From patchwork Tue Jun 19 06:21:10 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Shiyou Yin X-Patchwork-Id: 9443 Delivered-To: ffmpegpatchwork@gmail.com Received: by 2002:a02:11c:0:0:0:0:0 with SMTP id c28-v6csp4648795jad; Mon, 18 Jun 2018 23:21:56 -0700 (PDT) X-Google-Smtp-Source: ADUXVKJYTFd2FhFhg5GzoHRz4+ClOKLvfSvBgfzU5SvzKkxfT9+Ame2ka11I4U3YFQBEHDtjAZbX X-Received: by 2002:adf:8e30:: with SMTP id n45-v6mr12215243wrb.27.1529389316195; Mon, 18 Jun 2018 23:21:56 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1529389316; cv=none; d=google.com; s=arc-20160816; b=kfdlrZf3vgsx4tRePDLXMGAdNmL2XlTA1FgQkcMU2dpPWC9g3URXcBQieGLMG2xijm ZlGPhEV/COKXgCxGIBKTpzFoyrYAqJsOUqdxkFKiejrdJlBwesGYOMdQ8w8uOsioHJkX Pn860tWFj8I8S+DaSCHKU3AnSk7ObiSbLCKZ5QydP0M5Hsh8J5xW2MYY6UJDjjgs5kkv b1mK1hH1GYD+EV18J/hMrSuwa1JmGQ7XL6d3PgX5D6spvBaJ5ArNzhYkFZdPKyb6pMlW Al7FihVXlqNlXYdu7X8CYa2L0WhiEP2bauyKpclxdtgyat7YbqYUkrhKBXlMygOAEMCQ bsvg== 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:cc: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=9y5OvZ5Py/dBHIN4l9cq4GK7tYgMSi3OUj0esOdi4cU=; b=N4cF9JQyWVm0QJY6TMyCDPdYVsru2ixolwnnN0gFhgaxvxgVEvd22hSk4Sh5YnYEEH 9/GzrdOpsWlOvVMOIgnE/5TSWlNa43R1eUnmwVctJ7nyDxgC6pMeazONL044meifjhNd wevuZU5Nl73VROhdjDPvhk3nig4eEl/lhtpUFZrWOZhhHuwZPgt961uUS+wROQZS0j9s I+1KFLYulvMF654/V5npOWJa6GZmgm+Yyyi1//wVAaHOpw6AEayqWL3Sbjvyld7qpnU6 3joNKB9r7cRDUlArEnsJJR3RiCZkSMLtgW73YRjVa56pHiW85DvguoYI359d5tlu4Ceg Q3jw== 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 i137-v6si4209427wme.74.2018.06.18.23.21.46; Mon, 18 Jun 2018 23:21:56 -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 A5148689973; Tue, 19 Jun 2018 09:20:52 +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 49E9168050D for ; Tue, 19 Jun 2018 09:20:40 +0300 (EEST) Received: from localhost (unknown [202.141.160.96]) by mail (Coremail) with SMTP id QMiowPAxWOLjnyhb2s98AA--.17038S3; Tue, 19 Jun 2018 14:17:07 +0800 (CST) From: Shiyou Yin To: ffmpeg-devel@ffmpeg.org Date: Tue, 19 Jun 2018 14:21:10 +0800 Message-Id: <1529389270-32406-1-git-send-email-yinshiyou-hf@loongson.cn> X-Mailer: git-send-email 2.1.0 X-CM-TRANSID: QMiowPAxWOLjnyhb2s98AA--.17038S3 X-Coremail-Antispam: 1UD129KBjvAXoW3KF47Zw1UAF1fCw48JF4xWFg_yoW8CF4Duo W5Kay0g3ZrGw1xCrW8Aw1UAr4jyFW0qrWUZrs3ta13Gr90q3W29rWrCr15GrsxZw4agry3 Aa4qqFy3ZFyDJ3y8n29KB7ZKAUJUUUUU529EdanIXcx71UUUUU7v73VFW2AGmfu7bjvjm3 AaLaJ3UjIYCTnIWjp_UUUY57kC6x804xWl14x267AKxVWUJVW8JwAFc2x0x2IEx4CE42xK 8VAvwI8IcIk0rVWrJVCq3wAFIxvE14AKwVWUJVWUGwA2ocxC64kIII0Yj41l84x0c7CEw4 AK67xGY2AK021l84ACjcxK6xIIjxv20xvE14v26r1I6r4UM28EF7xvwVC0I7IYx2IY6xkF 7I0E14v26r1j6r4UM28EF7xvwVC2z280aVAFwI0_Gr1j6F4UJwA2z4x0Y4vEx4A2jsIEc7 CjxVAFwI0_Cr1j6rxdM2AIxVAIcxkEcVAq07x20xvEncxIr21l5I8CrVACY4xI64kE6c02 F40Ex7xfMcIj6xIIjxv20xvE14v26r1Y6r17McIj6I8E87Iv67AKxVWUJVW8JwAm72CE4I kC6x0Yz7v_Jr0_Gr1lF7xvr2IYc2Ij64vIr41lF7xvr2IYc2Ij64vIr40E4x8a64kEw24l c2xSY4AK67AK6r47MxAIw28IcxkI7VAKI48JMxC20s026xCaFVCjc4AY6r1j6r4UMI8I3I 0E5I8CrVAFwI0_Jr0_Jr4lx2IqxVCjr7xvwVAFwI0_JrI_JrWlx4CE17CEb7AF67AKxVWU XVWUAwCI42IY6xIIjxv20xvE14v26r1j6r1xMIIF0xvE2Ix0cI8IcVCY1x0267AKxVWUJV W8JwCI42IY6xAIw20EY4v20xvaj40_WFyUJVCq3wCI42IY6I8E87Iv67AKxVWUJVW8JwCI 42IY6I8E87Iv6xkF7I0E14v26r1j6r4UYxBIdaVFxhVjvjDU0xZFpf9x07jFg4hUUUUU= X-CM-SenderInfo: p1lq2x5l1r3gtki6z05rqj20fqof0/ Subject: [FFmpeg-devel] [PATCH] avcodec/mips: [loongson] optimize vp8 decoding in vp8dsp. 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 Cc: gxw MIME-Version: 1.0 Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" From: gxw Optimize vp8 decoding with mmi in function: 1. ff_vp8_h_loop_filter8uv_mmi. 2. ff_vp8_v_loop_filter8uv_mmi. 3. ff_vp8_h_loop_filter16_mmi. 4. ff_vp8_v_loop_filter16_mmi. Change-Id: Iba567c7ab4c6a284b5e8ccbc567575448a508350 Signed-off-by: gxw Signed-off-by: Shiyou Yin --- libavcodec/mips/vp8dsp_mmi.c | 408 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 408 insertions(+) diff --git a/libavcodec/mips/vp8dsp_mmi.c b/libavcodec/mips/vp8dsp_mmi.c index f972628..7103d5d 100644 --- a/libavcodec/mips/vp8dsp_mmi.c +++ b/libavcodec/mips/vp8dsp_mmi.c @@ -25,6 +25,251 @@ #include "constants.h" #include "libavutil/mips/mmiutils.h" +#define DECLARE_DOUBLE_1 double db_1 +#define DECLARE_DOUBLE_2 double db_2 +#define DECLARE_UINT32_T uint32_t it_1 +#define RESTRICT_ASM_DOUBLE_1 [db_1]"=&f"(db_1) +#define RESTRICT_ASM_DOUBLE_2 [db_2]"=&f"(db_2) +#define RESTRICT_ASM_UINT32_T [it_1]"=&r"(it_1) + +#define MMI_PSRAB(src, size) \ + "li %[it_1], 0x08 \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "psllh %[db_2], "#src", %[db_1] \n\t" \ + PTR_ADDU "%[it_1], "#size", %[it_1] \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "psrah %[db_2], %[db_2], %[db_1] \n\t" \ + "psrah "#src", "#src", %[db_1] \n\t" \ + PTR_SUBU "%[it_1], %[it_1], "#size" \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "psllh "#src", "#src", %[db_1] \n\t" \ + "xor %[db_1], %[db_1], %[db_1] \n\t" \ + "li %[it_1], 0x00ff00ff \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "punpcklwd %[db_1], %[db_1], %[db_1] \n\t" \ + "and %[db_2], %[db_2], %[db_1] \n\t" \ + "paddb "#src", "#src", %[db_2] \n\t" + +#define MMI_PSRLB(src, size) \ + "li %[it_1], 0x08 \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "psllh %[db_2], "#src", %[db_1] \n\t" \ + PTR_ADDU "%[it_1], "#size", %[it_1] \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "psrlh %[db_2], %[db_2], %[db_1] \n\t" \ + "psrlh "#src", "#src", %[db_1] \n\t" \ + PTR_SUBU "%[it_1], %[it_1], "#size" \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "psllh "#src", "#src", %[db_1] \n\t" \ + "paddb "#src", "#src", %[db_2] \n\t" + +#define MMI_PCMPGTUB(dst, src1, src2) \ + "pcmpeqb %[db_1], "#src1", "#src2" \n\t" \ + "pmaxub %[db_2], "#src1", "#src2" \n\t" \ + "pcmpeqb %[db_2], %[db_2], "#src1" \n\t" \ + "xor "#dst", %[db_2], %[db_1] \n\t" + +#define MMI_BTOH(dst_l, dst_r, src) \ + "xor %[db_1], %[db_1], %[db_1] \n\t" \ + "pcmpgtb %[db_2], %[db_1], "#src" \n\t" \ + "punpcklbh "#dst_r", "#src", %[db_2] \n\t" \ + "punpckhbh "#dst_l", "#src", %[db_2] \n\t" + +#define MMI_TRANSPOSE8x8_UB_UB(src_0, src_1, src_2, src_3, \ + src_4, src_5, src_6, src_7, \ + dst_0, dst_1, dst_2, dst_3, \ + dst_4, dst_5, dst_6, dst_7) \ + "li %[it_1], 0xe4 \n\t" \ + "dmtc1 %[it_1], %[db_1] \n\t" \ + "pshufh %[db_2], "#src_0", %[db_1] \n\t" \ + "punpcklbh "#dst_0", "#src_0", "#src_1" \n\t" \ + "punpckhbh "#dst_1", %[db_2], "#src_1" \n\t" \ + "pshufh %[db_2], "#src_2", %[db_1] \n\t" \ + "punpcklbh "#dst_2", "#src_2", "#src_3" \n\t" \ + "punpckhbh "#dst_3", %[db_2], "#src_3" \n\t" \ + "pshufh %[db_2], "#src_4", %[db_1] \n\t" \ + "punpcklbh "#dst_4", "#src_4", "#src_5" \n\t" \ + "punpckhbh "#dst_5", %[db_2], "#src_5" \n\t" \ + "pshufh %[db_2], "#src_6", %[db_1] \n\t" \ + "punpcklbh "#dst_6", "#src_6", "#src_7" \n\t" \ + "punpckhbh "#dst_7", %[db_2], "#src_7" \n\t" \ + \ + "pshufh %[db_2], "#dst_0", %[db_1] \n\t" \ + "punpcklhw "#dst_0", "#dst_0", "#dst_2" \n\t" \ + "punpckhhw "#dst_2", %[db_2], "#dst_2" \n\t" \ + "pshufh %[db_2], "#dst_1", %[db_1] \n\t" \ + "punpcklhw "#dst_1", "#dst_1", "#dst_3" \n\t" \ + "punpckhhw "#dst_3", %[db_2], "#dst_3" \n\t" \ + "pshufh %[db_2], "#dst_4", %[db_1] \n\t" \ + "punpcklhw "#dst_4", "#dst_4", "#dst_6" \n\t" \ + "punpckhhw "#dst_6", %[db_2], "#dst_6" \n\t" \ + "pshufh %[db_2], "#dst_5", %[db_1] \n\t" \ + "punpcklhw "#dst_5", "#dst_5", "#dst_7" \n\t" \ + "punpckhhw "#dst_7", %[db_2], "#dst_7" \n\t" \ + \ + "pshufh %[db_2], "#dst_0", %[db_1] \n\t" \ + "punpcklwd "#dst_0", "#dst_0", "#dst_4" \n\t" \ + "punpckhwd "#dst_4", %[db_2], "#dst_4" \n\t" \ + "pshufh %[db_2], "#dst_1", %[db_1] \n\t" \ + "punpcklwd "#dst_1", "#dst_1", "#dst_5" \n\t" \ + "punpckhwd "#dst_5", %[db_2], "#dst_5" \n\t" \ + "pshufh %[db_2], "#dst_2", %[db_1] \n\t" \ + "punpcklwd "#dst_2", "#dst_2", "#dst_6" \n\t" \ + "punpckhwd "#dst_6", %[db_2], "#dst_6" \n\t" \ + "pshufh %[db_2], "#dst_3", %[db_1] \n\t" \ + "punpcklwd "#dst_3", "#dst_3", "#dst_7" \n\t" \ + "punpckhwd "#dst_7", %[db_2], "#dst_7" \n\t" \ + \ + "pshufh %[db_2], "#dst_1", %[db_1] \n\t" \ + "pshufh "#dst_1", "#dst_4", %[db_1] \n\t" \ + "pshufh "#dst_4", %[db_2], %[db_1] \n\t" \ + "pshufh %[db_2], "#dst_3", %[db_1] \n\t" \ + "pshufh "#dst_3", "#dst_6", %[db_1] \n\t" \ + "pshufh "#dst_6", %[db_2], %[db_1] \n\t" + +#define MMI_VP8_LOOP_FILTER \ + /* Calculation of hev */ \ + "dmtc1 %[thresh], %[ftmp3] \n\t" \ + "punpcklbh %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "punpcklhw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "punpcklwd %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "pasubub %[ftmp0], %[p1], %[p0] \n\t" \ + "pasubub %[ftmp1], %[q1], %[q0] \n\t" \ + "pmaxub %[ftmp0], %[ftmp0], %[ftmp1] \n\t" \ + MMI_PCMPGTUB(%[hev], %[ftmp0], %[ftmp3]) \ + /* Calculation of mask */ \ + "pasubub %[ftmp1], %[p0], %[q0] \n\t" \ + "paddusb %[ftmp1], %[ftmp1], %[ftmp1] \n\t" \ + "pasubub %[ftmp2], %[p1], %[q1] \n\t" \ + "li %[tmp0], 0x01 \n\t" \ + MMI_PSRLB(%[ftmp2], %[tmp0]) \ + "paddusb %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \ + "dmtc1 %[e], %[ftmp3] \n\t" \ + "punpcklbh %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "punpcklhw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "punpcklwd %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + MMI_PCMPGTUB(%[mask], %[ftmp1], %[ftmp3]) \ + "pmaxub %[mask], %[mask], %[ftmp0] \n\t" \ + "pasubub %[ftmp1], %[p3], %[p2] \n\t" \ + "pasubub %[ftmp2], %[p2], %[p1] \n\t" \ + "pmaxub %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \ + "pmaxub %[mask], %[mask], %[ftmp1] \n\t" \ + "pasubub %[ftmp1], %[q3], %[q2] \n\t" \ + "pasubub %[ftmp2], %[q2], %[q1] \n\t" \ + "pmaxub %[ftmp1], %[ftmp1], %[ftmp2] \n\t" \ + "pmaxub %[mask], %[mask], %[ftmp1] \n\t" \ + "dmtc1 %[i], %[ftmp3] \n\t" \ + "punpcklbh %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "punpcklhw %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + "punpcklwd %[ftmp3], %[ftmp3], %[ftmp3] \n\t" \ + MMI_PCMPGTUB(%[mask], %[mask], %[ftmp3]) \ + "li %[tmp0], 0xffffffff \n\t" \ + "dmtc1 %[tmp0], %[ftmp3] \n\t" \ + "xor %[mask], %[mask], %[ftmp3] \n\t" \ + /* VP8_MBFILTER */ \ + "li %[tmp0], 0x80808080 \n\t" \ + "dmtc1 %[tmp0], %[ftmp7] \n\t" \ + "punpcklwd %[ftmp7], %[ftmp7], %[ftmp7] \n\t" \ + "xor %[p2], %[p2], %[ftmp7] \n\t" \ + "xor %[p1], %[p1], %[ftmp7] \n\t" \ + "xor %[p0], %[p0], %[ftmp7] \n\t" \ + "xor %[q0], %[q0], %[ftmp7] \n\t" \ + "xor %[q1], %[q1], %[ftmp7] \n\t" \ + "xor %[q2], %[q2], %[ftmp7] \n\t" \ + "psubsb %[ftmp4], %[p1], %[q1] \n\t" \ + "psubb %[ftmp5], %[q0], %[p0] \n\t" \ + "li %[tmp0], 0x00030003 \n\t" \ + "dmtc1 %[tmp0], %[ftmp6] \n\t" \ + "punpcklwd %[ftmp6], %[ftmp6], %[ftmp6] \n\t" \ + MMI_BTOH(%[ftmp1], %[ftmp0], %[ftmp5]) \ + MMI_BTOH(%[ftmp3], %[ftmp2], %[ftmp4]) \ + /* Right part */ \ + "pmullh %[ftmp0], %[ftmp0], %[ftmp6] \n\t" \ + "paddh %[ftmp0], %[ftmp2], %[ftmp0] \n\t" \ + /* Left part */ \ + "pmullh %[ftmp1], %[ftmp1], %[ftmp6] \n\t" \ + "paddh %[ftmp1], %[ftmp3], %[ftmp1] \n\t" \ + /* Combine left and right part */ \ + "packsshb %[ftmp1], %[ftmp0], %[ftmp1] \n\t" \ + "and %[ftmp1], %[ftmp1], %[mask] \n\t" \ + "and %[ftmp2], %[ftmp1], %[hev] \n\t" \ + "li %[tmp0], 0x04040404 \n\t" \ + "dmtc1 %[tmp0], %[ftmp0] \n\t" \ + "punpcklwd %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \ + "paddsb %[ftmp3], %[ftmp2], %[ftmp0] \n\t" \ + "li %[tmp0], 0x03 \n\t" \ + MMI_PSRAB(%[ftmp3], %[tmp0]) \ + "li %[tmp0], 0x03030303 \n\t" \ + "dmtc1 %[tmp0], %[ftmp0] \n\t" \ + "punpcklwd %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \ + "paddsb %[ftmp4], %[ftmp2], %[ftmp0] \n\t" \ + "li %[tmp0], 0x03 \n\t" \ + MMI_PSRAB(%[ftmp4], %[tmp0]) \ + "psubsb %[q0], %[q0], %[ftmp3] \n\t" \ + "paddsb %[p0], %[p0], %[ftmp4] \n\t" \ + /* filt_val &= ~hev */ \ + "pcmpeqw %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \ + "xor %[hev], %[hev], %[ftmp0] \n\t" \ + "and %[ftmp1], %[ftmp1], %[hev] \n\t" \ + MMI_BTOH(%[ftmp5], %[ftmp6], %[ftmp1]) \ + "li %[tmp0], 0x07 \n\t" \ + "dmtc1 %[tmp0], %[ftmp2] \n\t" \ + "li %[tmp0], 0x001b001b \n\t" \ + "dmtc1 %[tmp0], %[ftmp1] \n\t" \ + "punpcklwd %[ftmp1], %[ftmp1], %[ftmp1] \n\t" \ + "li %[tmp0], 0x003f003f \n\t" \ + "dmtc1 %[tmp0], %[ftmp0] \n\t" \ + "punpcklwd %[ftmp0], %[ftmp0], %[ftmp0] \n\t" \ + /* Right part */ \ + "pmullh %[ftmp3], %[ftmp6], %[ftmp1] \n\t" \ + "paddh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \ + "psrah %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \ + /* Left part */ \ + "pmullh %[ftmp4], %[ftmp5], %[ftmp1] \n\t" \ + "paddh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \ + "psrah %[ftmp4], %[ftmp4], %[ftmp2] \n\t" \ + /* Combine left and right part */ \ + "packsshb %[ftmp4], %[ftmp3], %[ftmp4] \n\t" \ + "psubsb %[q0], %[q0], %[ftmp4] \n\t" \ + "xor %[q0], %[q0], %[ftmp7] \n\t" \ + "paddsb %[p0], %[p0], %[ftmp4] \n\t" \ + "xor %[p0], %[p0], %[ftmp7] \n\t" \ + "li %[tmp0], 0x00120012 \n\t" \ + "dmtc1 %[tmp0], %[ftmp1] \n\t" \ + "punpcklwd %[ftmp1], %[ftmp1], %[ftmp1] \n\t" \ + /* Right part */ \ + "pmullh %[ftmp3], %[ftmp6], %[ftmp1] \n\t" \ + "paddh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \ + "psrah %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \ + /* Left part */ \ + "pmullh %[ftmp4], %[ftmp5], %[ftmp1] \n\t" \ + "paddh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \ + "psrah %[ftmp4], %[ftmp4], %[ftmp2] \n\t" \ + /* Combine left and right part */ \ + "packsshb %[ftmp4], %[ftmp3], %[ftmp4] \n\t" \ + "psubsb %[q1], %[q1], %[ftmp4] \n\t" \ + "xor %[q1], %[q1], %[ftmp7] \n\t" \ + "paddsb %[p1], %[p1], %[ftmp4] \n\t" \ + "xor %[p1], %[p1], %[ftmp7] \n\t" \ + "li %[tmp0], 0x03 \n\t" \ + "dmtc1 %[tmp0], %[ftmp1] \n\t" \ + /* Right part */ \ + "psllh %[ftmp3], %[ftmp6], %[ftmp1] \n\t" \ + "paddh %[ftmp3], %[ftmp3], %[ftmp6] \n\t" \ + "paddh %[ftmp3], %[ftmp3], %[ftmp0] \n\t" \ + "psrah %[ftmp3], %[ftmp3], %[ftmp2] \n\t" \ + /* Left part */ \ + "psllh %[ftmp4], %[ftmp5], %[ftmp1] \n\t" \ + "paddh %[ftmp4], %[ftmp4], %[ftmp5] \n\t" \ + "paddh %[ftmp4], %[ftmp4], %[ftmp0] \n\t" \ + "psrah %[ftmp4], %[ftmp4], %[ftmp2] \n\t" \ + /* Combine left and right part */ \ + "packsshb %[ftmp4], %[ftmp3], %[ftmp4] \n\t" \ + "psubsb %[q2], %[q2], %[ftmp4] \n\t" \ + "xor %[q2], %[q2], %[ftmp7] \n\t" \ + "paddsb %[p2], %[p2], %[ftmp4] \n\t" \ + "xor %[p2], %[p2], %[ftmp7] \n\t" + #define PUT_VP8_EPEL4_H6_MMI(src, dst) \ MMI_ULWC1(%[ftmp1], src, 0x00) \ "punpcklbh %[ftmp2], %[ftmp1], %[ftmp0] \n\t" \ @@ -621,6 +866,7 @@ static av_always_inline int vp8_normal_limit(uint8_t *p, ptrdiff_t stride, static av_always_inline void vp8_v_loop_filter8_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh) { +#if 0 int i; for (i = 0; i < 8; i++) @@ -630,6 +876,73 @@ static av_always_inline void vp8_v_loop_filter8_mmi(uint8_t *dst, else filter_mbedge(dst + i * 1, stride); } +#else + double ftmp[18]; + uint32_t tmp[1]; + DECLARE_DOUBLE_1; + DECLARE_DOUBLE_2; + DECLARE_UINT32_T; + __asm__ volatile( + /* Get data from dst */ + "gsldlc1 %[q0], 0x07(%[dst]) \n\t" + "gsldrc1 %[q0], 0x00(%[dst]) \n\t" + PTR_SUBU "%[tmp0], %[dst], %[stride] \n\t" + "gsldlc1 %[p0], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[p0], 0x00(%[tmp0]) \n\t" + PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[p1], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[p1], 0x00(%[tmp0]) \n\t" + PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[p2], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[p2], 0x00(%[tmp0]) \n\t" + PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[p3], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[p3], 0x00(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[dst], %[stride] \n\t" + "gsldlc1 %[q1], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[q1], 0x00(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[q2], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[q2], 0x00(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[q3], 0x07(%[tmp0]) \n\t" + "gsldrc1 %[q3], 0x00(%[tmp0]) \n\t" + MMI_VP8_LOOP_FILTER + /* Move to dst */ + "gssdlc1 %[q0], 0x07(%[dst]) \n\t" + "gssdrc1 %[q0], 0x00(%[dst]) \n\t" + PTR_SUBU "%[tmp0], %[dst], %[stride] \n\t" + "gssdlc1 %[p0], 0x07(%[tmp0]) \n\t" + "gssdrc1 %[p0], 0x00(%[tmp0]) \n\t" + PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t" + "gssdlc1 %[p1], 0x07(%[tmp0]) \n\t" + "gssdrc1 %[p1], 0x00(%[tmp0]) \n\t" + PTR_SUBU "%[tmp0], %[tmp0], %[stride] \n\t" + "gssdlc1 %[p2], 0x07(%[tmp0]) \n\t" + "gssdrc1 %[p2], 0x00(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[dst], %[stride] \n\t" + "gssdlc1 %[q1], 0x07(%[tmp0]) \n\t" + "gssdrc1 %[q1], 0x00(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gssdlc1 %[q2], 0x07(%[tmp0]) \n\t" + "gssdrc1 %[q2], 0x00(%[tmp0]) \n\t" + : [p3]"=&f"(ftmp[0]), [p2]"=&f"(ftmp[1]), + [p1]"=&f"(ftmp[2]), [p0]"=&f"(ftmp[3]), + [q0]"=&f"(ftmp[4]), [q1]"=&f"(ftmp[5]), + [q2]"=&f"(ftmp[6]), [q3]"=&f"(ftmp[7]), + [ftmp0]"=&f"(ftmp[8]), [ftmp1]"=&f"(ftmp[9]), + [ftmp2]"=&f"(ftmp[10]), [ftmp3]"=&f"(ftmp[11]), + [hev]"=&f"(ftmp[12]), [mask]"=&f"(ftmp[13]), + [ftmp4]"=&f"(ftmp[14]), [ftmp5]"=&f"(ftmp[15]), + [ftmp6]"=&f"(ftmp[16]), [ftmp7]"=&f"(ftmp[17]), + [dst]"+&r"(dst), [tmp0]"=&r"(tmp[0]), + RESTRICT_ASM_DOUBLE_1, RESTRICT_ASM_DOUBLE_2, + RESTRICT_ASM_UINT32_T + : [e]"r"((mips_reg)flim_E), [thresh]"r"((mips_reg)hev_thresh), + [i]"r"((mips_reg)flim_I), [stride]"r"((mips_reg)stride) + : "memory" + ); +#endif } static av_always_inline void vp8_v_loop_filter8_inner_mmi(uint8_t *dst, @@ -650,6 +963,7 @@ static av_always_inline void vp8_v_loop_filter8_inner_mmi(uint8_t *dst, static av_always_inline void vp8_h_loop_filter8_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh) { +#if 0 int i; for (i = 0; i < 8; i++) @@ -659,6 +973,89 @@ static av_always_inline void vp8_h_loop_filter8_mmi(uint8_t *dst, else filter_mbedge(dst + i * stride, 1); } +#else + double ftmp[18]; + uint32_t tmp[1]; + DECLARE_DOUBLE_1; + DECLARE_DOUBLE_2; + DECLARE_UINT32_T; + __asm__ volatile( + /* Get data from dst */ + "gsldlc1 %[p3], 0x03(%[dst]) \n\t" + "gsldrc1 %[p3], -0x04(%[dst]) \n\t" + PTR_ADDU "%[tmp0], %[dst], %[stride] \n\t" + "gsldlc1 %[p2], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[p2], -0x04(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[p1], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[p1], -0x04(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[p0], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[p0], -0x04(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[q0], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[q0], -0x04(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[q1], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[q1], -0x04(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[q2], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[q2], -0x04(%[tmp0]) \n\t" + PTR_ADDU "%[tmp0], %[tmp0], %[stride] \n\t" + "gsldlc1 %[q3], 0x03(%[tmp0]) \n\t" + "gsldrc1 %[q3], -0x04(%[tmp0]) \n\t" + /* Matrix transpose */ + MMI_TRANSPOSE8x8_UB_UB(%[p3], %[p2], %[p1], %[p0], + %[q0], %[q1], %[q2], %[q3], + %[p3], %[p2], %[p1], %[p0], + %[q0], %[q1], %[q2], %[q3]) + MMI_VP8_LOOP_FILTER + /* Matrix transpose */ + MMI_TRANSPOSE8x8_UB_UB(%[p3], %[p2], %[p1], %[p0], + %[q0], %[q1], %[q2], %[q3], + %[p3], %[p2], %[p1], %[p0], + %[q0], %[q1], %[q2], %[q3]) + /* Move to dst */ + "gssdlc1 %[p3], 0x03(%[dst]) \n\t" + "gssdrc1 %[p3], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[p2], 0x03(%[dst]) \n\t" + "gssdrc1 %[p2], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[p1], 0x03(%[dst]) \n\t" + "gssdrc1 %[p1], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[p0], 0x03(%[dst]) \n\t" + "gssdrc1 %[p0], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[q0], 0x03(%[dst]) \n\t" + "gssdrc1 %[q0], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[q1], 0x03(%[dst]) \n\t" + "gssdrc1 %[q1], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[q2], 0x03(%[dst]) \n\t" + "gssdrc1 %[q2], -0x04(%[dst]) \n\t" + PTR_ADDU "%[dst], %[dst], %[stride] \n\t" + "gssdlc1 %[q3], 0x03(%[dst]) \n\t" + "gssdrc1 %[q3], -0x04(%[dst]) \n\t" + : [p3]"=&f"(ftmp[0]), [p2]"=&f"(ftmp[1]), + [p1]"=&f"(ftmp[2]), [p0]"=&f"(ftmp[3]), + [q0]"=&f"(ftmp[4]), [q1]"=&f"(ftmp[5]), + [q2]"=&f"(ftmp[6]), [q3]"=&f"(ftmp[7]), + [ftmp0]"=&f"(ftmp[8]), [ftmp1]"=&f"(ftmp[9]), + [ftmp2]"=&f"(ftmp[10]), [ftmp3]"=&f"(ftmp[11]), + [hev]"=&f"(ftmp[12]), [mask]"=&f"(ftmp[13]), + [ftmp4]"=&f"(ftmp[14]), [ftmp5]"=&f"(ftmp[15]), + [ftmp6]"=&f"(ftmp[16]), [ftmp7]"=&f"(ftmp[17]), + [dst]"+&r"(dst), [tmp0]"=&r"(tmp[0]), + RESTRICT_ASM_DOUBLE_1, RESTRICT_ASM_DOUBLE_2, + RESTRICT_ASM_UINT32_T + : [e]"r"((mips_reg)flim_E), [thresh]"r"((mips_reg)hev_thresh), + [i]"r"((mips_reg)flim_I), [stride]"r"((mips_reg)stride) + : "memory" + ); +#endif } static av_always_inline void vp8_h_loop_filter8_inner_mmi(uint8_t *dst, @@ -1083,6 +1480,7 @@ void ff_vp8_idct_dc_add4uv_mmi(uint8_t *dst, int16_t block[4][16], void ff_vp8_v_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh) { +#if 0 int i; for (i = 0; i < 16; i++) @@ -1092,11 +1490,16 @@ void ff_vp8_v_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E, else filter_mbedge(dst + i * 1, stride); } +#else + vp8_v_loop_filter8_mmi(dst, stride, flim_E, flim_I, hev_thresh); + vp8_v_loop_filter8_mmi(dst + 8, stride, flim_E, flim_I, hev_thresh); +#endif } void ff_vp8_h_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E, int flim_I, int hev_thresh) { +#if 0 int i; for (i = 0; i < 16; i++) @@ -1106,6 +1509,11 @@ void ff_vp8_h_loop_filter16_mmi(uint8_t *dst, ptrdiff_t stride, int flim_E, else filter_mbedge(dst + i * stride, 1); } +#else + vp8_h_loop_filter8_mmi(dst, stride, flim_E, flim_I, hev_thresh); + vp8_h_loop_filter8_mmi(dst + 8 * stride, stride, flim_E, flim_I, + hev_thresh); +#endif } void ff_vp8_v_loop_filter8uv_mmi(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,