From patchwork Wed Oct 11 08:47:21 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: kaustubh.raste@imgtec.com X-Patchwork-Id: 5527 Delivered-To: ffmpegpatchwork@gmail.com Received: by 10.2.161.90 with SMTP id m26csp461291jah; Wed, 11 Oct 2017 01:46:03 -0700 (PDT) X-Google-Smtp-Source: AOwi7QAfmab4uRt7Uc97vemnoBS2o7Sm/nKvQEAKXO8ON+uikTRJddpkGBlaugMZdAx2caDp0PCC X-Received: by 10.223.157.136 with SMTP id p8mr15171216wre.28.1507711563433; Wed, 11 Oct 2017 01:46:03 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1507711563; cv=none; d=google.com; s=arc-20160816; b=X7HLMHx8JIjqfPRcSAR6Sj3+jHBgNO5av+/mkVbc72V/6oB6Dj6C6ySKbUIRg6kdAW 5n8Cbt3/NRoqwFIRNxBNjmyQnOR0VxTMGbaY+w/bO/Srm9zH9R8HuBRxpTXfXqJW+XKo OxglyVMgD8cPywPzi+5f1aVjI0BKh04g6PvAFmJr4KJC29r1tYinR4g+uBDSg/NZ7FqK oKH+XnkOhQzipHLW06MxqmCHG3xtep+DB6E4O3kavat3kAK06c6beX7yBXLdIzrVqSBX G98EBFahmBQj0218A6TvpeGQgbXwqPzc6BM9glMWRuLUFowjV/5r3AC+FIHFEpcZL9Lm OcLA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:cc:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:mime-version:message-id:date:to:from :delivered-to:arc-authentication-results; bh=5vgtEnAaWIsR6PfvoNtZA4QWEd6eZL7GmsmWpnhvdlU=; b=yjy0ybTCMu32Z5+o7aZ/bdh05y/wLuIQK3thhuKTX3YSmzTNlyxvWgCdaZM0r/V6r1 81xdnLbrJIBdhNU29ax98BPL5BevV/nfwFHGOhXE8TnomJyyS7zILrLXWkZzPxH65fkw VDKHKQmjOXYRma2rBVcq6TTPUdUjmorxNZOnU4l5aohNUtBdcyXWOzPJY4ZBweILv6d/ Vcb0l+J4Sa9HvE815f/0xoQvoC2XYLCdj+SAxQKMRyncrAIp5sYuKdTc5BTL/CP45p+v xzp51/1jb6PR41loZELILfJustqTWSKYxHPLpkMiZk9mSOM+grqI1W0ZySerzjTXSwIv dZOQ== 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 d6si6220210wrg.511.2017.10.11.01.46.02; Wed, 11 Oct 2017 01:46:03 -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 25279689CAD; Wed, 11 Oct 2017 11:45:59 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mailapp01.imgtec.com (mailapp01.imgtec.com [195.59.15.196]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id EFC156883AE for ; Wed, 11 Oct 2017 11:45:57 +0300 (EEST) Received: from hhmail02.hh.imgtec.org (unknown [10.100.10.20]) by Forcepoint Email with ESMTPS id 1E7333C869C8E for ; Wed, 11 Oct 2017 09:45:56 +0100 (IST) Received: from pudesk204.pu.imgtec.org (192.168.91.13) by hhmail02.hh.imgtec.org (10.100.10.20) with Microsoft SMTP Server (TLS) id 14.3.361.1; Wed, 11 Oct 2017 09:45:57 +0100 From: To: Date: Wed, 11 Oct 2017 14:17:21 +0530 Message-ID: <1507711641-32581-1-git-send-email-kaustubh.raste@imgtec.com> X-Mailer: git-send-email 1.7.9.5 MIME-Version: 1.0 X-Originating-IP: [192.168.91.13] Subject: [FFmpeg-devel] [PATCH] avcodec/mips: Improve avc put mc 12, 32 and 22 msa functions 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: Kaustubh Raste Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" From: Kaustubh Raste Remove loops and unroll as block sizes are known. Removed unused functions. Signed-off-by: Kaustubh Raste --- libavcodec/mips/h264qpel_msa.c | 1013 +++++++++++++++++++++++++++++----------- 1 file changed, 747 insertions(+), 266 deletions(-) diff --git a/libavcodec/mips/h264qpel_msa.c b/libavcodec/mips/h264qpel_msa.c index a22a482..f11fce8 100644 --- a/libavcodec/mips/h264qpel_msa.c +++ b/libavcodec/mips/h264qpel_msa.c @@ -171,241 +171,6 @@ static const uint8_t luma_mask_arr[16 * 8] = { out0_m; \ } ) -static void avc_luma_mid_4w_msa(const uint8_t *src, int32_t src_stride, - uint8_t *dst, int32_t dst_stride, - int32_t height) -{ - uint32_t loop_cnt; - v16i8 src0, src1, src2, src3, src4; - v16i8 mask0, mask1, mask2; - v8i16 hz_out0, hz_out1, hz_out2, hz_out3; - v8i16 hz_out4, hz_out5, hz_out6, hz_out7, hz_out8; - v8i16 dst0, dst1, dst2, dst3; - - LD_SB3(&luma_mask_arr[48], 16, mask0, mask1, mask2); - LD_SB5(src, src_stride, src0, src1, src2, src3, src4); - src += (5 * src_stride); - - XORI_B5_128_SB(src0, src1, src2, src3, src4); - - hz_out0 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src0, src1, - mask0, mask1, mask2); - hz_out2 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src2, src3, - mask0, mask1, mask2); - - PCKOD_D2_SH(hz_out0, hz_out0, hz_out2, hz_out2, hz_out1, hz_out3); - - hz_out4 = AVC_HORZ_FILTER_SH(src4, src4, mask0, mask1, mask2); - - for (loop_cnt = (height >> 2); loop_cnt--;) { - LD_SB4(src, src_stride, src0, src1, src2, src3); - src += (4 * src_stride); - - XORI_B4_128_SB(src0, src1, src2, src3); - - hz_out5 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src0, src1, - mask0, mask1, - mask2); - hz_out7 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src2, src3, - mask0, mask1, - mask2); - - PCKOD_D2_SH(hz_out5, hz_out5, hz_out7, hz_out7, hz_out6, hz_out8); - - dst0 = AVC_CALC_DPADD_H_6PIX_2COEFF_R_SH(hz_out0, hz_out1, hz_out2, - hz_out3, hz_out4, hz_out5); - dst1 = AVC_CALC_DPADD_H_6PIX_2COEFF_R_SH(hz_out1, hz_out2, hz_out3, - hz_out4, hz_out5, hz_out6); - dst2 = AVC_CALC_DPADD_H_6PIX_2COEFF_R_SH(hz_out2, hz_out3, hz_out4, - hz_out5, hz_out6, hz_out7); - dst3 = AVC_CALC_DPADD_H_6PIX_2COEFF_R_SH(hz_out3, hz_out4, hz_out5, - hz_out6, hz_out7, hz_out8); - - PCKEV_B2_SB(dst1, dst0, dst3, dst2, src0, src1); - XORI_B2_128_SB(src0, src1); - - ST4x4_UB(src0, src1, 0, 2, 0, 2, dst, dst_stride); - - dst += (4 * dst_stride); - - hz_out0 = hz_out4; - hz_out1 = hz_out5; - hz_out2 = hz_out6; - hz_out3 = hz_out7; - hz_out4 = hz_out8; - } -} - -static void avc_luma_mid_8w_msa(const uint8_t *src, int32_t src_stride, - uint8_t *dst, int32_t dst_stride, - int32_t height) -{ - uint32_t loop_cnt; - v16i8 src0, src1, src2, src3, src4; - v16i8 mask0, mask1, mask2; - v8i16 hz_out0, hz_out1, hz_out2, hz_out3; - v8i16 hz_out4, hz_out5, hz_out6, hz_out7, hz_out8; - v8i16 dst0, dst1, dst2, dst3; - v16u8 out0, out1; - - LD_SB3(&luma_mask_arr[0], 16, mask0, mask1, mask2); - - LD_SB5(src, src_stride, src0, src1, src2, src3, src4); - XORI_B5_128_SB(src0, src1, src2, src3, src4); - src += (5 * src_stride); - - hz_out0 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); - hz_out1 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); - hz_out2 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); - hz_out3 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); - hz_out4 = AVC_HORZ_FILTER_SH(src4, src4, mask0, mask1, mask2); - - for (loop_cnt = (height >> 2); loop_cnt--;) { - LD_SB4(src, src_stride, src0, src1, src2, src3); - XORI_B4_128_SB(src0, src1, src2, src3); - src += (4 * src_stride); - - hz_out5 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); - hz_out6 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); - hz_out7 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); - hz_out8 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); - dst0 = AVC_CALC_DPADD_H_6PIX_2COEFF_SH(hz_out0, hz_out1, hz_out2, - hz_out3, hz_out4, hz_out5); - dst1 = AVC_CALC_DPADD_H_6PIX_2COEFF_SH(hz_out1, hz_out2, hz_out3, - hz_out4, hz_out5, hz_out6); - dst2 = AVC_CALC_DPADD_H_6PIX_2COEFF_SH(hz_out2, hz_out3, hz_out4, - hz_out5, hz_out6, hz_out7); - dst3 = AVC_CALC_DPADD_H_6PIX_2COEFF_SH(hz_out3, hz_out4, hz_out5, - hz_out6, hz_out7, hz_out8); - out0 = PCKEV_XORI128_UB(dst0, dst1); - out1 = PCKEV_XORI128_UB(dst2, dst3); - ST8x4_UB(out0, out1, dst, dst_stride); - - dst += (4 * dst_stride); - hz_out3 = hz_out7; - hz_out1 = hz_out5; - hz_out5 = hz_out4; - hz_out4 = hz_out8; - hz_out2 = hz_out6; - hz_out0 = hz_out5; - } -} - -static void avc_luma_mid_16w_msa(const uint8_t *src, int32_t src_stride, - uint8_t *dst, int32_t dst_stride, - int32_t height) -{ - uint32_t multiple8_cnt; - - for (multiple8_cnt = 2; multiple8_cnt--;) { - avc_luma_mid_8w_msa(src, src_stride, dst, dst_stride, height); - src += 8; - dst += 8; - } -} - -static void avc_luma_midh_qrt_4w_msa(const uint8_t *src, int32_t src_stride, - uint8_t *dst, int32_t dst_stride, - int32_t height, uint8_t horiz_offset) -{ - uint32_t row; - v16i8 src0, src1, src2, src3, src4, src5, src6; - v8i16 vt_res0, vt_res1, vt_res2, vt_res3; - v4i32 hz_res0, hz_res1; - v8i16 dst0, dst1; - v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5; - v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; - v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; - v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; - v8i16 minus5h = __msa_ldi_h(-5); - v8i16 plus20h = __msa_ldi_h(20); - v8i16 zeros = { 0 }; - v16u8 out; - - LD_SB5(src, src_stride, src0, src1, src2, src3, src4); - src += (5 * src_stride); - XORI_B5_128_SB(src0, src1, src2, src3, src4); - - for (row = (height >> 1); row--;) { - LD_SB2(src, src_stride, src5, src6); - src += (2 * src_stride); - - XORI_B2_128_SB(src5, src6); - AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src0, src1, src2, src3, src4, src5, - vt_res0, vt_res1); - AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src1, src2, src3, src4, src5, src6, - vt_res2, vt_res3); - VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, - mask0, mask1, mask2, shf_vec0, shf_vec1, shf_vec2); - VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, - mask0, mask1, mask2, shf_vec3, shf_vec4, shf_vec5); - hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); - DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); - hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); - DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); - - SRARI_W2_SW(hz_res0, hz_res1, 10); - SAT_SW2_SW(hz_res0, hz_res1, 7); - - dst0 = __msa_srari_h(shf_vec2, 5); - dst1 = __msa_srari_h(shf_vec5, 5); - - SAT_SH2_SH(dst0, dst1, 7); - - if (horiz_offset) { - dst0 = __msa_ilvod_h(zeros, dst0); - dst1 = __msa_ilvod_h(zeros, dst1); - } else { - ILVEV_H2_SH(dst0, zeros, dst1, zeros, dst0, dst1); - } - - hz_res0 = __msa_aver_s_w(hz_res0, (v4i32) dst0); - hz_res1 = __msa_aver_s_w(hz_res1, (v4i32) dst1); - dst0 = __msa_pckev_h((v8i16) hz_res1, (v8i16) hz_res0); - - out = PCKEV_XORI128_UB(dst0, dst0); - ST4x2_UB(out, dst, dst_stride); - - dst += (2 * dst_stride); - - src0 = src2; - src1 = src3; - src2 = src4; - src3 = src5; - src4 = src6; - } -} - -static void avc_luma_midh_qrt_8w_msa(const uint8_t *src, int32_t src_stride, - uint8_t *dst, int32_t dst_stride, - int32_t height, uint8_t horiz_offset) -{ - uint32_t multiple8_cnt; - - for (multiple8_cnt = 2; multiple8_cnt--;) { - avc_luma_midh_qrt_4w_msa(src, src_stride, dst, dst_stride, height, - horiz_offset); - - src += 4; - dst += 4; - } -} - -static void avc_luma_midh_qrt_16w_msa(const uint8_t *src, int32_t src_stride, - uint8_t *dst, int32_t dst_stride, - int32_t height, uint8_t horiz_offset) -{ - uint32_t multiple8_cnt; - - for (multiple8_cnt = 4; multiple8_cnt--;) { - avc_luma_midh_qrt_4w_msa(src, src_stride, dst, dst_stride, height, - horiz_offset); - - src += 4; - dst += 4; - } -} - static void avc_luma_hv_qrt_4w_msa(const uint8_t *src_x, const uint8_t *src_y, int32_t src_stride, uint8_t *dst, int32_t dst_stride, int32_t height) @@ -3634,57 +3399,773 @@ void ff_put_h264_qpel4_mc02_msa(uint8_t *dst, const uint8_t *src, void ff_put_h264_qpel16_mc12_msa(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) { - avc_luma_midh_qrt_16w_msa(src - (2 * stride) - 2, - stride, dst, stride, 16, 0); + uint32_t row; + v16u8 out; + v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10; + v16i8 src11; + v8i16 vt_res0, vt_res1, vt_res2, vt_res3, dst0, dst1, dst2, dst3, mask3; + v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5, shf_vec6; + v8i16 shf_vec7, shf_vec8, shf_vec9, shf_vec10, shf_vec11, mask4, mask5; + v4i32 hz_res0, hz_res1, hz_res2, hz_res3; + v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; + v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; + v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; + v8i16 minus5h = __msa_ldi_h(-5); + v8i16 plus20h = __msa_ldi_h(20); + + mask3 = mask0 + 4; + mask4 = mask1 + 4; + mask5 = mask2 + 4; + + src -= ((2 * stride) + 2); + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + LD_SB5(src + 8, stride, src7, src8, src9, src10, src11); + src += (5 * stride); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + XORI_B5_128_SB(src7, src8, src9, src10, src11); + + for (row = 16; row--;) { + LD_SB2(src, 8, src5, src6); + src += stride; + XORI_B2_128_SB(src5, src6); + + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src0, src1, src2, src3, src4, src5, + vt_res0, vt_res1); + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src7, src8, src9, src10, src11, src6, + vt_res2, vt_res3); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec2); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec5); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask3, + mask4, mask5, shf_vec6, shf_vec7, shf_vec8); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask3, + mask4, mask5, shf_vec9, shf_vec10, shf_vec11); + hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); + hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); + hz_res2 = __msa_hadd_s_w(shf_vec6, shf_vec6); + hz_res3 = __msa_hadd_s_w(shf_vec9, shf_vec9); + DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); + DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); + DPADD_SH2_SW(shf_vec7, shf_vec8, minus5h, plus20h, hz_res2, hz_res2); + DPADD_SH2_SW(shf_vec10, shf_vec11, minus5h, plus20h, hz_res3, hz_res3); + SRARI_W4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 10); + SAT_SW4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 7); + dst0 = __msa_srari_h(shf_vec2, 5); + dst1 = __msa_srari_h(shf_vec5, 5); + dst2 = __msa_srari_h(shf_vec8, 5); + dst3 = __msa_srari_h(shf_vec11, 5); + SAT_SH4_SH(dst0, dst1, dst2, dst3, 7); + PCKEV_H2_SH(dst2, dst0, dst3, dst1, dst0, dst1); + PCKEV_H2_SH(hz_res2, hz_res0, hz_res3, hz_res1, dst2, dst3); + dst0 = __msa_aver_s_h(dst2, dst0); + dst1 = __msa_aver_s_h(dst3, dst1); + out = PCKEV_XORI128_UB(dst0, dst1); + ST_UB(out, dst); + dst += stride; + + src0 = src1; + src1 = src2; + src2 = src3; + src3 = src4; + src4 = src5; + src7 = src8; + src8 = src9; + src9 = src10; + src10 = src11; + src11 = src6; + } } void ff_put_h264_qpel16_mc32_msa(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) { - avc_luma_midh_qrt_16w_msa(src - (2 * stride) - 2, - stride, dst, stride, 16, 1); -} + uint32_t row; + v16u8 out; + v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10; + v16i8 src11; + v8i16 vt_res0, vt_res1, vt_res2, vt_res3, dst0, dst1, dst2, dst3, mask3; + v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5, shf_vec6; + v8i16 shf_vec7, shf_vec8, shf_vec9, shf_vec10, shf_vec11, mask4, mask5; + v4i32 hz_res0, hz_res1, hz_res2, hz_res3; + v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; + v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; + v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; + v8i16 minus5h = __msa_ldi_h(-5); + v8i16 plus20h = __msa_ldi_h(20); -void ff_put_h264_qpel8_mc12_msa(uint8_t *dst, const uint8_t *src, - ptrdiff_t stride) -{ - avc_luma_midh_qrt_8w_msa(src - (2 * stride) - 2, stride, dst, stride, 8, 0); -} + mask3 = mask0 + 4; + mask4 = mask1 + 4; + mask5 = mask2 + 4; -void ff_put_h264_qpel8_mc32_msa(uint8_t *dst, const uint8_t *src, - ptrdiff_t stride) -{ - avc_luma_midh_qrt_8w_msa(src - (2 * stride) - 2, stride, dst, stride, 8, 1); -} + src -= ((2 * stride) + 2); -void ff_put_h264_qpel4_mc12_msa(uint8_t *dst, const uint8_t *src, - ptrdiff_t stride) -{ - avc_luma_midh_qrt_4w_msa(src - (2 * stride) - 2, stride, dst, stride, 4, 0); -} + LD_SB5(src, stride, src0, src1, src2, src3, src4); + LD_SB5(src + 8, stride, src7, src8, src9, src10, src11); + src += (5 * stride); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + XORI_B5_128_SB(src7, src8, src9, src10, src11); -void ff_put_h264_qpel4_mc32_msa(uint8_t *dst, const uint8_t *src, - ptrdiff_t stride) -{ - avc_luma_midh_qrt_4w_msa(src - (2 * stride) - 2, stride, dst, stride, 4, 1); -} + for (row = 16; row--;) { + LD_SB2(src, 8, src5, src6); + src += stride; + XORI_B2_128_SB(src5, src6); -void ff_put_h264_qpel16_mc22_msa(uint8_t *dst, const uint8_t *src, - ptrdiff_t stride) -{ - avc_luma_mid_16w_msa(src - (2 * stride) - 2, stride, dst, stride, 16); + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src0, src1, src2, src3, src4, src5, + vt_res0, vt_res1); + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src7, src8, src9, src10, src11, src6, + vt_res2, vt_res3); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec2); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec5); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask3, + mask4, mask5, shf_vec6, shf_vec7, shf_vec8); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask3, + mask4, mask5, shf_vec9, shf_vec10, shf_vec11); + hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); + hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); + hz_res2 = __msa_hadd_s_w(shf_vec6, shf_vec6); + hz_res3 = __msa_hadd_s_w(shf_vec9, shf_vec9); + DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); + DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); + DPADD_SH2_SW(shf_vec7, shf_vec8, minus5h, plus20h, hz_res2, hz_res2); + DPADD_SH2_SW(shf_vec10, shf_vec11, minus5h, plus20h, hz_res3, hz_res3); + SRARI_W4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 10); + SAT_SW4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 7); + dst0 = __msa_srari_h(shf_vec2, 5); + dst1 = __msa_srari_h(shf_vec5, 5); + dst2 = __msa_srari_h(shf_vec8, 5); + dst3 = __msa_srari_h(shf_vec11, 5); + SAT_SH4_SH(dst0, dst1, dst2, dst3, 7); + dst0 = __msa_pckod_h(dst2, dst0); + dst1 = __msa_pckod_h(dst3, dst1); + PCKEV_H2_SH(hz_res2, hz_res0, hz_res3, hz_res1, dst2, dst3); + dst0 = __msa_aver_s_h(dst2, dst0); + dst1 = __msa_aver_s_h(dst3, dst1); + out = PCKEV_XORI128_UB(dst0, dst1); + ST_UB(out, dst); + dst += stride; + + src0 = src1; + src1 = src2; + src2 = src3; + src3 = src4; + src4 = src5; + src7 = src8; + src8 = src9; + src9 = src10; + src10 = src11; + src11 = src6; + } +} + +void ff_put_h264_qpel8_mc12_msa(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + uint32_t row; + v16u8 out; + v16i8 src0, src1, src2, src3, src4, src5, src6; + v8i16 vt_res0, vt_res1, vt_res2, vt_res3, dst0, dst1, dst2, dst3; + v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5, shf_vec6; + v8i16 shf_vec7, shf_vec8, shf_vec9, shf_vec10, shf_vec11; + v8i16 mask3, mask4, mask5; + v4i32 hz_res0, hz_res1, hz_res2, hz_res3; + v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; + v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; + v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; + v8i16 minus5h = __msa_ldi_h(-5); + v8i16 plus20h = __msa_ldi_h(20); + + mask3 = mask0 + 4; + mask4 = mask1 + 4; + mask5 = mask2 + 4; + + src -= ((2 * stride) + 2); + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + src += (5 * stride); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + + for (row = 4; row--;) { + LD_SB2(src, stride, src5, src6); + src += (2 * stride); + XORI_B2_128_SB(src5, src6); + + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src0, src1, src2, src3, src4, src5, + vt_res0, vt_res1); + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src1, src2, src3, src4, src5, src6, + vt_res2, vt_res3); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec2); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec5); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask3, + mask4, mask5, shf_vec6, shf_vec7, shf_vec8); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask3, + mask4, mask5, shf_vec9, shf_vec10, shf_vec11); + hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); + hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); + hz_res2 = __msa_hadd_s_w(shf_vec6, shf_vec6); + hz_res3 = __msa_hadd_s_w(shf_vec9, shf_vec9); + DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); + DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); + DPADD_SH2_SW(shf_vec7, shf_vec8, minus5h, plus20h, hz_res2, hz_res2); + DPADD_SH2_SW(shf_vec10, shf_vec11, minus5h, plus20h, hz_res3, hz_res3); + SRARI_W4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 10); + SAT_SW4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 7); + dst0 = __msa_srari_h(shf_vec2, 5); + dst1 = __msa_srari_h(shf_vec5, 5); + dst2 = __msa_srari_h(shf_vec8, 5); + dst3 = __msa_srari_h(shf_vec11, 5); + SAT_SH4_SH(dst0, dst1, dst2, dst3, 7); + PCKEV_H2_SH(dst2, dst0, dst3, dst1, dst0, dst1); + PCKEV_H2_SH(hz_res2, hz_res0, hz_res3, hz_res1, dst2, dst3); + dst0 = __msa_aver_s_h(dst2, dst0); + dst1 = __msa_aver_s_h(dst3, dst1); + out = PCKEV_XORI128_UB(dst0, dst1); + ST8x2_UB(out, dst, stride); + dst += (2 * stride); + + src0 = src2; + src1 = src3; + src2 = src4; + src3 = src5; + src4 = src6; + } +} + +void ff_put_h264_qpel8_mc32_msa(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + uint32_t row; + v16u8 out; + v16i8 src0, src1, src2, src3, src4, src5, src6; + v8i16 vt_res0, vt_res1, vt_res2, vt_res3, dst0, dst1, dst2, dst3; + v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5, shf_vec6; + v8i16 shf_vec7, shf_vec8, shf_vec9, shf_vec10, shf_vec11; + v8i16 mask3, mask4, mask5; + v4i32 hz_res0, hz_res1, hz_res2, hz_res3; + v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; + v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; + v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; + v8i16 minus5h = __msa_ldi_h(-5); + v8i16 plus20h = __msa_ldi_h(20); + + mask3 = mask0 + 4; + mask4 = mask1 + 4; + mask5 = mask2 + 4; + + src -= ((2 * stride) + 2); + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + src += (5 * stride); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + + for (row = 4; row--;) { + LD_SB2(src, stride, src5, src6); + src += (2 * stride); + XORI_B2_128_SB(src5, src6); + + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src0, src1, src2, src3, src4, src5, + vt_res0, vt_res1); + AVC_CALC_DPADD_B_6PIX_2COEFF_SH(src1, src2, src3, src4, src5, src6, + vt_res2, vt_res3); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec2); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec5); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask3, + mask4, mask5, shf_vec6, shf_vec7, shf_vec8); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask3, + mask4, mask5, shf_vec9, shf_vec10, shf_vec11); + hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); + hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); + hz_res2 = __msa_hadd_s_w(shf_vec6, shf_vec6); + hz_res3 = __msa_hadd_s_w(shf_vec9, shf_vec9); + DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); + DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); + DPADD_SH2_SW(shf_vec7, shf_vec8, minus5h, plus20h, hz_res2, hz_res2); + DPADD_SH2_SW(shf_vec10, shf_vec11, minus5h, plus20h, hz_res3, hz_res3); + SRARI_W4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 10); + SAT_SW4_SW(hz_res0, hz_res1, hz_res2, hz_res3, 7); + dst0 = __msa_srari_h(shf_vec2, 5); + dst1 = __msa_srari_h(shf_vec5, 5); + dst2 = __msa_srari_h(shf_vec8, 5); + dst3 = __msa_srari_h(shf_vec11, 5); + SAT_SH4_SH(dst0, dst1, dst2, dst3, 7); + dst0 = __msa_pckod_h(dst2, dst0); + dst1 = __msa_pckod_h(dst3, dst1); + PCKEV_H2_SH(hz_res2, hz_res0, hz_res3, hz_res1, dst2, dst3); + dst0 = __msa_aver_s_h(dst2, dst0); + dst1 = __msa_aver_s_h(dst3, dst1); + out = PCKEV_XORI128_UB(dst0, dst1); + ST8x2_UB(out, dst, stride); + dst += (2 * stride); + + src0 = src2; + src1 = src3; + src2 = src4; + src3 = src5; + src4 = src6; + } +} + +void ff_put_h264_qpel4_mc12_msa(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + const int16_t filt_const0 = 0xfb01; + const int16_t filt_const1 = 0x1414; + const int16_t filt_const2 = 0x1fb; + v16u8 out; + v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8; + v16i8 src10_r, src21_r, src32_r, src43_r, src54_r, src65_r, src76_r; + v16i8 src87_r, src10_l, src21_l, src32_l, src43_l, src54_l, src65_l; + v16i8 src76_l, src87_l, filt0, filt1, filt2; + v8i16 vt_res0, vt_res1, vt_res2, vt_res3, dst0, dst1, dst2, dst3, shf_vec7; + v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5, shf_vec6; + v4i32 hz_res0, hz_res1, hz_res2, hz_res3; + v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; + v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; + v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; + v8i16 minus5h = __msa_ldi_h(-5); + v8i16 plus20h = __msa_ldi_h(20); + v8i16 zeros = { 0 }; + + filt0 = (v16i8) __msa_fill_h(filt_const0); + filt1 = (v16i8) __msa_fill_h(filt_const1); + filt2 = (v16i8) __msa_fill_h(filt_const2); + + src -= ((2 * stride) + 2); + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + src += (5 * stride); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + LD_SB4(src, stride, src5, src6, src7, src8); + XORI_B4_128_SB(src5, src6, src7, src8); + + ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_r, src21_r, + src32_r, src43_r); + ILVR_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_r, src65_r, + src76_r, src87_r); + ILVL_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_l, src21_l, + src32_l, src43_l); + ILVL_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_l, src65_l, + src76_l, src87_l); + vt_res0 = AVC_DOT_SH3_SH(src10_r, src32_r, src54_r, filt0, filt1, filt2); + vt_res1 = AVC_DOT_SH3_SH(src10_l, src32_l, src54_l, filt0, filt1, filt2); + vt_res2 = AVC_DOT_SH3_SH(src21_r, src43_r, src65_r, filt0, filt1, filt2); + vt_res3 = AVC_DOT_SH3_SH(src21_l, src43_l, src65_l, filt0, filt1, filt2); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec2); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec5); + hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); + DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); + hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); + DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); + + vt_res0 = AVC_DOT_SH3_SH(src32_r, src54_r, src76_r, filt0, filt1, filt2); + vt_res1 = AVC_DOT_SH3_SH(src32_l, src54_l, src76_l, filt0, filt1, filt2); + vt_res2 = AVC_DOT_SH3_SH(src43_r, src65_r, src87_r, filt0, filt1, filt2); + vt_res3 = AVC_DOT_SH3_SH(src43_l, src65_l, src87_l, filt0, filt1, filt2); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec6); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec7); + hz_res2 = __msa_hadd_s_w(shf_vec0, shf_vec0); + DPADD_SH2_SW(shf_vec1, shf_vec6, minus5h, plus20h, hz_res2, hz_res2); + hz_res3 = __msa_hadd_s_w(shf_vec3, shf_vec3); + DPADD_SH2_SW(shf_vec4, shf_vec7, minus5h, plus20h, hz_res3, hz_res3); + + SRARI_W2_SW(hz_res0, hz_res1, 10); + SAT_SW2_SW(hz_res0, hz_res1, 7); + SRARI_W2_SW(hz_res2, hz_res3, 10); + SAT_SW2_SW(hz_res2, hz_res3, 7); + + dst0 = __msa_srari_h(shf_vec2, 5); + dst1 = __msa_srari_h(shf_vec5, 5); + dst2 = __msa_srari_h(shf_vec6, 5); + dst3 = __msa_srari_h(shf_vec7, 5); + + SAT_SH2_SH(dst0, dst1, 7); + SAT_SH2_SH(dst2, dst3, 7); + ILVEV_H2_SH(dst0, zeros, dst1, zeros, dst0, dst1); + ILVEV_H2_SH(dst2, zeros, dst3, zeros, dst2, dst3); + + hz_res0 = __msa_aver_s_w(hz_res0, (v4i32) dst0); + hz_res1 = __msa_aver_s_w(hz_res1, (v4i32) dst1); + hz_res2 = __msa_aver_s_w(hz_res2, (v4i32) dst2); + hz_res3 = __msa_aver_s_w(hz_res3, (v4i32) dst3); + + PCKEV_H2_SH(hz_res1, hz_res0, hz_res3, hz_res2, dst0, dst2); + out = PCKEV_XORI128_UB(dst0, dst2); + ST4x4_UB(out, out, 0, 1, 2, 3, dst, stride); +} + +void ff_put_h264_qpel4_mc32_msa(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + const int16_t filt_const0 = 0xfb01; + const int16_t filt_const1 = 0x1414; + const int16_t filt_const2 = 0x1fb; + v16u8 out; + v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8; + v16i8 src10_r, src21_r, src32_r, src43_r, src54_r, src65_r, src76_r; + v16i8 src87_r, src10_l, src21_l, src32_l, src43_l, src54_l, src65_l; + v16i8 src76_l, src87_l, filt0, filt1, filt2; + v8i16 vt_res0, vt_res1, vt_res2, vt_res3, dst0, dst1, dst2, dst3, shf_vec7; + v8i16 shf_vec0, shf_vec1, shf_vec2, shf_vec3, shf_vec4, shf_vec5, shf_vec6; + v4i32 hz_res0, hz_res1, hz_res2, hz_res3; + v8i16 mask0 = { 0, 5, 1, 6, 2, 7, 3, 8 }; + v8i16 mask1 = { 1, 4, 2, 5, 3, 6, 4, 7 }; + v8i16 mask2 = { 2, 3, 3, 4, 4, 5, 5, 6 }; + v8i16 minus5h = __msa_ldi_h(-5); + v8i16 plus20h = __msa_ldi_h(20); + v8i16 zeros = { 0 }; + + filt0 = (v16i8) __msa_fill_h(filt_const0); + filt1 = (v16i8) __msa_fill_h(filt_const1); + filt2 = (v16i8) __msa_fill_h(filt_const2); + + src -= ((2 * stride) + 2); + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + src += (5 * stride); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + LD_SB4(src, stride, src5, src6, src7, src8); + XORI_B4_128_SB(src5, src6, src7, src8); + + ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_r, src21_r, + src32_r, src43_r); + ILVR_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_r, src65_r, + src76_r, src87_r); + ILVL_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_l, src21_l, + src32_l, src43_l); + ILVL_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_l, src65_l, + src76_l, src87_l); + + vt_res0 = AVC_DOT_SH3_SH(src10_r, src32_r, src54_r, filt0, filt1, filt2); + vt_res1 = AVC_DOT_SH3_SH(src10_l, src32_l, src54_l, filt0, filt1, filt2); + vt_res2 = AVC_DOT_SH3_SH(src21_r, src43_r, src65_r, filt0, filt1, filt2); + vt_res3 = AVC_DOT_SH3_SH(src21_l, src43_l, src65_l, filt0, filt1, filt2); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec2); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec5); + hz_res0 = __msa_hadd_s_w(shf_vec0, shf_vec0); + DPADD_SH2_SW(shf_vec1, shf_vec2, minus5h, plus20h, hz_res0, hz_res0); + hz_res1 = __msa_hadd_s_w(shf_vec3, shf_vec3); + DPADD_SH2_SW(shf_vec4, shf_vec5, minus5h, plus20h, hz_res1, hz_res1); + + vt_res0 = AVC_DOT_SH3_SH(src32_r, src54_r, src76_r, filt0, filt1, filt2); + vt_res1 = AVC_DOT_SH3_SH(src32_l, src54_l, src76_l, filt0, filt1, filt2); + vt_res2 = AVC_DOT_SH3_SH(src43_r, src65_r, src87_r, filt0, filt1, filt2); + vt_res3 = AVC_DOT_SH3_SH(src43_l, src65_l, src87_l, filt0, filt1, filt2); + VSHF_H3_SH(vt_res0, vt_res1, vt_res0, vt_res1, vt_res0, vt_res1, mask0, + mask1, mask2, shf_vec0, shf_vec1, shf_vec6); + VSHF_H3_SH(vt_res2, vt_res3, vt_res2, vt_res3, vt_res2, vt_res3, mask0, + mask1, mask2, shf_vec3, shf_vec4, shf_vec7); + hz_res2 = __msa_hadd_s_w(shf_vec0, shf_vec0); + DPADD_SH2_SW(shf_vec1, shf_vec6, minus5h, plus20h, hz_res2, hz_res2); + hz_res3 = __msa_hadd_s_w(shf_vec3, shf_vec3); + DPADD_SH2_SW(shf_vec4, shf_vec7, minus5h, plus20h, hz_res3, hz_res3); + + SRARI_W2_SW(hz_res0, hz_res1, 10); + SAT_SW2_SW(hz_res0, hz_res1, 7); + SRARI_W2_SW(hz_res2, hz_res3, 10); + SAT_SW2_SW(hz_res2, hz_res3, 7); + + dst0 = __msa_srari_h(shf_vec2, 5); + dst1 = __msa_srari_h(shf_vec5, 5); + dst2 = __msa_srari_h(shf_vec6, 5); + dst3 = __msa_srari_h(shf_vec7, 5); + + SAT_SH2_SH(dst0, dst1, 7); + SAT_SH2_SH(dst2, dst3, 7); + + dst0 = __msa_ilvod_h(zeros, dst0); + dst1 = __msa_ilvod_h(zeros, dst1); + dst2 = __msa_ilvod_h(zeros, dst2); + dst3 = __msa_ilvod_h(zeros, dst3); + + hz_res0 = __msa_aver_s_w(hz_res0, (v4i32) dst0); + hz_res1 = __msa_aver_s_w(hz_res1, (v4i32) dst1); + hz_res2 = __msa_aver_s_w(hz_res2, (v4i32) dst2); + hz_res3 = __msa_aver_s_w(hz_res3, (v4i32) dst3); + + PCKEV_H2_SH(hz_res1, hz_res0, hz_res3, hz_res2, dst0, dst2); + out = PCKEV_XORI128_UB(dst0, dst2); + ST4x4_UB(out, out, 0, 1, 2, 3, dst, stride); +} + +void ff_put_h264_qpel16_mc22_msa(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + const int32_t filt_const0 = 0xfffb0001; + const int32_t filt_const1 = 0x140014; + const int32_t filt_const2 = 0x1fffb; + const uint8_t *src_tmp = src - (2 * stride) - 2; + uint8_t *dst_tmp = dst; + uint32_t multiple8_cnt, loop_cnt; + v16u8 out0, out1; + v16i8 src0, src1, src2, src3, src4, mask0, mask1, mask2; + v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6; + v8i16 hz_out7, hz_out8, dst0, dst1, dst2, dst3; + v8i16 hz_out10_r, hz_out21_r, hz_out32_r, hz_out43_r, hz_out54_r; + v8i16 hz_out65_r, hz_out76_r, hz_out87_r, hz_out10_l, hz_out21_l; + v8i16 hz_out32_l, hz_out43_l, hz_out54_l, hz_out65_l, hz_out76_l; + v8i16 hz_out87_l, filt0, filt1, filt2; + v4i32 tmp0, tmp1; + + filt0 = (v8i16) __msa_fill_w(filt_const0); + filt1 = (v8i16) __msa_fill_w(filt_const1); + filt2 = (v8i16) __msa_fill_w(filt_const2); + + LD_SB3(&luma_mask_arr[0], 16, mask0, mask1, mask2); + + for (multiple8_cnt = 2; multiple8_cnt--;) { + src = src_tmp; + dst = dst_tmp; + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + src += (5 * stride); + + hz_out0 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); + hz_out1 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); + hz_out2 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); + hz_out3 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); + hz_out4 = AVC_HORZ_FILTER_SH(src4, src4, mask0, mask1, mask2); + + for (loop_cnt = 4; loop_cnt--;) { + LD_SB4(src, stride, src0, src1, src2, src3); + XORI_B4_128_SB(src0, src1, src2, src3); + src += (4 * stride); + + hz_out5 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); + hz_out6 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); + hz_out7 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); + hz_out8 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); + + ILVR_H4_SH(hz_out1, hz_out0, hz_out2, hz_out1, hz_out3, hz_out2, + hz_out4, hz_out3, hz_out10_r, hz_out21_r, hz_out32_r, + hz_out43_r); + ILVL_H4_SH(hz_out1, hz_out0, hz_out2, hz_out1, hz_out3, hz_out2, + hz_out4, hz_out3, hz_out10_l, hz_out21_l, hz_out32_l, + hz_out43_l); + ILVR_H4_SH(hz_out5, hz_out4, hz_out6, hz_out5, hz_out7, hz_out6, + hz_out8, hz_out7, hz_out54_r, hz_out65_r, hz_out76_r, + hz_out87_r); + ILVL_H4_SH(hz_out5, hz_out4, hz_out6, hz_out5, hz_out7, hz_out6, + hz_out8, hz_out7, hz_out54_l, hz_out65_l, hz_out76_l, + hz_out87_l); + + tmp0 = AVC_DOT_SW3_SW(hz_out10_r, hz_out32_r, hz_out54_r, filt0, + filt1, filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out10_l, hz_out32_l, hz_out54_l, filt0, + filt1, filt2); + dst0 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out21_r, hz_out43_r, hz_out65_r, filt0, + filt1, filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out21_l, hz_out43_l, hz_out65_l, filt0, + filt1, filt2); + dst1 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out32_r, hz_out54_r, hz_out76_r, filt0, + filt1, filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out32_l, hz_out54_l, hz_out76_l, filt0, + filt1, filt2); + dst2 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out43_r, hz_out65_r, hz_out87_r, filt0, + filt1, filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out43_l, hz_out65_l, hz_out87_l, filt0, + filt1, filt2); + dst3 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + + out0 = PCKEV_XORI128_UB(dst0, dst1); + out1 = PCKEV_XORI128_UB(dst2, dst3); + ST8x4_UB(out0, out1, dst, stride); + dst += (4 * stride); + + hz_out0 = hz_out4; + hz_out1 = hz_out5; + hz_out2 = hz_out6; + hz_out3 = hz_out7; + hz_out4 = hz_out8; + } + + src_tmp += 8; + dst_tmp += 8; + } } void ff_put_h264_qpel8_mc22_msa(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) { - avc_luma_mid_8w_msa(src - (2 * stride) - 2, stride, dst, stride, 8); + const int32_t filt_const0 = 0xfffb0001; + const int32_t filt_const1 = 0x140014; + const int32_t filt_const2 = 0x1fffb; + v16u8 out0, out1; + v16i8 src0, src1, src2, src3, src4, mask0, mask1, mask2; + v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6; + v8i16 hz_out7, hz_out8, hz_out9, hz_out10, hz_out11, hz_out12; + v8i16 hz_out10_r, hz_out21_r, hz_out32_r, hz_out43_r, hz_out54_r; + v8i16 hz_out65_r, hz_out76_r, hz_out87_r, hz_out89_r, hz_out910_r; + v8i16 hz_out1110_r, hz_out1211_r, dst0, dst1, dst2, dst3; + v8i16 hz_out10_l, hz_out21_l, hz_out32_l, hz_out43_l, hz_out54_l; + v8i16 hz_out65_l, hz_out76_l, hz_out87_l, hz_out89_l, hz_out910_l; + v8i16 hz_out1110_l, hz_out1211_l, filt0, filt1, filt2; + v4i32 tmp0, tmp1; + + filt0 = (v8i16) __msa_fill_w(filt_const0); + filt1 = (v8i16) __msa_fill_w(filt_const1); + filt2 = (v8i16) __msa_fill_w(filt_const2); + + LD_SB3(&luma_mask_arr[0], 16, mask0, mask1, mask2); + + src -= ((2 * stride) + 2); + LD_SB5(src, stride, src0, src1, src2, src3, src4); + XORI_B5_128_SB(src0, src1, src2, src3, src4); + src += (5 * stride); + + hz_out0 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); + hz_out1 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); + hz_out2 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); + hz_out3 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); + hz_out4 = AVC_HORZ_FILTER_SH(src4, src4, mask0, mask1, mask2); + + LD_SB4(src, stride, src0, src1, src2, src3); + XORI_B4_128_SB(src0, src1, src2, src3); + src += (4 * stride); + hz_out5 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); + hz_out6 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); + hz_out7 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); + hz_out8 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); + ILVR_H4_SH(hz_out1, hz_out0, hz_out2, hz_out1, hz_out3, hz_out2, hz_out4, + hz_out3, hz_out10_r, hz_out21_r, hz_out32_r, hz_out43_r); + ILVL_H4_SH(hz_out1, hz_out0, hz_out2, hz_out1, hz_out3, hz_out2, hz_out4, + hz_out3, hz_out10_l, hz_out21_l, hz_out32_l, hz_out43_l); + ILVR_H4_SH(hz_out5, hz_out4, hz_out6, hz_out5, hz_out7, hz_out6, hz_out8, + hz_out7, hz_out54_r, hz_out65_r, hz_out76_r, hz_out87_r); + ILVL_H4_SH(hz_out5, hz_out4, hz_out6, hz_out5, hz_out7, hz_out6, hz_out8, + hz_out7, hz_out54_l, hz_out65_l, hz_out76_l, hz_out87_l); + + tmp0 = AVC_DOT_SW3_SW(hz_out10_r, hz_out32_r, hz_out54_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out10_l, hz_out32_l, hz_out54_l, filt0, filt1, + filt2); + dst0 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out21_r, hz_out43_r, hz_out65_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out21_l, hz_out43_l, hz_out65_l, filt0, filt1, + filt2); + dst1 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out32_r, hz_out54_r, hz_out76_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out32_l, hz_out54_l, hz_out76_l, filt0, filt1, + filt2); + dst2 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out43_r, hz_out65_r, hz_out87_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out43_l, hz_out65_l, hz_out87_l, filt0, filt1, + filt2); + dst3 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + out0 = PCKEV_XORI128_UB(dst0, dst1); + out1 = PCKEV_XORI128_UB(dst2, dst3); + ST8x4_UB(out0, out1, dst, stride); + dst += (4 * stride); + + LD_SB4(src, stride, src0, src1, src2, src3); + XORI_B4_128_SB(src0, src1, src2, src3); + hz_out9 = AVC_HORZ_FILTER_SH(src0, src0, mask0, mask1, mask2); + hz_out10 = AVC_HORZ_FILTER_SH(src1, src1, mask0, mask1, mask2); + hz_out11 = AVC_HORZ_FILTER_SH(src2, src2, mask0, mask1, mask2); + hz_out12 = AVC_HORZ_FILTER_SH(src3, src3, mask0, mask1, mask2); + ILVR_H4_SH(hz_out9, hz_out8, hz_out10, hz_out9, hz_out11, hz_out10, + hz_out12, hz_out11, hz_out89_r, hz_out910_r, hz_out1110_r, + hz_out1211_r); + ILVL_H4_SH(hz_out9, hz_out8, hz_out10, hz_out9, hz_out11, hz_out10, + hz_out12, hz_out11, hz_out89_l, hz_out910_l, hz_out1110_l, + hz_out1211_l); + tmp0 = AVC_DOT_SW3_SW(hz_out54_r, hz_out76_r, hz_out89_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out54_l, hz_out76_l, hz_out89_l, filt0, filt1, + filt2); + dst0 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out65_r, hz_out87_r, hz_out910_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out65_l, hz_out87_l, hz_out910_l, filt0, filt1, + filt2); + dst1 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out76_r, hz_out89_r, hz_out1110_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out76_l, hz_out89_l, hz_out1110_l, filt0, filt1, + filt2); + dst2 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out87_r, hz_out910_r, hz_out1211_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out87_l, hz_out910_l, hz_out1211_l, filt0, filt1, + filt2); + dst3 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + out0 = PCKEV_XORI128_UB(dst0, dst1); + out1 = PCKEV_XORI128_UB(dst2, dst3); + ST8x4_UB(out0, out1, dst, stride); } void ff_put_h264_qpel4_mc22_msa(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) { - avc_luma_mid_4w_msa(src - (2 * stride) - 2, stride, dst, stride, 4); + const int32_t filt_const0 = 0xfffb0001; + const int32_t filt_const1 = 0x140014; + const int32_t filt_const2 = 0x1fffb; + v16u8 res; + v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8; + v16i8 mask0, mask1, mask2; + v8i16 hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6; + v8i16 hz_out7, hz_out8, dst0, dst1, filt0, filt1, filt2; + v8i16 hz_out10_r, hz_out21_r, hz_out32_r, hz_out43_r, hz_out54_r; + v8i16 hz_out65_r, hz_out76_r, hz_out87_r; + v4i32 tmp0, tmp1; + + LD_SB3(&luma_mask_arr[48], 16, mask0, mask1, mask2); + + filt0 = (v8i16) __msa_fill_w(filt_const0); + filt1 = (v8i16) __msa_fill_w(filt_const1); + filt2 = (v8i16) __msa_fill_w(filt_const2); + + src -= ((2 * stride) + 2); + + LD_SB5(src, stride, src0, src1, src2, src3, src4); + src += (5 * stride); + LD_SB4(src, stride, src5, src6, src7, src8); + + XORI_B5_128_SB(src0, src1, src2, src3, src4); + XORI_B4_128_SB(src5, src6, src7, src8); + hz_out0 = AVC_HORZ_FILTER_SH(src0, src1, mask0, mask1, mask2); + hz_out2 = AVC_HORZ_FILTER_SH(src2, src3, mask0, mask1, mask2); + hz_out4 = AVC_HORZ_FILTER_SH(src4, src5, mask0, mask1, mask2); + hz_out6 = AVC_HORZ_FILTER_SH(src6, src7, mask0, mask1, mask2); + hz_out8 = AVC_HORZ_FILTER_SH(src8, src8, mask0, mask1, mask2); + PCKOD_D2_SH(hz_out0, hz_out0, hz_out2, hz_out2, hz_out1, hz_out3); + PCKOD_D2_SH(hz_out4, hz_out4, hz_out6, hz_out6, hz_out5, hz_out7); + ILVR_H4_SH(hz_out1, hz_out0, hz_out2, hz_out1, hz_out3, hz_out2, hz_out4, + hz_out3, hz_out10_r, hz_out21_r, hz_out32_r, hz_out43_r); + ILVR_H4_SH(hz_out5, hz_out4, hz_out6, hz_out5, hz_out7, hz_out6, hz_out8, + hz_out7, hz_out54_r, hz_out65_r, hz_out76_r, hz_out87_r); + + tmp0 = AVC_DOT_SW3_SW(hz_out10_r, hz_out32_r, hz_out54_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out21_r, hz_out43_r, hz_out65_r, filt0, filt1, + filt2); + dst0 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + tmp0 = AVC_DOT_SW3_SW(hz_out32_r, hz_out54_r, hz_out76_r, filt0, filt1, + filt2); + tmp1 = AVC_DOT_SW3_SW(hz_out43_r, hz_out65_r, hz_out87_r, filt0, filt1, + filt2); + dst1 = __msa_pckev_h((v8i16) tmp1, (v8i16) tmp0); + res = PCKEV_XORI128_UB(dst0, dst1); + ST4x4_UB(res, res, 0, 1, 2, 3, dst, stride); } void ff_avg_h264_qpel16_mc10_msa(uint8_t *dst, const uint8_t *src,