(addr[1]), [block]"+&r"(block), [pixels]"+&r"(pixels), [h]"+&r"(h) : [line_size]"r"((mips_reg)line_size) @@ -69,37 +68,35 @@ void ff_put_pixels4_8_mmi(uint8_t *block, const uint8_t *pixels, void ff_put_pixels8_8_mmi(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) { - double ftmp[2]; - mips_reg addr[3]; + double ftmp[4]; DECLARE_VAR_ALL64; __asm__ volatile ( - PTR_ADDU "%[addr1], %[line_size], %[line_size] \n\t" "1: \n\t" MMI_ULDC1(%[ftmp0], %[pixels], 0x00) - PTR_ADDU "%[addr0], %[pixels], %[line_size] \n\t" - MMI_ULDC1(%[ftmp1], %[addr0], 0x00) - MMI_SDC1(%[ftmp0], %[block], 0x00) - PTR_ADDU "%[addr2], %[block], %[line_size] \n\t" - MMI_SDC1(%[ftmp1], %[addr2], 0x00) - PTR_ADDU "%[pixels], %[pixels], %[addr1] \n\t" - PTR_ADDU "%[block], %[block], %[addr1] \n\t" + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_ULDC1(%[ftmp1], %[pixels], 0x00) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_ULDC1(%[ftmp2], %[pixels], 0x00) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_ULDC1(%[ftmp3], %[pixels], 0x00) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + + PTR_ADDI "%[h], %[h], -0x04 \n\t" - MMI_ULDC1(%[ftmp0], %[pixels], 0x00) - PTR_ADDU "%[addr0], %[pixels], %[line_size] \n\t" - MMI_ULDC1(%[ftmp1], %[addr0], 0x00) MMI_SDC1(%[ftmp0], %[block], 0x00) - PTR_ADDU "%[addr2], %[block], %[line_size] \n\t" - MMI_SDC1(%[ftmp1], %[addr2], 0x00) - PTR_ADDU "%[pixels], %[pixels], %[addr1] \n\t" - PTR_ADDU "%[block], %[block], %[addr1] \n\t" + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SDC1(%[ftmp1], %[block], 0x00) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SDC1(%[ftmp2], %[block], 0x00) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SDC1(%[ftmp3], %[block], 0x00) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" - PTR_ADDI "%[h], %[h], -0x04 \n\t" "bnez %[h], 1b \n\t" : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), + [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), RESTRICT_ASM_ALL64 - [addr0]"=&r"(addr[0]), [addr1]"=&r"(addr[1]), - [addr2]"=&r"(addr[2]), [block]"+&r"(block), [pixels]"+&r"(pixels), [h]"+&r"(h) : [line_size]"r"((mips_reg)line_size) @@ -110,7 +107,7 @@ void ff_put_pixels8_8_mmi(uint8_t *block, const uint8_t *pixels, void ff_put_pixels16_8_mmi(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) { - double ftmp[4]; + double ftmp[8]; DECLARE_VAR_ALL64; __asm__ volatile ( @@ -120,31 +117,34 @@ void ff_put_pixels16_8_mmi(uint8_t *block, const uint8_t *pixels, PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" MMI_ULDC1(%[ftmp1], %[pixels], 0x00) MMI_ULDC1(%[ftmp3], %[pixels], 0x08) - MMI_SDC1(%[ftmp0], %[block], 0x00) - MMI_SDC1(%[ftmp2], %[block], 0x08) - PTR_ADDU "%[block], %[block], %[line_size] \n\t" - MMI_SDC1(%[ftmp1], %[block], 0x00) - MMI_SDC1(%[ftmp3], %[block], 0x08) PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" - PTR_ADDU "%[block], %[block], %[line_size] \n\t" - - MMI_ULDC1(%[ftmp0], %[pixels], 0x00) - MMI_ULDC1(%[ftmp2], %[pixels], 0x08) + MMI_ULDC1(%[ftmp4], %[pixels], 0x00) + MMI_ULDC1(%[ftmp6], %[pixels], 0x08) PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" - MMI_ULDC1(%[ftmp1], %[pixels], 0x00) - MMI_ULDC1(%[ftmp3], %[pixels], 0x08) + MMI_ULDC1(%[ftmp5], %[pixels], 0x00) + MMI_ULDC1(%[ftmp7], %[pixels], 0x08) + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + + PTR_ADDI "%[h], %[h], -0x04 \n\t" + MMI_SDC1(%[ftmp0], %[block], 0x00) MMI_SDC1(%[ftmp2], %[block], 0x08) PTR_ADDU "%[block], %[block], %[line_size] \n\t" MMI_SDC1(%[ftmp1], %[block], 0x00) MMI_SDC1(%[ftmp3], %[block], 0x08) - PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SDC1(%[ftmp4], %[block], 0x00) + MMI_SDC1(%[ftmp6], %[block], 0x08) + PTR_ADDU "%[block], %[block], %[line_size] \n\t" + MMI_SDC1(%[ftmp5], %[block], 0x00) + MMI_SDC1(%[ftmp7], %[block], 0x08) PTR_ADDU "%[block], %[block], %[line_size] \n\t" - PTR_ADDI "%[h], %[h], -0x04 \n\t" "bnez %[h], 1b \n\t" : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), + [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), + [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), RESTRICT_ASM_ALL64 [block]"+&r"(block), [pixels]"+&r"(pixels), [h]"+&r"(h) diff --git a/libavcodec/mips/idctdsp_mmi.c b/libavcodec/mips/idctdsp_mmi.c index b797965..a022e18 100644 --- a/libavcodec/mips/idctdsp_mmi.c +++ b/libavcodec/mips/idctdsp_mmi.c @@ -174,49 +174,41 @@ void ff_put_signed_pixels_clamped_mmi(const int16_t *block, void ff_add_pixels_clamped_mmi(const int16_t *block, uint8_t *av_restrict pixels, ptrdiff_t line_size) { - double ftmp[8]; + double ftmp[9]; uint64_t tmp[1]; - mips_reg addr[1]; - DECLARE_VAR_ALL64; - DECLARE_VAR_ADDRT; - __asm__ volatile ( - "li %[tmp0], 0x04 \n\t" - "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" - "1: \n\t" - MMI_LDC1(%[ftmp1], %[block], 0x00) - MMI_LDC1(%[ftmp2], %[block], 0x08) - MMI_LDC1(%[ftmp3], %[block], 0x10) - MMI_LDC1(%[ftmp4], %[block], 0x18) + "li %[tmp0], 0x04 \n\t" + "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" + "1: \n\t" + + "gslqc1 %[ftmp2], %[ftmp1], 0x00(%[block]) \n\t" + "gslqc1 %[ftmp4], %[ftmp3], 0x10(%[block]) \n\t" + PTR_ADDIU "%[block], %[block], 0x20 \n\t" MMI_LDC1(%[ftmp5], %[pixels], 0x00) - MMI_LDXC1(%[ftmp6], %[pixels], %[line_size], 0x00) - "mov.d %[ftmp7], %[ftmp5] \n\t" - "punpcklbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t" - "punpckhbh %[ftmp7], %[ftmp7], %[ftmp0] \n\t" - "paddh %[ftmp1], %[ftmp1], %[ftmp5] \n\t" - "paddh %[ftmp2], %[ftmp2], %[ftmp7] \n\t" - "mov.d %[ftmp7], %[ftmp6] \n\t" - "punpcklbh %[ftmp6], %[ftmp6], %[ftmp0] \n\t" - "punpckhbh %[ftmp7], %[ftmp7], %[ftmp0] \n\t" - "paddh %[ftmp3], %[ftmp3], %[ftmp6] \n\t" - "paddh %[ftmp4], %[ftmp4], %[ftmp7] \n\t" - "packushb %[ftmp1], %[ftmp1], %[ftmp2] \n\t" - "packushb %[ftmp3], %[ftmp3], %[ftmp4] \n\t" + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_LDC1(%[ftmp6], %[pixels], 0x00) + PTR_SUBU "%[pixels], %[pixels], %[line_size] \n\t" + "punpckhbh %[ftmp7], %[ftmp5], %[ftmp0] \n\t" + "punpcklbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t" + "punpckhbh %[ftmp8], %[ftmp6], %[ftmp0] \n\t" + "punpcklbh %[ftmp6], %[ftmp6], %[ftmp0] \n\t" + "paddh %[ftmp1], %[ftmp1], %[ftmp5] \n\t" + "paddh %[ftmp2], %[ftmp2], %[ftmp7] \n\t" + "paddh %[ftmp3], %[ftmp3], %[ftmp6] \n\t" + "paddh %[ftmp4], %[ftmp4], %[ftmp8] \n\t" + "packushb %[ftmp1], %[ftmp1], %[ftmp2] \n\t" + "packushb %[ftmp3], %[ftmp3], %[ftmp4] \n\t" MMI_SDC1(%[ftmp1], %[pixels], 0x00) - MMI_SDXC1(%[ftmp3], %[pixels], %[line_size], 0x00) - "addi %[tmp0], %[tmp0], -0x01 \n\t" - PTR_ADDIU "%[block], %[block], 0x20 \n\t" - PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" - PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" - "bnez %[tmp0], 1b" + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + MMI_SDC1(%[ftmp3], %[pixels], 0x00) + "addi %[tmp0], %[tmp0], -0x01 \n\t" + PTR_ADDU "%[pixels], %[pixels], %[line_size] \n\t" + "bnez %[tmp0], 1b \n\t" : [ftmp0]"=&f"(ftmp[0]), [ftmp1]"=&f"(ftmp[1]), [ftmp2]"=&f"(ftmp[2]), [ftmp3]"=&f"(ftmp[3]), [ftmp4]"=&f"(ftmp[4]), [ftmp5]"=&f"(ftmp[5]), [ftmp6]"=&f"(ftmp[6]), [ftmp7]"=&f"(ftmp[7]), - [tmp0]"=&r"(tmp[0]), - RESTRICT_ASM_ALL64 - RESTRICT_ASM_ADDRT - [addr0]"=&r"(addr[0]), + [ftmp8]"=&f"(ftmp[8]), [tmp0]"=&r"(tmp[0]), [pixels]"+&r"(pixels), [block]"+&r"(block) : [line_size]"r"((mips_reg)line_size) : "memory"