| /*! |
| ********************************************************************************** |
| * Copyright (c) 2021 Loongson Technology Corporation Limited |
| * Contributed by Lu Wang <wanglu@loongson.cn> |
| * |
| * \copy |
| * Copyright (c) 2013, Cisco Systems |
| * All rights reserved. |
| * |
| * Redistribution and use in source and binary forms, with or without |
| * modification, are permitted provided that the following conditions |
| * are met: |
| * |
| * * Redistributions of source code must retain the above copyright |
| * notice, this list of conditions and the following disclaimer. |
| * |
| * * Redistributions in binary form must reproduce the above copyright |
| * notice, this list of conditions and the following disclaimer in |
| * the documentation and/or other materials provided with the |
| * distribution. |
| * |
| * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS |
| * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE |
| * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
| * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, |
| * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
| * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN |
| * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| * POSSIBILITY OF SUCH DAMAGE. |
| * \file vaa_lsx.c |
| * |
| * \brief Loongson optimization |
| * |
| * \date 12/10/2021 Created |
| * |
| ********************************************************************************** |
| */ |
| |
| #include "stdint.h" |
| #include "loongson_intrinsics.h" |
| |
| void VAACalcSad_lsx (const uint8_t* pCurData, const uint8_t* pRefData, |
| int32_t iPicWidth, int32_t iPicHeight, int32_t iPicStride, |
| int32_t* pFrameSad, int32_t* pSad8x8) { |
| const uint8_t* tmp_ref = pRefData; |
| const uint8_t* tmp_cur = pCurData; |
| int32_t iMbWidth = (iPicWidth >> 4); |
| int32_t mb_height = (iPicHeight >> 4); |
| int32_t mb_index = 0; |
| int32_t pic_stride_x8 = iPicStride << 3; |
| int32_t step = (iPicStride << 4) - iPicWidth; |
| |
| *pFrameSad = 0; |
| for (int32_t i = 0; i < mb_height; i++) { |
| for (int32_t j = 0; j < iMbWidth; j++) { |
| int32_t k, l_sad; |
| const uint8_t* tmp_cur_row; |
| const uint8_t* tmp_ref_row; |
| int32_t tmp_mb_index = mb_index << 2; |
| int32_t tmp_mb_index1 = tmp_mb_index + 1; |
| int32_t tmp_mb_index2 = tmp_mb_index + 2; |
| int32_t tmp_mb_index3 = tmp_mb_index + 3; |
| __m128i cur, ref; |
| __m128i vec_abs_diff, tmp_l_sad; |
| __m128i zero = __lsx_vreplgr2vr_b(0); |
| __m128i vec_l_sad = zero; |
| |
| l_sad = 0; |
| tmp_cur_row = tmp_cur; |
| tmp_ref_row = tmp_ref; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| tmp_l_sad = __lsx_vhaddw_w_h(vec_l_sad, vec_l_sad); |
| tmp_l_sad = __lsx_vhaddw_d_w(tmp_l_sad, tmp_l_sad); |
| tmp_l_sad = __lsx_vhaddw_q_d(tmp_l_sad, tmp_l_sad); |
| l_sad = __lsx_vpickve2gr_d(tmp_l_sad, 0); |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index] = l_sad; |
| |
| l_sad = 0; |
| tmp_cur_row = tmp_cur + 8; |
| tmp_ref_row = tmp_ref + 8; |
| vec_l_sad = zero; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| tmp_l_sad = __lsx_vhaddw_w_h(vec_l_sad, vec_l_sad); |
| tmp_l_sad = __lsx_vhaddw_d_w(tmp_l_sad, tmp_l_sad); |
| tmp_l_sad = __lsx_vhaddw_q_d(tmp_l_sad, tmp_l_sad); |
| l_sad = __lsx_vpickve2gr_d(tmp_l_sad, 0); |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index1] = l_sad; |
| |
| l_sad = 0; |
| tmp_cur_row = tmp_cur + pic_stride_x8; |
| tmp_ref_row = tmp_ref + pic_stride_x8; |
| vec_l_sad = zero; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| tmp_l_sad = __lsx_vhaddw_w_h(vec_l_sad, vec_l_sad); |
| tmp_l_sad = __lsx_vhaddw_d_w(tmp_l_sad, tmp_l_sad); |
| tmp_l_sad = __lsx_vhaddw_q_d(tmp_l_sad, tmp_l_sad); |
| l_sad = __lsx_vpickve2gr_d(tmp_l_sad, 0); |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index2] = l_sad; |
| |
| l_sad = 0; |
| tmp_cur_row = tmp_cur + pic_stride_x8 + 8; |
| tmp_ref_row = tmp_ref + pic_stride_x8 + 8; |
| vec_l_sad = zero; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| tmp_l_sad = __lsx_vhaddw_w_h(vec_l_sad, vec_l_sad); |
| tmp_l_sad = __lsx_vhaddw_d_w(tmp_l_sad, tmp_l_sad); |
| tmp_l_sad = __lsx_vhaddw_q_d(tmp_l_sad, tmp_l_sad); |
| l_sad = __lsx_vpickve2gr_d(tmp_l_sad, 0); |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index3] = l_sad; |
| |
| tmp_ref += 16; |
| tmp_cur += 16; |
| ++mb_index; |
| } |
| tmp_ref += step; |
| tmp_cur += step; |
| } |
| } |
| |
| void VAACalcSadBgd_lsx (const uint8_t* pCurData, const uint8_t* pRefData, |
| int32_t iPicWidth, int32_t iPicHeight, int32_t iPicStride, |
| int32_t* pFrameSad, int32_t* pSad8x8, int32_t* pSd8x8, |
| uint8_t* pMad8x8) { |
| const uint8_t* tmp_ref = pRefData; |
| const uint8_t* tmp_cur = pCurData; |
| int32_t iMbWidth = (iPicWidth >> 4); |
| int32_t mb_height = (iPicHeight >> 4); |
| int32_t mb_index = 0; |
| int32_t pic_stride_x8 = iPicStride << 3; |
| int32_t step = (iPicStride << 4) - iPicWidth; |
| |
| *pFrameSad = 0; |
| for (int32_t i = 0; i < mb_height; i++) { |
| for (int32_t j = 0; j < iMbWidth; j++) { |
| int32_t k; |
| int32_t l_sad, l_sd, l_mad; |
| const uint8_t* tmp_cur_row; |
| const uint8_t* tmp_ref_row; |
| int32_t tmp_mb_index = mb_index << 2; |
| int32_t tmp_mb_index1 = tmp_mb_index + 1; |
| int32_t tmp_mb_index2 = tmp_mb_index + 2; |
| int32_t tmp_mb_index3 = tmp_mb_index + 3; |
| __m128i cur, ref; |
| __m128i vec_diff, vec_abs_diff, tmp_l_sd, tmp_l_sad, tmp_l_mad; |
| __m128i zero = __lsx_vreplgr2vr_b(0); |
| __m128i vec_l_sd = zero; |
| __m128i vec_l_sad = zero; |
| __m128i vec_l_mad = zero; |
| |
| l_mad = l_sd = l_sad = 0; |
| tmp_cur_row = tmp_cur; |
| tmp_ref_row = tmp_ref; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_diff = __lsx_vsub_h(cur, ref); |
| vec_l_sd = __lsx_vadd_h(vec_l_sd, vec_diff); |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| |
| DUP2_ARG2(__lsx_vhaddw_w_h, vec_l_sd, vec_l_sd, vec_l_sad, vec_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_d_w, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_q_d, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vpickve2gr_d, tmp_l_sd, 0, tmp_l_sad, 0, l_sd, l_sad); |
| |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 8); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 4); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 2); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| l_mad = __lsx_vpickve2gr_h(vec_l_mad, 0); |
| |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index] = l_sad; |
| pSd8x8[tmp_mb_index] = l_sd; |
| pMad8x8[tmp_mb_index] = l_mad; |
| |
| l_mad = l_sd = l_sad = 0; |
| tmp_cur_row = tmp_cur + 8; |
| tmp_ref_row = tmp_ref + 8; |
| vec_l_sd = vec_l_sad = vec_l_mad = zero; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_diff = __lsx_vsub_h(cur, ref); |
| vec_l_sd = __lsx_vadd_h(vec_l_sd, vec_diff); |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| |
| DUP2_ARG2(__lsx_vhaddw_w_h, vec_l_sd, vec_l_sd, vec_l_sad, vec_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_d_w, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_q_d, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vpickve2gr_d, tmp_l_sd, 0, tmp_l_sad, 0, l_sd, l_sad); |
| |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 8); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 4); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 2); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| l_mad = __lsx_vpickve2gr_h(vec_l_mad, 0); |
| |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index1] = l_sad; |
| pSd8x8[tmp_mb_index1] = l_sd; |
| pMad8x8[tmp_mb_index1] = l_mad; |
| |
| l_mad = l_sd = l_sad = 0; |
| tmp_cur_row = tmp_cur + pic_stride_x8; |
| tmp_ref_row = tmp_ref + pic_stride_x8; |
| vec_l_sd = vec_l_sad = vec_l_mad = zero; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_diff = __lsx_vsub_h(cur, ref); |
| vec_l_sd = __lsx_vadd_h(vec_l_sd, vec_diff); |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| |
| DUP2_ARG2(__lsx_vhaddw_w_h, vec_l_sd, vec_l_sd, vec_l_sad, vec_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_d_w, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_q_d, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vpickve2gr_d, tmp_l_sd, 0, tmp_l_sad, 0, l_sd, l_sad); |
| |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 8); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 4); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 2); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| l_mad = __lsx_vpickve2gr_h(vec_l_mad, 0); |
| |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index2] = l_sad; |
| pSd8x8[tmp_mb_index2] = l_sd; |
| pMad8x8[tmp_mb_index2] = l_mad; |
| |
| l_mad = l_sd = l_sad = 0; |
| tmp_cur_row = tmp_cur + pic_stride_x8 + 8; |
| tmp_ref_row = tmp_ref + pic_stride_x8 + 8; |
| vec_l_sd = vec_l_sad = vec_l_mad = zero; |
| for (k = 0; k < 8; k ++) { |
| DUP2_ARG2(__lsx_vld, tmp_cur_row, 0, tmp_ref_row, 0, cur, ref); |
| DUP2_ARG2(__lsx_vilvl_b, zero, cur, zero, ref, cur, ref); |
| |
| vec_diff = __lsx_vsub_h(cur, ref); |
| vec_l_sd = __lsx_vadd_h(vec_l_sd, vec_diff); |
| vec_abs_diff = __lsx_vabsd_h(cur, ref); |
| vec_l_sad = __lsx_vadd_h(vec_l_sad, vec_abs_diff); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, vec_abs_diff); |
| tmp_cur_row += iPicStride; |
| tmp_ref_row += iPicStride; |
| } |
| |
| DUP2_ARG2(__lsx_vhaddw_w_h, vec_l_sd, vec_l_sd, vec_l_sad, vec_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_d_w, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vhaddw_q_d, tmp_l_sd, tmp_l_sd, tmp_l_sad, tmp_l_sad, |
| tmp_l_sd, tmp_l_sad); |
| DUP2_ARG2(__lsx_vpickve2gr_d, tmp_l_sd, 0, tmp_l_sad, 0, l_sd, l_sad); |
| |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 8); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 4); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| tmp_l_mad = __lsx_vbsrl_v(vec_l_mad, 2); |
| vec_l_mad = __lsx_vmax_h(vec_l_mad, tmp_l_mad); |
| l_mad = __lsx_vpickve2gr_h(vec_l_mad, 0); |
| |
| *pFrameSad += l_sad; |
| pSad8x8[tmp_mb_index3] = l_sad; |
| pSd8x8[tmp_mb_index3] = l_sd; |
| pMad8x8[tmp_mb_index3] = l_mad; |
| |
| tmp_ref += 16; |
| tmp_cur += 16; |
| ++mb_index; |
| } |
| tmp_ref += step; |
| tmp_cur += step; |
| } |
| } |
