From patchwork Fri Jan 18 06:42:20 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Shiyou Yin X-Patchwork-Id: 11787 Return-Path: X-Original-To: patchwork@ffaux-bg.ffmpeg.org Delivered-To: patchwork@ffaux-bg.ffmpeg.org Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100]) by ffaux.localdomain (Postfix) with ESMTP id AD69B44DCFD for ; Fri, 18 Jan 2019 08:42:31 +0200 (EET) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id C440C68ABC6; Fri, 18 Jan 2019 08:42:19 +0200 (EET) 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 B6DBF68ABB9 for ; Fri, 18 Jan 2019 08:42:10 +0200 (EET) Received: from localhost (unknown [210.45.123.188]) by mail (Coremail) with SMTP id QMiowPDxhuRLdUFc0c1zAA--.22250S3; Fri, 18 Jan 2019 14:42:20 +0800 (CST) From: Shiyou Yin To: ffmpeg-devel@ffmpeg.org Date: Fri, 18 Jan 2019 14:42:20 +0800 Message-Id: <1547793740-21403-1-git-send-email-yinshiyou-hf@loongson.cn> X-Mailer: git-send-email 2.1.0 X-CM-TRANSID: QMiowPDxhuRLdUFc0c1zAA--.22250S3 X-Coremail-Antispam: 1UD129KBjvJXoW3Gryxuw13Jw4fZF15Aw4Utwb_yoWfZrWxpF ykCwsav3WxtFsruw47Jr1UWr13Arn8XF1akFZFgFy8Xw1YqFn7XrZ7Jry8G3WagFyj9Fyf ZrnxWFyayw18Cw7anT9S1TB71UUUUU7qnTZGkaVYY2UrUUUUjbIjqfuFe4nvWSU5nxnvy2 9KBjDU0xBIdaVrnRJUUUvKb7Iv0xC_Kw4lb4IE77IF4wAFF20E14v26r1j6r4UM7CY07I2 0VC2zVCF04k26cxKx2IYs7xG6rWj6s0DM7CIcVAFz4kK6r1j6r18M28lY4IEw2IIxxk0rw A2F7IY1VAKz4vEj48ve4kI8wA2z4x0Y4vE2Ix0cI8IcVAFwI0_Xr0_Ar1l84ACjcxK6xII jxv20xvEc7CjxVAFwI0_Gr0_Cr1l84ACjcxK6I8E87Iv67AKxVW8Jr0_Cr1UM28EF7xvwV C2z280aVCY1x0267AKxVWxJr0_GcWle2I262IYc4CY6c8Ij28IcVAaY2xG8wAqx4xG64xv F2IEw4CE5I8CrVC2j2WlYx0E2Ix0cI8IcVAFwI0_JrI_JrylYx0Ex4A2jsIE14v26r1j6r 4UMcvjeVCFs4IE7xkEbVWUJVW8JwACjcxG0xvY0x0EwIxGrwACjcxG0xvY0x0EwIxGrVCF 72vEw4AK0wCY02Avz4vE14v_GF1l42xK82IYc2Ij64vIr41l4I8I3I0E4IkC6x0Yz7v_Jr 0_Gr1lx2IqxVAqx4xG67AKxVWUJVWUGwC20s026x8GjcxK67AKxVWUGVWUWwC2zVAF1VAY 17CE14v26r1j6r15MIIYrxkI7VAKI48JMIIF0xvE2Ix0cI8IcVAFwI0_Jr0_JF4lIxAIcV C0I7IYx2IY6xkF7I0E14v26r1j6r4UMIIF0xvE42xK8VAvwI8IcIk0rVWrJr0_WFyUJwCI 42IY6I8E87Iv67AKxVWUJVW8JwCI42IY6I8E87Iv6xkF7I0E14v26r1j6r4UYxBIdaVFxh VjvjDU0xZFpf9x07bzuWJUUUUU= X-CM-SenderInfo: p1lq2x5l1r3gtki6z05rqj20fqof0/ Subject: [FFmpeg-devel] [PATCH] avcodec/mips: [loongson] optimize put_hevc_pel_bi_pixels_8 with mmi. 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" Optimize put_hevc_pel_bi_pixels_8 with mmi in the case width=8/16/24/32/48/64. This optimization improved HEVC decoding performance 2%(1.77x to 1.81x, tested on loongson 3A3000). --- libavcodec/mips/Makefile | 1 + libavcodec/mips/hevcdsp_init_mips.c | 23 ++++++ libavcodec/mips/hevcdsp_mips.h | 19 +++++ libavcodec/mips/hevcdsp_mmi.c | 135 ++++++++++++++++++++++++++++++++++++ 4 files changed, 178 insertions(+) create mode 100644 libavcodec/mips/hevcdsp_mmi.c diff --git a/libavcodec/mips/Makefile b/libavcodec/mips/Makefile index 3571207..3029872 100644 --- a/libavcodec/mips/Makefile +++ b/libavcodec/mips/Makefile @@ -86,3 +86,4 @@ MMI-OBJS-$(CONFIG_VP8_DECODER) += mips/vp8dsp_mmi.o MMI-OBJS-$(CONFIG_HPELDSP) += mips/hpeldsp_mmi.o MMI-OBJS-$(CONFIG_VC1_DECODER) += mips/vc1dsp_mmi.o MMI-OBJS-$(CONFIG_WMV2DSP) += mips/wmv2dsp_mmi.o +MMI-OBJS-$(CONFIG_HEVC_DECODER) += mips/hevcdsp_mmi.o diff --git a/libavcodec/mips/hevcdsp_init_mips.c b/libavcodec/mips/hevcdsp_init_mips.c index 776d13e..41c9001 100644 --- a/libavcodec/mips/hevcdsp_init_mips.c +++ b/libavcodec/mips/hevcdsp_init_mips.c @@ -20,6 +20,26 @@ #include "libavcodec/mips/hevcdsp_mips.h" +#if HAVE_MMI +static av_cold void hevc_dsp_init_mmi(HEVCDSPContext *c, + const int bit_depth) +{ + if (8 == bit_depth) { + c->put_hevc_qpel_bi[3][0][0] = ff_hevc_put_hevc_pel_bi_pixels8_8_mmi; + c->put_hevc_qpel_bi[5][0][0] = ff_hevc_put_hevc_pel_bi_pixels16_8_mmi; + c->put_hevc_qpel_bi[6][0][0] = ff_hevc_put_hevc_pel_bi_pixels24_8_mmi; + c->put_hevc_qpel_bi[7][0][0] = ff_hevc_put_hevc_pel_bi_pixels32_8_mmi; + c->put_hevc_qpel_bi[8][0][0] = ff_hevc_put_hevc_pel_bi_pixels48_8_mmi; + c->put_hevc_qpel_bi[9][0][0] = ff_hevc_put_hevc_pel_bi_pixels64_8_mmi; + + c->put_hevc_epel_bi[3][0][0] = ff_hevc_put_hevc_pel_bi_pixels8_8_mmi; + c->put_hevc_epel_bi[5][0][0] = ff_hevc_put_hevc_pel_bi_pixels16_8_mmi; + c->put_hevc_epel_bi[6][0][0] = ff_hevc_put_hevc_pel_bi_pixels24_8_mmi; + c->put_hevc_epel_bi[7][0][0] = ff_hevc_put_hevc_pel_bi_pixels32_8_mmi; + } +} +#endif // #if HAVE_MMI + #if HAVE_MSA static av_cold void hevc_dsp_init_msa(HEVCDSPContext *c, const int bit_depth) @@ -448,6 +468,9 @@ static av_cold void hevc_dsp_init_msa(HEVCDSPContext *c, void ff_hevc_dsp_init_mips(HEVCDSPContext *c, const int bit_depth) { +#if HAVE_MMI + hevc_dsp_init_mmi(c, bit_depth); +#endif // #if HAVE_MMI #if HAVE_MSA hevc_dsp_init_msa(c, bit_depth); #endif // #if HAVE_MSA diff --git a/libavcodec/mips/hevcdsp_mips.h b/libavcodec/mips/hevcdsp_mips.h index 1573d1c..ff9401c 100644 --- a/libavcodec/mips/hevcdsp_mips.h +++ b/libavcodec/mips/hevcdsp_mips.h @@ -479,4 +479,23 @@ void ff_hevc_addblk_32x32_msa(uint8_t *dst, int16_t *pi16Coeffs, ptrdiff_t stride); void ff_hevc_idct_luma_4x4_msa(int16_t *pi16Coeffs); +/* Loongson optimization */ +#define L_BI_MC(PEL, DIR, WIDTH, TYPE) \ +void ff_hevc_put_hevc_##PEL##_bi_##DIR##WIDTH##_8_##TYPE(uint8_t *dst, \ + ptrdiff_t dst_stride, \ + uint8_t *src, \ + ptrdiff_t src_stride, \ + int16_t *src_16bit, \ + int height, \ + intptr_t mx, \ + intptr_t my, \ + int width) + +L_BI_MC(pel, pixels, 8, mmi); +L_BI_MC(pel, pixels, 16, mmi); +L_BI_MC(pel, pixels, 24, mmi); +L_BI_MC(pel, pixels, 32, mmi); +L_BI_MC(pel, pixels, 48, mmi); +L_BI_MC(pel, pixels, 64, mmi); +#undef L_BI_MC #endif // #ifndef AVCODEC_MIPS_HEVCDSP_MIPS_H diff --git a/libavcodec/mips/hevcdsp_mmi.c b/libavcodec/mips/hevcdsp_mmi.c new file mode 100644 index 0000000..e9403c5 --- /dev/null +++ b/libavcodec/mips/hevcdsp_mmi.c @@ -0,0 +1,135 @@ +/* + * Copyright (c) 2019 Shiyou Yin (yinshiyou-hf@loongson.cn) + * + * 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 "libavcodec/bit_depth_template.c" +#include "libavcodec/mips/hevcdsp_mips.h" +#include "libavutil/mips/mmiutils.h" + +#define PUT_HEVC_PEL_BI_PIXELS(w, x_step, src_step, dst_step, src2_step) \ +void ff_hevc_put_hevc_pel_bi_pixels##w##_8_mmi(uint8_t *_dst, \ + ptrdiff_t _dststride, \ + uint8_t *_src, \ + ptrdiff_t _srcstride, \ + int16_t *src2, int height, \ + intptr_t mx, intptr_t my, \ + int width) \ +{ \ + int x, y; \ + pixel *src = (pixel *)_src; \ + ptrdiff_t srcstride = _srcstride / sizeof(pixel); \ + pixel *dst = (pixel *)_dst; \ + ptrdiff_t dststride = _dststride / sizeof(pixel); \ + uint64_t ftmp[12]; \ + uint64_t rtmp[1]; \ + int shift = 7; \ + \ + y = height; \ + x = width >> 3; \ + __asm__ volatile( \ + "xor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \ + "li %[rtmp0], 0x06 \n\t" \ + "dmtc1 %[rtmp0], %[ftmp1] \n\t" \ + "li %[rtmp0], 0x10 \n\t" \ + "dmtc1 %[rtmp0], %[ftmp10] \n\t" \ + "li %[rtmp0], 0x40 \n\t" \ + "dmtc1 %[rtmp0], %[offset] \n\t" \ + "punpcklhw %[offset], %[offset], %[offset] \n\t" \ + "punpcklwd %[offset], %[offset], %[offset] \n\t" \ + \ + "1: \n\t" \ + "2: \n\t" \ + "gsldlc1 %[ftmp5], 0x07(%[src]) \n\t" \ + "gsldrc1 %[ftmp5], 0x00(%[src]) \n\t" \ + "gsldlc1 %[ftmp2], 0x07(%[src2]) \n\t" \ + "gsldrc1 %[ftmp2], 0x00(%[src2]) \n\t" \ + "gsldlc1 %[ftmp3], 0x0f(%[src2]) \n\t" \ + "gsldrc1 %[ftmp3], 0x08(%[src2]) \n\t" \ + "punpcklbh %[ftmp4], %[ftmp5], %[ftmp0] \n\t" \ + "punpckhbh %[ftmp5], %[ftmp5], %[ftmp0] \n\t" \ + "psllh %[ftmp4], %[ftmp4], %[ftmp1] \n\t" \ + "psllh %[ftmp5], %[ftmp5], %[ftmp1] \n\t" \ + "paddh %[ftmp4], %[ftmp4], %[offset] \n\t" \ + "paddh %[ftmp5], %[ftmp5], %[offset] \n\t" \ + "punpcklhw %[ftmp6], %[ftmp4], %[ftmp0] \n\t" \ + "punpckhhw %[ftmp7], %[ftmp4], %[ftmp0] \n\t" \ + "punpcklhw %[ftmp8], %[ftmp5], %[ftmp0] \n\t" \ + "punpckhhw %[ftmp9], %[ftmp5], %[ftmp0] \n\t" \ + "punpcklhw %[ftmp4], %[ftmp0], %[ftmp3] \n\t" \ + "punpckhhw %[ftmp5], %[ftmp0], %[ftmp3] \n\t" \ + "punpckhhw %[ftmp3], %[ftmp0], %[ftmp2] \n\t" \ + "punpcklhw %[ftmp2], %[ftmp0], %[ftmp2] \n\t" \ + "psraw %[ftmp2], %[ftmp2], %[ftmp10] \n\t" \ + "psraw %[ftmp3], %[ftmp3], %[ftmp10] \n\t" \ + "psraw %[ftmp4], %[ftmp4], %[ftmp10] \n\t" \ + "psraw %[ftmp5], %[ftmp5], %[ftmp10] \n\t" \ + "paddw %[ftmp2], %[ftmp2], %[ftmp6] \n\t" \ + "paddw %[ftmp3], %[ftmp3], %[ftmp7] \n\t" \ + "paddw %[ftmp4], %[ftmp4], %[ftmp8] \n\t" \ + "paddw %[ftmp5], %[ftmp5], %[ftmp9] \n\t" \ + "psraw %[ftmp2], %[ftmp2], %[shift] \n\t" \ + "psraw %[ftmp3], %[ftmp3], %[shift] \n\t" \ + "psraw %[ftmp4], %[ftmp4], %[shift] \n\t" \ + "psraw %[ftmp5], %[ftmp5], %[shift] \n\t" \ + "packsswh %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \ + "packsswh %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \ + "pcmpgth %[ftmp3], %[ftmp2], %[ftmp0] \n\t" \ + "pcmpgth %[ftmp5], %[ftmp4], %[ftmp0] \n\t" \ + "and %[ftmp2], %[ftmp2], %[ftmp3] \n\t" \ + "and %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \ + "packushb %[ftmp2], %[ftmp2], %[ftmp4] \n\t" \ + "gssdlc1 %[ftmp2], 0x07(%[dst]) \n\t" \ + "gssdrc1 %[ftmp2], 0x00(%[dst]) \n\t" \ + \ + "daddi %[x], %[x], -0x01 \n\t" \ + PTR_ADDIU "%[src], %[src], 0x08 \n\t" \ + PTR_ADDIU "%[dst], %[dst], 0x08 \n\t" \ + PTR_ADDIU "%[src2], %[src2], 0x10 \n\t" \ + "bnez %[x], 2b \n\t" \ + \ + PTR_ADDIU "%[src], %[src], " #src_step " \n\t" \ + PTR_ADDIU "%[dst], %[dst], " #dst_step " \n\t" \ + PTR_ADDIU "%[src2], %[src2], " #src2_step " \n\t" \ + "li %[x], " #x_step " \n\t" \ + "daddi %[y], %[y], -0x01 \n\t" \ + PTR_ADDU "%[src], %[src], %[srcstride] \n\t" \ + PTR_ADDU "%[dst], %[dst], %[dststride] \n\t" \ + PTR_ADDIU "%[src2], %[src2], 0x80 \n\t" \ + "bnez %[y], 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]), \ + [ftmp8]"=&f"(ftmp[8]), [ftmp9]"=&f"(ftmp[9]), \ + [ftmp10]"=&f"(ftmp[10]), [offset]"=&f"(ftmp[11]), \ + [src2]"+&r"(src2), [dst]"+&r"(dst), [src]"+&r"(src), \ + [x]"+&r"(x), [y]"+&r"(y), [rtmp0]"=&r"(rtmp[0]) \ + : [dststride]"r"(dststride), [shift]"f"(shift), \ + [srcstride]"r"(srcstride) \ + : "memory" \ + ); \ +} \ + +PUT_HEVC_PEL_BI_PIXELS(8, 1, -8, -8, -16); +PUT_HEVC_PEL_BI_PIXELS(16, 2, -16, -16, -32); +PUT_HEVC_PEL_BI_PIXELS(24, 3, -24, -24, -48); +PUT_HEVC_PEL_BI_PIXELS(32, 4, -32, -32, -64); +PUT_HEVC_PEL_BI_PIXELS(48, 6, -48, -48, -96); +PUT_HEVC_PEL_BI_PIXELS(64, 8, -64, -64, -128); +