/****************************************************************************** * * Copyright (C) 2015 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at: * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. * ***************************************************************************** * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore */ /*****************************************************************************/ /* */ /* File Name : ih264_deblk_luma_ssse3.c */ /* */ /* Description : Contains function definitions for deblocking */ /* */ /* List of Functions : ih264_deblk_luma_vert_bs4_ssse3() */ /* ih264_deblk_luma_horz_bs4_ssse3() */ /* ih264_deblk_luma_vert_bslt4_ssse3() */ /* ih264_deblk_luma_horz_bslt4_ssse3() */ /* ih264_deblk_luma_vert_bs4_mbaff_ssse3() */ /* ih264_deblk_luma_vert_bslt4_mbaff_ssse3() */ /* */ /* Issues / Problems : None */ /* */ /* Revision History : */ /* */ /* DD MM YYYY Author(s) Changes (Describe the changes made) */ /* 12 02 2015 Naveen Kumar P Added luma deblocking ssse3 */ /* intrinsics */ /* */ /*****************************************************************************/ /*****************************************************************************/ /* File Includes */ /*****************************************************************************/ /* System include files */ #include /* User include files */ #include "ih264_typedefs.h" #include "ih264_platform_macros.h" #include "ih264_deblk_edge_filters.h" #include "ih264_macros.h" /*****************************************************************************/ /* Function Definitions */ /*****************************************************************************/ /*****************************************************************************/ /* */ /* Function Name : ih264_deblk_luma_vert_bs4_ssse3() */ /* */ /* Description : This function performs filtering of a luma block */ /* vertical edge when the boundary strength is set to 4. */ /* */ /* Inputs : pu1_src - pointer to the src sample q0 */ /* src_strd - source stride */ /* alpha - alpha value for the boundary */ /* beta - beta value for the boundary */ /* */ /* Globals : None */ /* */ /* Processing : This operation is described in Sec. 8.7.2.4 under the */ /* title "Filtering process for edges for bS equal to 4" in */ /* ITU T Rec H.264. */ /* */ /* Outputs : None */ /* */ /* Returns : None */ /* */ /* Issues : None */ /* */ /* Revision History: */ /* */ /* DD MM YYYY Author(s) Changes (Describe the changes made) */ /* 12 02 2015 Naveen Kumar P Initial version */ /* */ /*****************************************************************************/ void ih264_deblk_luma_vert_bs4_ssse3(UWORD8 *pu1_src, WORD32 src_strd, WORD32 alpha, WORD32 beta) { __m128i zero = _mm_setzero_si128(); __m128i q0_16x8, q1_16x8, q2_16x8, q3_16x8; __m128i p0_16x8, p1_16x8, p2_16x8, p3_16x8; __m128i q0_8x16, q1_8x16, q2_8x16, q3_8x16; __m128i p0_8x16, p1_8x16, p2_8x16, p3_8x16; __m128i q0_16x8_1; __m128i p0_16x8_1; __m128i q0_16x8_2, q1_16x8_2, q2_16x8_2; __m128i p0_16x8_2, p1_16x8_2, p2_16x8_2; __m128i temp1, temp2, temp3, temp4, temp5, temp6; __m128i Alpha_8x16, Beta_8x16; __m128i flag1_16x8, flag2_16x8, flag3_16x8, flag4_16x8; __m128i const_val2_16x8 = _mm_set1_epi16(2); __m128i line1, line2, line3, line4, line5, line6, line7, line8; Alpha_8x16 = _mm_set1_epi16(alpha); Beta_8x16 = _mm_set1_epi16(beta); line1 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 0 * src_strd)); line2 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 1 * src_strd)); line3 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 2 * src_strd)); line4 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 3 * src_strd)); line5 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 4 * src_strd)); line6 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 5 * src_strd)); line7 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 6 * src_strd)); line8 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 7 * src_strd)); temp1 = _mm_unpacklo_epi8(line1, line2); temp2 = _mm_unpacklo_epi8(line3, line4); temp3 = _mm_unpacklo_epi8(line5, line6); temp4 = _mm_unpacklo_epi8(line7, line8); line1 = _mm_unpacklo_epi16(temp1, temp2); line2 = _mm_unpackhi_epi16(temp1, temp2); line3 = _mm_unpacklo_epi16(temp3, temp4); line4 = _mm_unpackhi_epi16(temp3, temp4); p1_8x16 = _mm_unpacklo_epi32(line1, line3); p0_8x16 = _mm_unpackhi_epi32(line1, line3); q0_8x16 = _mm_unpacklo_epi32(line2, line4); q1_8x16 = _mm_unpackhi_epi32(line2, line4); line1 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 8 * src_strd)); line2 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 9 * src_strd)); line3 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 10 * src_strd)); line4 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 11 * src_strd)); line5 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 12 * src_strd)); line6 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 13 * src_strd)); line7 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 14 * src_strd)); line8 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 15 * src_strd)); temp1 = _mm_unpacklo_epi8(line1, line2); temp2 = _mm_unpacklo_epi8(line3, line4); temp3 = _mm_unpacklo_epi8(line5, line6); temp4 = _mm_unpacklo_epi8(line7, line8); line1 = _mm_unpacklo_epi16(temp1, temp2); line2 = _mm_unpackhi_epi16(temp1, temp2); line3 = _mm_unpacklo_epi16(temp3, temp4); line4 = _mm_unpackhi_epi16(temp3, temp4); temp1 = _mm_unpacklo_epi32(line1, line3); temp2 = _mm_unpackhi_epi32(line1, line3); temp3 = _mm_unpacklo_epi32(line2, line4); temp4 = _mm_unpackhi_epi32(line2, line4); p3_16x8 = _mm_unpacklo_epi64(p1_8x16, temp1); p2_16x8 = _mm_unpackhi_epi64(p1_8x16, temp1); q2_16x8 = _mm_unpacklo_epi64(q1_8x16, temp4); q3_16x8 = _mm_unpackhi_epi64(q1_8x16, temp4); p1_16x8 = _mm_unpacklo_epi64(p0_8x16, temp2); p0_16x8 = _mm_unpackhi_epi64(p0_8x16, temp2); q0_16x8 = _mm_unpacklo_epi64(q0_8x16, temp3); q1_16x8 = _mm_unpackhi_epi64(q0_8x16, temp3); //Cond1 (ABS(p0 - q0) < alpha) temp1 = _mm_subs_epu8(q0_16x8, p0_16x8); temp2 = _mm_subs_epu8(p0_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag1_16x8 = _mm_packs_epi16(temp2, temp1); //Cond2 (ABS(q1 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q1_16x8); temp2 = _mm_subs_epu8(q1_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); //Cond3 (ABS(p1 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p1_16x8); temp2 = _mm_subs_epu8(p1_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); // !((ABS(p0 - q0) < alpha) || (ABS(q1 - q0) < beta) || (ABS(p1 - p0) < beta)) flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p0 - q0) < ((alpha >> 2) + 2)) temp1 = _mm_subs_epu8(p0_16x8, q0_16x8); temp2 = _mm_subs_epu8(q0_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); Alpha_8x16 = _mm_srai_epi16(Alpha_8x16, 2); Alpha_8x16 = _mm_add_epi16(Alpha_8x16, const_val2_16x8); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag2_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p2 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p2_16x8); temp2 = _mm_subs_epu8(p2_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag3_16x8 = _mm_packs_epi16(temp2, temp1); flag3_16x8 = _mm_and_si128(flag3_16x8, flag2_16x8); // (ABS(q2 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q2_16x8); temp2 = _mm_subs_epu8(q2_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag4_16x8 = _mm_packs_epi16(temp2, temp1); flag4_16x8 = _mm_and_si128(flag4_16x8, flag2_16x8); // First 8 pixels p3_8x16 = _mm_unpacklo_epi8(p3_16x8, zero); p2_8x16 = _mm_unpacklo_epi8(p2_16x8, zero); p1_8x16 = _mm_unpacklo_epi8(p1_16x8, zero); p0_8x16 = _mm_unpacklo_epi8(p0_16x8, zero); q0_8x16 = _mm_unpacklo_epi8(q0_16x8, zero); q1_8x16 = _mm_unpacklo_epi8(q1_16x8, zero); q2_8x16 = _mm_unpacklo_epi8(q2_16x8, zero); q3_8x16 = _mm_unpacklo_epi8(q3_16x8, zero); // p0_1 and q0_1 temp1 = _mm_add_epi16(p0_8x16, q1_8x16); temp2 = _mm_add_epi16(p1_8x16, q0_8x16); temp5 = _mm_add_epi16(temp1, const_val2_16x8); temp6 = _mm_add_epi16(temp2, const_val2_16x8); temp3 = _mm_slli_epi16(p1_8x16, 1); temp4 = _mm_slli_epi16(q1_8x16, 1); temp1 = _mm_add_epi16(temp5, temp3); temp2 = _mm_add_epi16(temp6, temp4); p0_16x8_1 = _mm_srai_epi16(temp1, 2); q0_16x8_1 = _mm_srai_epi16(temp2, 2); // p1_2 and q1_2 temp6 = _mm_add_epi16(temp6, p0_8x16); temp5 = _mm_add_epi16(temp5, q0_8x16); temp1 = _mm_add_epi16(temp6, p2_8x16); temp2 = _mm_add_epi16(temp5, q2_8x16); p1_16x8_2 = _mm_srai_epi16(temp1, 2); q1_16x8_2 = _mm_srai_epi16(temp2, 2); // p0_2 and q0_2 temp1 = _mm_add_epi16(temp3, p2_8x16); temp2 = _mm_add_epi16(temp4, q2_8x16); temp1 = _mm_add_epi16(temp1, q1_8x16); temp2 = _mm_add_epi16(temp2, p1_8x16); temp3 = _mm_add_epi16(p0_8x16, q0_8x16); temp3 = _mm_slli_epi16(temp3, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp3); temp1 = _mm_add_epi16(temp1, _mm_set1_epi16(4)); temp2 = _mm_add_epi16(temp2, _mm_set1_epi16(4)); p0_16x8_2 = _mm_srai_epi16(temp1, 3); q0_16x8_2 = _mm_srai_epi16(temp2, 3); // p2_2 and q2_2 temp1 = _mm_add_epi16(temp6, const_val2_16x8); temp2 = _mm_add_epi16(temp5, const_val2_16x8); temp3 = _mm_slli_epi16(p2_8x16, 1); temp4 = _mm_slli_epi16(q2_8x16, 1); temp3 = _mm_add_epi16(p2_8x16, temp3); temp4 = _mm_add_epi16(q2_8x16, temp4); temp5 = _mm_slli_epi16(p3_8x16, 1); temp6 = _mm_slli_epi16(q3_8x16, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp4); temp1 = _mm_add_epi16(temp1, temp5); temp2 = _mm_add_epi16(temp2, temp6); p2_16x8_2 = _mm_srai_epi16(temp1, 3); q2_16x8_2 = _mm_srai_epi16(temp2, 3); // Second 8 pixels and packing with first 8 pixels p3_8x16 = _mm_unpackhi_epi8(p3_16x8, zero); p2_8x16 = _mm_unpackhi_epi8(p2_16x8, zero); p1_8x16 = _mm_unpackhi_epi8(p1_16x8, zero); p0_8x16 = _mm_unpackhi_epi8(p0_16x8, zero); q0_8x16 = _mm_unpackhi_epi8(q0_16x8, zero); q1_8x16 = _mm_unpackhi_epi8(q1_16x8, zero); q2_8x16 = _mm_unpackhi_epi8(q2_16x8, zero); q3_8x16 = _mm_unpackhi_epi8(q3_16x8, zero); // p0_1 and q0_1 temp1 = _mm_add_epi16(p0_8x16, q1_8x16); temp2 = _mm_add_epi16(p1_8x16, q0_8x16); temp5 = _mm_add_epi16(temp1, const_val2_16x8); temp6 = _mm_add_epi16(temp2, const_val2_16x8); temp3 = _mm_slli_epi16(p1_8x16, 1); temp4 = _mm_slli_epi16(q1_8x16, 1); temp1 = _mm_add_epi16(temp5, temp3); temp2 = _mm_add_epi16(temp6, temp4); temp1 = _mm_srai_epi16(temp1, 2); temp2 = _mm_srai_epi16(temp2, 2); p0_16x8_1 = _mm_packus_epi16(p0_16x8_1, temp1); q0_16x8_1 = _mm_packus_epi16(q0_16x8_1, temp2); // p1_2 and q1_2 temp6 = _mm_add_epi16(temp6, p0_8x16); temp5 = _mm_add_epi16(temp5, q0_8x16); temp1 = _mm_add_epi16(temp6, p2_8x16); temp2 = _mm_add_epi16(temp5, q2_8x16); temp1 = _mm_srai_epi16(temp1, 2); temp2 = _mm_srai_epi16(temp2, 2); p1_16x8_2 = _mm_packus_epi16(p1_16x8_2, temp1); q1_16x8_2 = _mm_packus_epi16(q1_16x8_2, temp2); // p0_2 and q0_2 temp1 = _mm_add_epi16(temp3, p2_8x16); temp2 = _mm_add_epi16(temp4, q2_8x16); temp1 = _mm_add_epi16(temp1, q1_8x16); temp2 = _mm_add_epi16(temp2, p1_8x16); temp3 = _mm_add_epi16(p0_8x16, q0_8x16); temp3 = _mm_slli_epi16(temp3, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp3); temp1 = _mm_add_epi16(temp1, _mm_set1_epi16(4)); temp2 = _mm_add_epi16(temp2, _mm_set1_epi16(4)); temp1 = _mm_srai_epi16(temp1, 3); temp2 = _mm_srai_epi16(temp2, 3); p0_16x8_2 = _mm_packus_epi16(p0_16x8_2, temp1); q0_16x8_2 = _mm_packus_epi16(q0_16x8_2, temp2); // p2_2 and q2_2 temp1 = _mm_add_epi16(temp6, const_val2_16x8); temp2 = _mm_add_epi16(temp5, const_val2_16x8); temp3 = _mm_slli_epi16(p2_8x16, 1); temp4 = _mm_slli_epi16(q2_8x16, 1); temp3 = _mm_add_epi16(p2_8x16, temp3); temp4 = _mm_add_epi16(q2_8x16, temp4); temp5 = _mm_slli_epi16(p3_8x16, 1); temp6 = _mm_slli_epi16(q3_8x16, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp4); temp1 = _mm_add_epi16(temp1, temp5); temp2 = _mm_add_epi16(temp2, temp6); temp1 = _mm_srai_epi16(temp1, 3); temp2 = _mm_srai_epi16(temp2, 3); p2_16x8_2 = _mm_packus_epi16(p2_16x8_2, temp1); q2_16x8_2 = _mm_packus_epi16(q2_16x8_2, temp2); // p0 and q0 p0_16x8 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi8(0xFF))); p0_16x8_1 = _mm_and_si128(p0_16x8_1, flag1_16x8); p0_16x8 = _mm_add_epi8(p0_16x8, p0_16x8_1); q0_16x8 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi8(0xFF))); q0_16x8_1 = _mm_and_si128(q0_16x8_1, flag1_16x8); q0_16x8 = _mm_add_epi8(q0_16x8, q0_16x8_1); // p0 and q0 p0_16x8 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p0_16x8_2 = _mm_and_si128(p0_16x8_2, flag3_16x8); p0_16x8 = _mm_add_epi8(p0_16x8, p0_16x8_2); q0_16x8 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q0_16x8_2 = _mm_and_si128(q0_16x8_2, flag4_16x8); q0_16x8 = _mm_add_epi8(q0_16x8, q0_16x8_2); // p1 and q1 p1_16x8 = _mm_and_si128(p1_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p1_16x8_2 = _mm_and_si128(p1_16x8_2, flag3_16x8); p1_16x8 = _mm_add_epi8(p1_16x8, p1_16x8_2); q1_16x8 = _mm_and_si128(q1_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q1_16x8_2 = _mm_and_si128(q1_16x8_2, flag4_16x8); q1_16x8 = _mm_add_epi8(q1_16x8, q1_16x8_2); // p2 and q2 p2_16x8 = _mm_and_si128(p2_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p2_16x8_2 = _mm_and_si128(p2_16x8_2, flag3_16x8); p2_16x8 = _mm_add_epi8(p2_16x8, p2_16x8_2); q2_16x8 = _mm_and_si128(q2_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q2_16x8_2 = _mm_and_si128(q2_16x8_2, flag4_16x8); q2_16x8 = _mm_add_epi8(q2_16x8, q2_16x8_2); temp1 = _mm_unpacklo_epi8(p3_16x8, p2_16x8); temp2 = _mm_unpacklo_epi8(p1_16x8, p0_16x8); temp3 = _mm_unpacklo_epi8(q0_16x8, q1_16x8); temp4 = _mm_unpacklo_epi8(q2_16x8, q3_16x8); p3_8x16 = _mm_unpacklo_epi16(temp1, temp2); p2_8x16 = _mm_unpackhi_epi16(temp1, temp2); q2_8x16 = _mm_unpacklo_epi16(temp3, temp4); q3_8x16 = _mm_unpackhi_epi16(temp3, temp4); line1 = _mm_unpacklo_epi32(p3_8x16, q2_8x16); line2 = _mm_srli_si128(line1, 8); line3 = _mm_unpackhi_epi32(p3_8x16, q2_8x16); line4 = _mm_srli_si128(line3, 8); line5 = _mm_unpacklo_epi32(p2_8x16, q3_8x16); line6 = _mm_srli_si128(line5, 8); line7 = _mm_unpackhi_epi32(p2_8x16, q3_8x16); line8 = _mm_srli_si128(line7, 8); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 0 * src_strd), line1); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 1 * src_strd), line2); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 2 * src_strd), line3); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 3 * src_strd), line4); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 4 * src_strd), line5); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 5 * src_strd), line6); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 6 * src_strd), line7); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 7 * src_strd), line8); temp1 = _mm_unpackhi_epi8(p3_16x8, p2_16x8); temp2 = _mm_unpackhi_epi8(p1_16x8, p0_16x8); temp3 = _mm_unpackhi_epi8(q0_16x8, q1_16x8); temp4 = _mm_unpackhi_epi8(q2_16x8, q3_16x8); p3_8x16 = _mm_unpacklo_epi16(temp1, temp2); p2_8x16 = _mm_unpackhi_epi16(temp1, temp2); q2_8x16 = _mm_unpacklo_epi16(temp3, temp4); q3_8x16 = _mm_unpackhi_epi16(temp3, temp4); line1 = _mm_unpacklo_epi32(p3_8x16, q2_8x16); line2 = _mm_srli_si128(line1, 8); line3 = _mm_unpackhi_epi32(p3_8x16, q2_8x16); line4 = _mm_srli_si128(line3, 8); line5 = _mm_unpacklo_epi32(p2_8x16, q3_8x16); line6 = _mm_srli_si128(line5, 8); line7 = _mm_unpackhi_epi32(p2_8x16, q3_8x16); line8 = _mm_srli_si128(line7, 8); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 8 * src_strd), line1); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 9 * src_strd), line2); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 10 * src_strd), line3); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 11 * src_strd), line4); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 12 * src_strd), line5); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 13 * src_strd), line6); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 14 * src_strd), line7); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 15 * src_strd), line8); } /*****************************************************************************/ /* */ /* Function Name : ih264_deblk_luma_horz_bs4_ssse3() */ /* */ /* Description : This function performs filtering of a luma block */ /* horizontal edge when the boundary strength is set to 4. */ /* */ /* Inputs : pu1_src - pointer to the src sample q0 */ /* src_strd - source stride */ /* alpha - alpha value for the boundary */ /* beta - beta value for the boundary */ /* */ /* Globals : None */ /* */ /* Processing : This operation is described in Sec. 8.7.2.4 under the */ /* title "Filtering process for edges for bS equal to 4" in */ /* ITU T Rec H.264. */ /* */ /* Outputs : None */ /* */ /* Returns : None */ /* */ /* Issues : None */ /* */ /* Revision History: */ /* */ /* DD MM YYYY Author(s) Changes (Describe the changes made) */ /* 12 02 2015 Naveen Kumar P Initial version */ /* */ /*****************************************************************************/ void ih264_deblk_luma_horz_bs4_ssse3(UWORD8 *pu1_src, WORD32 src_strd, WORD32 alpha, WORD32 beta) { WORD16 i16_posP3, i16_posP2, i16_posP1, i16_posP0; WORD16 i16_posQ1, i16_posQ2, i16_posQ3; UWORD8 *pu1_HorzPixel; __m128i zero = _mm_setzero_si128(); __m128i q0_16x8, q1_16x8, q2_16x8, q3_16x8; __m128i p0_16x8, p1_16x8, p2_16x8, p3_16x8; __m128i q0_8x16, q1_8x16, q2_8x16, q3_8x16; __m128i p0_8x16, p1_8x16, p2_8x16, p3_8x16; __m128i q0_16x8_1; __m128i p0_16x8_1; __m128i q0_16x8_2, q1_16x8_2, q2_16x8_2; __m128i p0_16x8_2, p1_16x8_2, p2_16x8_2; __m128i temp1, temp2, temp3, temp4, temp5, temp6; __m128i Alpha_8x16, Beta_8x16; __m128i flag1_16x8, flag2_16x8, flag3_16x8, flag4_16x8; __m128i const_val2_16x8 = _mm_set1_epi16(2); pu1_HorzPixel = pu1_src - (src_strd << 2); i16_posQ1 = src_strd; i16_posQ2 = X2(src_strd); i16_posQ3 = X3(src_strd); i16_posP0 = X3(src_strd); i16_posP1 = X2(src_strd); i16_posP2 = src_strd; i16_posP3 = 0; Alpha_8x16 = _mm_set1_epi16(alpha); Beta_8x16 = _mm_set1_epi16(beta); p3_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP3)); p2_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP2)); p1_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP1)); p0_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP0)); q0_16x8 = _mm_loadu_si128((__m128i *)(pu1_src)); q1_16x8 = _mm_loadu_si128((__m128i *)(pu1_src + i16_posQ1)); q2_16x8 = _mm_loadu_si128((__m128i *)(pu1_src + i16_posQ2)); q3_16x8 = _mm_loadu_si128((__m128i *)(pu1_src + i16_posQ3)); //Cond1 (ABS(p0 - q0) < alpha) temp1 = _mm_subs_epu8(q0_16x8, p0_16x8); temp2 = _mm_subs_epu8(p0_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag1_16x8 = _mm_packs_epi16(temp2, temp1); //Cond2 (ABS(q1 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q1_16x8); temp2 = _mm_subs_epu8(q1_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); //Cond3 (ABS(p1 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p1_16x8); temp2 = _mm_subs_epu8(p1_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); // !((ABS(p0 - q0) < alpha) || (ABS(q1 - q0) < beta) || (ABS(p1 - p0) < beta)) flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p0 - q0) < ((alpha >> 2) + 2)) temp1 = _mm_subs_epu8(p0_16x8, q0_16x8); temp2 = _mm_subs_epu8(q0_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); Alpha_8x16 = _mm_srai_epi16(Alpha_8x16, 2); Alpha_8x16 = _mm_add_epi16(Alpha_8x16, const_val2_16x8); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag2_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p2 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p2_16x8); temp2 = _mm_subs_epu8(p2_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag3_16x8 = _mm_packs_epi16(temp2, temp1); flag3_16x8 = _mm_and_si128(flag3_16x8, flag2_16x8); // (ABS(q2 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q2_16x8); temp2 = _mm_subs_epu8(q2_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag4_16x8 = _mm_packs_epi16(temp2, temp1); flag4_16x8 = _mm_and_si128(flag4_16x8, flag2_16x8); // First 8 pixels p3_8x16 = _mm_unpacklo_epi8(p3_16x8, zero); p2_8x16 = _mm_unpacklo_epi8(p2_16x8, zero); p1_8x16 = _mm_unpacklo_epi8(p1_16x8, zero); p0_8x16 = _mm_unpacklo_epi8(p0_16x8, zero); q0_8x16 = _mm_unpacklo_epi8(q0_16x8, zero); q1_8x16 = _mm_unpacklo_epi8(q1_16x8, zero); q2_8x16 = _mm_unpacklo_epi8(q2_16x8, zero); q3_8x16 = _mm_unpacklo_epi8(q3_16x8, zero); // p0_1 and q0_1 temp1 = _mm_add_epi16(p0_8x16, q1_8x16); temp2 = _mm_add_epi16(p1_8x16, q0_8x16); temp5 = _mm_add_epi16(temp1, const_val2_16x8); temp6 = _mm_add_epi16(temp2, const_val2_16x8); temp3 = _mm_slli_epi16(p1_8x16, 1); temp4 = _mm_slli_epi16(q1_8x16, 1); temp1 = _mm_add_epi16(temp5, temp3); temp2 = _mm_add_epi16(temp6, temp4); p0_16x8_1 = _mm_srai_epi16(temp1, 2); q0_16x8_1 = _mm_srai_epi16(temp2, 2); // p1_2 and q1_2 temp6 = _mm_add_epi16(temp6, p0_8x16); temp5 = _mm_add_epi16(temp5, q0_8x16); temp1 = _mm_add_epi16(temp6, p2_8x16); temp2 = _mm_add_epi16(temp5, q2_8x16); p1_16x8_2 = _mm_srai_epi16(temp1, 2); q1_16x8_2 = _mm_srai_epi16(temp2, 2); // p0_2 and q0_2 temp1 = _mm_add_epi16(temp3, p2_8x16); temp2 = _mm_add_epi16(temp4, q2_8x16); temp1 = _mm_add_epi16(temp1, q1_8x16); temp2 = _mm_add_epi16(temp2, p1_8x16); temp3 = _mm_add_epi16(p0_8x16, q0_8x16); temp3 = _mm_slli_epi16(temp3, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp3); temp1 = _mm_add_epi16(temp1, _mm_set1_epi16(4)); temp2 = _mm_add_epi16(temp2, _mm_set1_epi16(4)); p0_16x8_2 = _mm_srai_epi16(temp1, 3); q0_16x8_2 = _mm_srai_epi16(temp2, 3); // p2_2 and q2_2 temp1 = _mm_add_epi16(temp6, const_val2_16x8); temp2 = _mm_add_epi16(temp5, const_val2_16x8); temp3 = _mm_slli_epi16(p2_8x16, 1); temp4 = _mm_slli_epi16(q2_8x16, 1); temp3 = _mm_add_epi16(p2_8x16, temp3); temp4 = _mm_add_epi16(q2_8x16, temp4); temp5 = _mm_slli_epi16(p3_8x16, 1); temp6 = _mm_slli_epi16(q3_8x16, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp4); temp1 = _mm_add_epi16(temp1, temp5); temp2 = _mm_add_epi16(temp2, temp6); p2_16x8_2 = _mm_srai_epi16(temp1, 3); q2_16x8_2 = _mm_srai_epi16(temp2, 3); // Second 8 pixels and packing with first 8 pixels p3_8x16 = _mm_unpackhi_epi8(p3_16x8, zero); p2_8x16 = _mm_unpackhi_epi8(p2_16x8, zero); p1_8x16 = _mm_unpackhi_epi8(p1_16x8, zero); p0_8x16 = _mm_unpackhi_epi8(p0_16x8, zero); q0_8x16 = _mm_unpackhi_epi8(q0_16x8, zero); q1_8x16 = _mm_unpackhi_epi8(q1_16x8, zero); q2_8x16 = _mm_unpackhi_epi8(q2_16x8, zero); q3_8x16 = _mm_unpackhi_epi8(q3_16x8, zero); // p0_1 and q0_1 temp1 = _mm_add_epi16(p0_8x16, q1_8x16); temp2 = _mm_add_epi16(p1_8x16, q0_8x16); temp5 = _mm_add_epi16(temp1, const_val2_16x8); temp6 = _mm_add_epi16(temp2, const_val2_16x8); temp3 = _mm_slli_epi16(p1_8x16, 1); temp4 = _mm_slli_epi16(q1_8x16, 1); temp1 = _mm_add_epi16(temp5, temp3); temp2 = _mm_add_epi16(temp6, temp4); temp1 = _mm_srai_epi16(temp1, 2); temp2 = _mm_srai_epi16(temp2, 2); p0_16x8_1 = _mm_packus_epi16(p0_16x8_1, temp1); q0_16x8_1 = _mm_packus_epi16(q0_16x8_1, temp2); // p1_2 and q1_2 temp6 = _mm_add_epi16(temp6, p0_8x16); temp5 = _mm_add_epi16(temp5, q0_8x16); temp1 = _mm_add_epi16(temp6, p2_8x16); temp2 = _mm_add_epi16(temp5, q2_8x16); temp1 = _mm_srai_epi16(temp1, 2); temp2 = _mm_srai_epi16(temp2, 2); p1_16x8_2 = _mm_packus_epi16(p1_16x8_2, temp1); q1_16x8_2 = _mm_packus_epi16(q1_16x8_2, temp2); // p0_2 and q0_2 temp1 = _mm_add_epi16(temp3, p2_8x16); temp2 = _mm_add_epi16(temp4, q2_8x16); temp1 = _mm_add_epi16(temp1, q1_8x16); temp2 = _mm_add_epi16(temp2, p1_8x16); temp3 = _mm_add_epi16(p0_8x16, q0_8x16); temp3 = _mm_slli_epi16(temp3, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp3); temp1 = _mm_add_epi16(temp1, _mm_set1_epi16(4)); temp2 = _mm_add_epi16(temp2, _mm_set1_epi16(4)); temp1 = _mm_srai_epi16(temp1, 3); temp2 = _mm_srai_epi16(temp2, 3); p0_16x8_2 = _mm_packus_epi16(p0_16x8_2, temp1); q0_16x8_2 = _mm_packus_epi16(q0_16x8_2, temp2); // p2_2 and q2_2 temp1 = _mm_add_epi16(temp6, const_val2_16x8); temp2 = _mm_add_epi16(temp5, const_val2_16x8); temp3 = _mm_slli_epi16(p2_8x16, 1); temp4 = _mm_slli_epi16(q2_8x16, 1); temp3 = _mm_add_epi16(p2_8x16, temp3); temp4 = _mm_add_epi16(q2_8x16, temp4); temp5 = _mm_slli_epi16(p3_8x16, 1); temp6 = _mm_slli_epi16(q3_8x16, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp4); temp1 = _mm_add_epi16(temp1, temp5); temp2 = _mm_add_epi16(temp2, temp6); temp1 = _mm_srai_epi16(temp1, 3); temp2 = _mm_srai_epi16(temp2, 3); p2_16x8_2 = _mm_packus_epi16(p2_16x8_2, temp1); q2_16x8_2 = _mm_packus_epi16(q2_16x8_2, temp2); // p0 and q0 p0_16x8 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi8(0xFF))); p0_16x8_1 = _mm_and_si128(p0_16x8_1, flag1_16x8); p0_16x8 = _mm_add_epi8(p0_16x8, p0_16x8_1); q0_16x8 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi8(0xFF))); q0_16x8_1 = _mm_and_si128(q0_16x8_1, flag1_16x8); q0_16x8 = _mm_add_epi8(q0_16x8, q0_16x8_1); // p0 and q0 p0_16x8 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p0_16x8_2 = _mm_and_si128(p0_16x8_2, flag3_16x8); p0_16x8 = _mm_add_epi8(p0_16x8, p0_16x8_2); q0_16x8 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q0_16x8_2 = _mm_and_si128(q0_16x8_2, flag4_16x8); q0_16x8 = _mm_add_epi8(q0_16x8, q0_16x8_2); // p1 and q1 p1_16x8 = _mm_and_si128(p1_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p1_16x8_2 = _mm_and_si128(p1_16x8_2, flag3_16x8); p1_16x8 = _mm_add_epi8(p1_16x8, p1_16x8_2); q1_16x8 = _mm_and_si128(q1_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q1_16x8_2 = _mm_and_si128(q1_16x8_2, flag4_16x8); q1_16x8 = _mm_add_epi8(q1_16x8, q1_16x8_2); // p2 and q2 p2_16x8 = _mm_and_si128(p2_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p2_16x8_2 = _mm_and_si128(p2_16x8_2, flag3_16x8); p2_16x8 = _mm_add_epi8(p2_16x8, p2_16x8_2); q2_16x8 = _mm_and_si128(q2_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q2_16x8_2 = _mm_and_si128(q2_16x8_2, flag4_16x8); q2_16x8 = _mm_add_epi8(q2_16x8, q2_16x8_2); _mm_storeu_si128((__m128i *)(pu1_HorzPixel + i16_posP2), p2_16x8); _mm_storeu_si128((__m128i *)(pu1_HorzPixel + i16_posP1), p1_16x8); _mm_storeu_si128((__m128i *)(pu1_HorzPixel + i16_posP0), p0_16x8); _mm_storeu_si128((__m128i *)(pu1_src), q0_16x8); _mm_storeu_si128((__m128i *)(pu1_src + i16_posQ1), q1_16x8); _mm_storeu_si128((__m128i *)(pu1_src + i16_posQ2), q2_16x8); } /*****************************************************************************/ /* */ /* Function Name : ih264_deblk_luma_vert_bslt4_ssse3() */ /* */ /* Description : This function performs filtering of a luma block */ /* vertical edge when the boundary strength is less than 4. */ /* */ /* Inputs : pu1_src - pointer to the src sample q0 */ /* src_strd - source stride */ /* alpha - alpha value for the boundary */ /* beta - beta value for the boundary */ /* u4_bs - packed Boundary strength array */ /* pu1_cliptab - tc0_table */ /* */ /* Globals : None */ /* */ /* Processing : This operation is described in Sec. 8.7.2.3 under the */ /* title "Filtering process for edges for bS less than 4" */ /* in ITU T Rec H.264. */ /* */ /* Outputs : None */ /* */ /* Returns : None */ /* */ /* Issues : None */ /* */ /* Revision History: */ /* */ /* DD MM YYYY Author(s) Changes (Describe the changes made) */ /* 12 02 2015 Naveen Kumar P Initial version */ /* */ /*****************************************************************************/ void ih264_deblk_luma_vert_bslt4_ssse3(UWORD8 *pu1_src, WORD32 src_strd, WORD32 alpha, WORD32 beta, UWORD32 u4_bs, const UWORD8 *pu1_cliptab) { UWORD8 u1_Bs, u1_Bs1; WORD32 j = 0; __m128i linea, lineb, linec, lined, linee, linef, lineg, lineh; __m128i int1, int2, int3, int4, high1, high2; __m128i flag, flag1, i_C, i_C0; __m128i i_Ap, i_Aq, diff, const1, const2, in_macro, in_macrotemp, temp, temp1; __m128i zero = _mm_setzero_si128(); for(j = 0; j <= 8 * src_strd; j += 8 * src_strd) { //Transpose linea = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + j)); lineb = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + src_strd + j)); linec = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + 2 * src_strd + j)); lined = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + 3 * src_strd + j)); linea = _mm_unpacklo_epi8(linea, zero); lineb = _mm_unpacklo_epi8(lineb, zero); linec = _mm_unpacklo_epi8(linec, zero); lined = _mm_unpacklo_epi8(lined, zero); int1 = _mm_unpacklo_epi16(linea, lineb); lineb = _mm_unpackhi_epi16(linea, lineb); int2 = _mm_unpacklo_epi16(linec, lined); lined = _mm_unpackhi_epi16(linec, lined); linea = _mm_unpacklo_epi16(int1, int2); int1 = _mm_unpackhi_epi16(int1, int2); linec = _mm_unpacklo_epi16(lineb, lined); high1 = _mm_unpackhi_epi16(lineb, lined); linee = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + 4 * src_strd + j)); linef = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + 5 * src_strd + j)); lineg = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + 6 * src_strd + j)); lineh = _mm_loadl_epi64((__m128i *)(pu1_src - 3 + 7 * src_strd + j)); linee = _mm_unpacklo_epi8(linee, zero); linef = _mm_unpacklo_epi8(linef, zero); lineg = _mm_unpacklo_epi8(lineg, zero); lineh = _mm_unpacklo_epi8(lineh, zero); int2 = _mm_unpacklo_epi16(linee, linef); linef = _mm_unpackhi_epi16(linee, linef); int3 = _mm_unpacklo_epi16(lineg, lineh); lineh = _mm_unpackhi_epi16(lineg, lineh); linee = _mm_unpacklo_epi16(int2, int3); int2 = _mm_unpackhi_epi16(int2, int3); lineg = _mm_unpacklo_epi16(linef, lineh); high2 = _mm_unpackhi_epi16(linef, lineh); int4 = _mm_unpacklo_epi16(linea, linee); lineb = _mm_unpackhi_epi16(linea, linee); int3 = _mm_unpacklo_epi16(int1, int2); lined = _mm_unpackhi_epi16(int1, int2); int2 = _mm_unpacklo_epi16(linec, lineg); linef = _mm_unpackhi_epi16(linec, lineg); linea = int4; linec = int3; linee = int2; lineg = _mm_unpacklo_epi16(high1, high2); lineh = _mm_unpackhi_epi16(high1, high2); //end of transpose u1_Bs = (u4_bs >> 24) & 0xff; u1_Bs1 = (u4_bs >> 16) & 0xff; u4_bs <<= 16; flag1 = _mm_set_epi16(u1_Bs1, u1_Bs, u1_Bs1, u1_Bs, u1_Bs1, u1_Bs, u1_Bs1, u1_Bs); flag1 = _mm_cmpeq_epi16(flag1, zero); //Set flag to 1s and 0s flag1 = _mm_xor_si128(flag1, _mm_set1_epi16(0xFFFF)); //Invert for required mask i_C0 = _mm_set_epi16(pu1_cliptab[u1_Bs1], pu1_cliptab[u1_Bs], pu1_cliptab[u1_Bs1], pu1_cliptab[u1_Bs], pu1_cliptab[u1_Bs1], pu1_cliptab[u1_Bs], pu1_cliptab[u1_Bs1], pu1_cliptab[u1_Bs]); diff = _mm_subs_epi16(linec, lined); //Condn 1 diff = _mm_abs_epi16(diff); const1 = _mm_set1_epi16(alpha); flag = _mm_cmpgt_epi16(const1, diff); diff = _mm_subs_epi16(linee, lined); //Condtn 2 diff = _mm_abs_epi16(diff); const1 = _mm_set1_epi16(beta); flag = _mm_and_si128(flag, _mm_cmpgt_epi16(const1, diff)); diff = _mm_subs_epi16(lineb, linec); //Condtn 3 diff = _mm_abs_epi16(diff); flag = _mm_and_si128(flag, _mm_cmpgt_epi16(const1, diff)); //Const 1= Beta from now on flag = _mm_and_si128(flag, flag1); //Final flag (ui_B condition + other 3 conditions) //Adding Ap> 24) & 0xff; u1_Bs1 = (u4_bs >> 16) & 0xff; u1_Bs2 = (u4_bs >> 8) & 0xff; u1_Bs3 = (u4_bs >> 0) & 0xff; clip0 = pu1_cliptab[u1_Bs0]; clip1 = pu1_cliptab[u1_Bs1]; clip2 = pu1_cliptab[u1_Bs2]; clip3 = pu1_cliptab[u1_Bs3]; Alpha_8x16 = _mm_set1_epi16(alpha); Beta_8x16 = _mm_set1_epi16(beta); bs_flag_16x8b = _mm_set_epi8(u1_Bs3, u1_Bs3, u1_Bs3, u1_Bs3, u1_Bs2, u1_Bs2, u1_Bs2, u1_Bs2, u1_Bs1, u1_Bs1, u1_Bs1, u1_Bs1, u1_Bs0, u1_Bs0, u1_Bs0, u1_Bs0); C0_16x8 = _mm_set_epi8(clip3, clip3, clip3, clip3, clip2, clip2, clip2, clip2, clip1, clip1, clip1, clip1, clip0, clip0, clip0, clip0); bs_flag_16x8b = _mm_cmpeq_epi8(bs_flag_16x8b, zero); bs_flag_16x8b = _mm_xor_si128(bs_flag_16x8b, _mm_set1_epi8(0xFF)); //Invert for required mask C0_8x16 = _mm_unpacklo_epi8(C0_16x8, zero); C0_hi_8x16 = _mm_unpackhi_epi8(C0_16x8, zero); p1_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP1)); p0_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP0)); p2_16x8 = _mm_loadu_si128((__m128i *)(pu1_HorzPixel + i16_posP2)); q2_16x8 = _mm_loadu_si128((__m128i *)(pu1_src + i16_posQ2)); //Cond1 (ABS(p0 - q0) < alpha) temp1 = _mm_subs_epu8(q0_16x8, p0_16x8); temp2 = _mm_subs_epu8(p0_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag1_16x8 = _mm_packs_epi16(temp2, temp1); flag1_16x8 = _mm_and_si128(flag1_16x8, bs_flag_16x8b); //Cond2 (ABS(q1 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q1_16x8); temp2 = _mm_subs_epu8(q1_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); //Cond3 (ABS(p1 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p1_16x8); temp2 = _mm_subs_epu8(p1_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); // !((ABS(p0 - q0) < alpha) || (ABS(q1 - q0) < beta) || (ABS(p1 - p0) < beta)) flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p2 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p2_16x8); temp2 = _mm_subs_epu8(p2_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag2_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); temp2 = _mm_subs_epi16(zero, temp2); temp1 = _mm_subs_epi16(zero, temp1); C_8x16 = _mm_add_epi16(C0_8x16, temp2); C_hi_8x16 = _mm_add_epi16(C0_hi_8x16, temp1); // (ABS(q2 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q2_16x8); temp2 = _mm_subs_epu8(q2_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag3_16x8 = _mm_packs_epi16(temp2, temp1); flag3_16x8 = _mm_and_si128(flag1_16x8, flag3_16x8); temp2 = _mm_subs_epi16(zero, temp2); temp1 = _mm_subs_epi16(zero, temp1); C_8x16 = _mm_add_epi16(C_8x16, temp2); C_hi_8x16 = _mm_add_epi16(C_hi_8x16, temp1); const_val4_8x16 = _mm_set1_epi16(4); temp1 = _mm_subs_epi16(_mm_unpacklo_epi8(q0_16x8, zero), _mm_unpacklo_epi8(p0_16x8, zero)); temp2 = _mm_subs_epi16(_mm_unpacklo_epi8(p1_16x8, zero), _mm_unpacklo_epi8(q1_16x8, zero)); temp1 = _mm_slli_epi16(temp1, 2); temp1 = _mm_add_epi16(temp1, temp2); temp1 = _mm_add_epi16(temp1, const_val4_8x16); in_macro_16x8 = _mm_srai_epi16(temp1, 3); temp1 = _mm_subs_epi16(_mm_unpackhi_epi8(q0_16x8, zero), _mm_unpackhi_epi8(p0_16x8, zero)); temp2 = _mm_subs_epi16(_mm_unpackhi_epi8(p1_16x8, zero), _mm_unpackhi_epi8(q1_16x8, zero)); temp1 = _mm_slli_epi16(temp1, 2); temp1 = _mm_add_epi16(temp1, temp2); temp1 = _mm_add_epi16(temp1, const_val4_8x16); in_macro_hi_16x8 = _mm_srai_epi16(temp1, 3); in_macro_16x8 = _mm_min_epi16(C_8x16, in_macro_16x8); //CLIP3 in_macro_hi_16x8 = _mm_min_epi16(C_hi_8x16, in_macro_hi_16x8); //CLIP3 C_8x16 = _mm_subs_epi16(zero, C_8x16); C_hi_8x16 = _mm_subs_epi16(zero, C_hi_8x16); in_macro_16x8 = _mm_max_epi16(C_8x16, in_macro_16x8); //CLIP3 in_macro_hi_16x8 = _mm_max_epi16(C_hi_8x16, in_macro_hi_16x8); //CLIP3 temp1 = _mm_add_epi16(_mm_unpacklo_epi8(p0_16x8, zero), in_macro_16x8); temp2 = _mm_add_epi16(_mm_unpackhi_epi8(p0_16x8, zero), in_macro_hi_16x8); temp1 = _mm_packus_epi16(temp1, temp2); temp1 = _mm_and_si128(temp1, flag1_16x8); temp2 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi16(0xFFFF))); temp1 = _mm_add_epi8(temp1, temp2); _mm_storeu_si128((__m128i *)(pu1_HorzPixel + i16_posP0), temp1); temp1 = _mm_sub_epi16(_mm_unpacklo_epi8(q0_16x8, zero), in_macro_16x8); temp2 = _mm_sub_epi16(_mm_unpackhi_epi8(q0_16x8, zero), in_macro_hi_16x8); temp1 = _mm_packus_epi16(temp1, temp2); temp1 = _mm_and_si128(temp1, flag1_16x8); temp2 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi16(0xFFFF))); temp1 = _mm_add_epi8(temp1, temp2); _mm_storeu_si128((__m128i *)(pu1_src), temp1); //if(Ap < Beta) temp1 = _mm_avg_epu16(_mm_unpacklo_epi8(q0_16x8, zero), _mm_unpacklo_epi8(p0_16x8, zero)); temp2 = _mm_slli_epi16(_mm_unpacklo_epi8(p1_16x8, zero), 1); //temp2 = _mm_subs_epi16(zero,temp2); temp2 = _mm_subs_epi16(_mm_unpacklo_epi8(p2_16x8, zero), temp2); temp2 = _mm_add_epi16(temp1, temp2); in_macro_16x8 = _mm_srai_epi16(temp2, 1); temp1 = _mm_avg_epu16(_mm_unpackhi_epi8(q0_16x8, zero), _mm_unpackhi_epi8(p0_16x8, zero)); temp2 = _mm_slli_epi16(_mm_unpackhi_epi8(p1_16x8, zero), 1); //temp2 = _mm_subs_epi16(zero,temp2); temp2 = _mm_subs_epi16(_mm_unpackhi_epi8(p2_16x8, zero), temp2); temp2 = _mm_add_epi16(temp1, temp2); in_macro_hi_16x8 = _mm_srai_epi16(temp2, 1); in_macro_16x8 = _mm_min_epi16(C0_8x16, in_macro_16x8); //CLIP3 in_macro_hi_16x8 = _mm_min_epi16(C0_hi_8x16, in_macro_hi_16x8); //CLIP3 C0_8x16 = _mm_subs_epi16(zero, C0_8x16); C0_hi_8x16 = _mm_subs_epi16(zero, C0_hi_8x16); in_macro_16x8 = _mm_max_epi16(C0_8x16, in_macro_16x8); //CLIP3 in_macro_hi_16x8 = _mm_max_epi16(C0_hi_8x16, in_macro_hi_16x8); //CLIP3 temp1 = _mm_add_epi16(_mm_unpacklo_epi8(p1_16x8, zero), in_macro_16x8); temp2 = _mm_add_epi16(_mm_unpackhi_epi8(p1_16x8, zero), in_macro_hi_16x8); temp1 = _mm_packus_epi16(temp1, temp2); temp1 = _mm_and_si128(temp1, flag2_16x8); temp2 = _mm_and_si128(p1_16x8, _mm_xor_si128(flag2_16x8, _mm_set1_epi16(0xFFFF))); temp1 = _mm_add_epi8(temp1, temp2); _mm_storeu_si128((__m128i *)(pu1_HorzPixel + i16_posP1), temp1); //if(Aq < Beta) temp1 = _mm_avg_epu16(_mm_unpacklo_epi8(q0_16x8, zero), _mm_unpacklo_epi8(p0_16x8, zero)); temp2 = _mm_slli_epi16(_mm_unpacklo_epi8(q1_16x8, zero), 1); //temp2 = _mm_slli_epi16 (temp2, 1); temp2 = _mm_subs_epi16(_mm_unpacklo_epi8(q2_16x8, zero), temp2); temp2 = _mm_add_epi16(temp1, temp2); in_macro_16x8 = _mm_srai_epi16(temp2, 1); temp1 = _mm_avg_epu16(_mm_unpackhi_epi8(q0_16x8, zero), _mm_unpackhi_epi8(p0_16x8, zero)); temp2 = _mm_slli_epi16(_mm_unpackhi_epi8(q1_16x8, zero), 1); //temp2 = _mm_slli_epi16 (temp2, 1); temp2 = _mm_subs_epi16(_mm_unpackhi_epi8(q2_16x8, zero), temp2); temp2 = _mm_add_epi16(temp1, temp2); in_macro_hi_16x8 = _mm_srai_epi16(temp2, 1); in_macro_16x8 = _mm_max_epi16(C0_8x16, in_macro_16x8); //CLIP3 in_macro_hi_16x8 = _mm_max_epi16(C0_hi_8x16, in_macro_hi_16x8); //CLIP3 C0_8x16 = _mm_subs_epi16(zero, C0_8x16); C0_hi_8x16 = _mm_subs_epi16(zero, C0_hi_8x16); in_macro_16x8 = _mm_min_epi16(C0_8x16, in_macro_16x8); //CLIP3 in_macro_hi_16x8 = _mm_min_epi16(C0_hi_8x16, in_macro_hi_16x8); //CLIP3 temp1 = _mm_add_epi16(_mm_unpacklo_epi8(q1_16x8, zero), in_macro_16x8); temp2 = _mm_add_epi16(_mm_unpackhi_epi8(q1_16x8, zero), in_macro_hi_16x8); temp1 = _mm_packus_epi16(temp1, temp2); temp1 = _mm_and_si128(temp1, flag3_16x8); temp2 = _mm_and_si128(q1_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi16(0xFFFF))); temp1 = _mm_add_epi8(temp1, temp2); _mm_storeu_si128((__m128i *)(pu1_src + i16_posQ1), temp1); } /*****************************************************************************/ /* */ /* Function Name : ih264_deblk_luma_vert_bs4_mbaff_ssse3() */ /* */ /* Description : This function performs filtering of a luma block */ /* vertical edge when boundary strength is set to 4. */ /* */ /* Inputs : pu1_src - pointer to the src sample q0 */ /* src_strd - source stride */ /* alpha - alpha value for the boundary */ /* beta - beta value for the boundary */ /* */ /* Globals : None */ /* */ /* Processing : When the function is called twice, this operation is as */ /* described in Sec. 8.7.2.3 under the title "Filtering */ /* process for edges for bS equal to 4" in ITU T Rec H.264. */ /* */ /* Outputs : None */ /* */ /* Returns : None */ /* */ /* Issues : None */ /* */ /* Revision History: */ /* */ /* DD MM YYYY Author(s) Changes (Describe the changes made) */ /* 12 02 2015 Naveen Kumar P Initial version */ /* */ /*****************************************************************************/ void ih264_deblk_luma_vert_bs4_mbaff_ssse3(UWORD8 *pu1_src, WORD32 src_strd, WORD32 alpha, WORD32 beta) { __m128i zero = _mm_setzero_si128(); __m128i q0_16x8, q1_16x8, q2_16x8, q3_16x8; __m128i p0_16x8, p1_16x8, p2_16x8, p3_16x8; __m128i q0_8x16, q1_8x16, q2_8x16, q3_8x16; __m128i p0_8x16, p1_8x16, p2_8x16, p3_8x16; __m128i q0_16x8_1; __m128i p0_16x8_1; __m128i q0_16x8_2, q1_16x8_2, q2_16x8_2; __m128i p0_16x8_2, p1_16x8_2, p2_16x8_2; __m128i temp1, temp2, temp3, temp4, temp5, temp6; __m128i Alpha_8x16, Beta_8x16; __m128i flag1_16x8, flag2_16x8, flag3_16x8, flag4_16x8; __m128i const_val2_16x8 = _mm_set1_epi16(2); __m128i line1, line2, line3, line4, line5, line6, line7, line8; Alpha_8x16 = _mm_set1_epi16(alpha); Beta_8x16 = _mm_set1_epi16(beta); line1 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 0 * src_strd)); line2 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 1 * src_strd)); line3 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 2 * src_strd)); line4 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 3 * src_strd)); line5 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 4 * src_strd)); line6 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 5 * src_strd)); line7 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 6 * src_strd)); line8 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 7 * src_strd)); temp1 = _mm_unpacklo_epi8(line1, line2); temp2 = _mm_unpacklo_epi8(line3, line4); temp3 = _mm_unpacklo_epi8(line5, line6); temp4 = _mm_unpacklo_epi8(line7, line8); line1 = _mm_unpacklo_epi16(temp1, temp2); line2 = _mm_unpackhi_epi16(temp1, temp2); line3 = _mm_unpacklo_epi16(temp3, temp4); line4 = _mm_unpackhi_epi16(temp3, temp4); p1_8x16 = _mm_unpacklo_epi32(line1, line3); p0_8x16 = _mm_unpackhi_epi32(line1, line3); q0_8x16 = _mm_unpacklo_epi32(line2, line4); q1_8x16 = _mm_unpackhi_epi32(line2, line4); p3_16x8 = _mm_unpacklo_epi64(p1_8x16, zero); p2_16x8 = _mm_unpackhi_epi64(p1_8x16, zero); q2_16x8 = _mm_unpacklo_epi64(q1_8x16, zero); q3_16x8 = _mm_unpackhi_epi64(q1_8x16, zero); p1_16x8 = _mm_unpacklo_epi64(p0_8x16, zero); p0_16x8 = _mm_unpackhi_epi64(p0_8x16, zero); q0_16x8 = _mm_unpacklo_epi64(q0_8x16, zero); q1_16x8 = _mm_unpackhi_epi64(q0_8x16, zero); //Cond1 (ABS(p0 - q0) < alpha) temp1 = _mm_subs_epu8(q0_16x8, p0_16x8); temp2 = _mm_subs_epu8(p0_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag1_16x8 = _mm_packs_epi16(temp2, temp1); //Cond2 (ABS(q1 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q1_16x8); temp2 = _mm_subs_epu8(q1_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); //Cond3 (ABS(p1 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p1_16x8); temp2 = _mm_subs_epu8(p1_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); // !((ABS(p0 - q0) < alpha) || (ABS(q1 - q0) < beta) || (ABS(p1 - p0) < beta)) flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p0 - q0) < ((alpha >> 2) + 2)) temp1 = _mm_subs_epu8(p0_16x8, q0_16x8); temp2 = _mm_subs_epu8(q0_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); Alpha_8x16 = _mm_srai_epi16(Alpha_8x16, 2); Alpha_8x16 = _mm_add_epi16(Alpha_8x16, const_val2_16x8); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); temp1 = _mm_cmpgt_epi16(Alpha_8x16, temp1); flag2_16x8 = _mm_packs_epi16(temp2, temp1); flag2_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p2 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p2_16x8); temp2 = _mm_subs_epu8(p2_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag3_16x8 = _mm_packs_epi16(temp2, temp1); flag3_16x8 = _mm_and_si128(flag3_16x8, flag2_16x8); // (ABS(q2 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q2_16x8); temp2 = _mm_subs_epu8(q2_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp1 = _mm_unpackhi_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); temp1 = _mm_cmpgt_epi16(Beta_8x16, temp1); flag4_16x8 = _mm_packs_epi16(temp2, temp1); flag4_16x8 = _mm_and_si128(flag4_16x8, flag2_16x8); // First 8 pixels p3_8x16 = _mm_unpacklo_epi8(p3_16x8, zero); p2_8x16 = _mm_unpacklo_epi8(p2_16x8, zero); p1_8x16 = _mm_unpacklo_epi8(p1_16x8, zero); p0_8x16 = _mm_unpacklo_epi8(p0_16x8, zero); q0_8x16 = _mm_unpacklo_epi8(q0_16x8, zero); q1_8x16 = _mm_unpacklo_epi8(q1_16x8, zero); q2_8x16 = _mm_unpacklo_epi8(q2_16x8, zero); q3_8x16 = _mm_unpacklo_epi8(q3_16x8, zero); // p0_1 and q0_1 temp1 = _mm_add_epi16(p0_8x16, q1_8x16); temp2 = _mm_add_epi16(p1_8x16, q0_8x16); temp5 = _mm_add_epi16(temp1, const_val2_16x8); temp6 = _mm_add_epi16(temp2, const_val2_16x8); temp3 = _mm_slli_epi16(p1_8x16, 1); temp4 = _mm_slli_epi16(q1_8x16, 1); temp1 = _mm_add_epi16(temp5, temp3); temp2 = _mm_add_epi16(temp6, temp4); p0_16x8_1 = _mm_srai_epi16(temp1, 2); q0_16x8_1 = _mm_srai_epi16(temp2, 2); // p1_2 and q1_2 temp6 = _mm_add_epi16(temp6, p0_8x16); temp5 = _mm_add_epi16(temp5, q0_8x16); temp1 = _mm_add_epi16(temp6, p2_8x16); temp2 = _mm_add_epi16(temp5, q2_8x16); p1_16x8_2 = _mm_srai_epi16(temp1, 2); q1_16x8_2 = _mm_srai_epi16(temp2, 2); // p0_2 and q0_2 temp1 = _mm_add_epi16(temp3, p2_8x16); temp2 = _mm_add_epi16(temp4, q2_8x16); temp1 = _mm_add_epi16(temp1, q1_8x16); temp2 = _mm_add_epi16(temp2, p1_8x16); temp3 = _mm_add_epi16(p0_8x16, q0_8x16); temp3 = _mm_slli_epi16(temp3, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp3); temp1 = _mm_add_epi16(temp1, _mm_set1_epi16(4)); temp2 = _mm_add_epi16(temp2, _mm_set1_epi16(4)); p0_16x8_2 = _mm_srai_epi16(temp1, 3); q0_16x8_2 = _mm_srai_epi16(temp2, 3); // p2_2 and q2_2 temp1 = _mm_add_epi16(temp6, const_val2_16x8); temp2 = _mm_add_epi16(temp5, const_val2_16x8); temp3 = _mm_slli_epi16(p2_8x16, 1); temp4 = _mm_slli_epi16(q2_8x16, 1); temp3 = _mm_add_epi16(p2_8x16, temp3); temp4 = _mm_add_epi16(q2_8x16, temp4); temp5 = _mm_slli_epi16(p3_8x16, 1); temp6 = _mm_slli_epi16(q3_8x16, 1); temp1 = _mm_add_epi16(temp1, temp3); temp2 = _mm_add_epi16(temp2, temp4); temp1 = _mm_add_epi16(temp1, temp5); temp2 = _mm_add_epi16(temp2, temp6); p2_16x8_2 = _mm_srai_epi16(temp1, 3); q2_16x8_2 = _mm_srai_epi16(temp2, 3); // p0_1 and q0_1 p0_16x8_1 = _mm_packus_epi16(p0_16x8_1, zero); q0_16x8_1 = _mm_packus_epi16(q0_16x8_1, zero); // p1_2 and q1_2 p1_16x8_2 = _mm_packus_epi16(p1_16x8_2, zero); q1_16x8_2 = _mm_packus_epi16(q1_16x8_2, zero); // p0_2 and q0_2 p0_16x8_2 = _mm_packus_epi16(p0_16x8_2, zero); q0_16x8_2 = _mm_packus_epi16(q0_16x8_2, zero); // p2_2 and q2_2 p2_16x8_2 = _mm_packus_epi16(p2_16x8_2, zero); q2_16x8_2 = _mm_packus_epi16(q2_16x8_2, zero); // p0 and q0 p0_16x8 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi8(0xFF))); p0_16x8_1 = _mm_and_si128(p0_16x8_1, flag1_16x8); p0_16x8 = _mm_add_epi8(p0_16x8, p0_16x8_1); q0_16x8 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi8(0xFF))); q0_16x8_1 = _mm_and_si128(q0_16x8_1, flag1_16x8); q0_16x8 = _mm_add_epi8(q0_16x8, q0_16x8_1); // p0 and q0 p0_16x8 = _mm_and_si128(p0_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p0_16x8_2 = _mm_and_si128(p0_16x8_2, flag3_16x8); p0_16x8 = _mm_add_epi8(p0_16x8, p0_16x8_2); q0_16x8 = _mm_and_si128(q0_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q0_16x8_2 = _mm_and_si128(q0_16x8_2, flag4_16x8); q0_16x8 = _mm_add_epi8(q0_16x8, q0_16x8_2); // p1 and q1 p1_16x8 = _mm_and_si128(p1_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p1_16x8_2 = _mm_and_si128(p1_16x8_2, flag3_16x8); p1_16x8 = _mm_add_epi8(p1_16x8, p1_16x8_2); q1_16x8 = _mm_and_si128(q1_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q1_16x8_2 = _mm_and_si128(q1_16x8_2, flag4_16x8); q1_16x8 = _mm_add_epi8(q1_16x8, q1_16x8_2); // p2 and q2 p2_16x8 = _mm_and_si128(p2_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi8(0xFF))); p2_16x8_2 = _mm_and_si128(p2_16x8_2, flag3_16x8); p2_16x8 = _mm_add_epi8(p2_16x8, p2_16x8_2); q2_16x8 = _mm_and_si128(q2_16x8, _mm_xor_si128(flag4_16x8, _mm_set1_epi8(0xFF))); q2_16x8_2 = _mm_and_si128(q2_16x8_2, flag4_16x8); q2_16x8 = _mm_add_epi8(q2_16x8, q2_16x8_2); temp1 = _mm_unpacklo_epi8(p3_16x8, p2_16x8); temp2 = _mm_unpacklo_epi8(p1_16x8, p0_16x8); temp3 = _mm_unpacklo_epi8(q0_16x8, q1_16x8); temp4 = _mm_unpacklo_epi8(q2_16x8, q3_16x8); p3_8x16 = _mm_unpacklo_epi16(temp1, temp2); p2_8x16 = _mm_unpackhi_epi16(temp1, temp2); q2_8x16 = _mm_unpacklo_epi16(temp3, temp4); q3_8x16 = _mm_unpackhi_epi16(temp3, temp4); line1 = _mm_unpacklo_epi32(p3_8x16, q2_8x16); line2 = _mm_srli_si128(line1, 8); line3 = _mm_unpackhi_epi32(p3_8x16, q2_8x16); line4 = _mm_srli_si128(line3, 8); line5 = _mm_unpacklo_epi32(p2_8x16, q3_8x16); line6 = _mm_srli_si128(line5, 8); line7 = _mm_unpackhi_epi32(p2_8x16, q3_8x16); line8 = _mm_srli_si128(line7, 8); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 0 * src_strd), line1); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 1 * src_strd), line2); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 2 * src_strd), line3); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 3 * src_strd), line4); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 4 * src_strd), line5); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 5 * src_strd), line6); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 6 * src_strd), line7); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 7 * src_strd), line8); } /*****************************************************************************/ /* */ /* Function Name : ih264_deblk_luma_vert_bslt4_mbaff_ssse3() */ /* */ /* Description : This function performs filtering of a luma block */ /* vertical edge when boundary strength is less than 4. */ /* */ /* Inputs : pu1_src - pointer to the src sample q0 */ /* src_strd - source stride */ /* alpha - alpha value for the boundary */ /* beta - beta value for the boundary */ /* u4_bs - packed Boundary strength array */ /* pu1_cliptab - tc0_table */ /* */ /* Globals : None */ /* */ /* Processing : When the function is called twice, this operation is as */ /* described in Sec. 8.7.2.3 under the title "Filtering */ /* process for edges for bS less than 4" in ITU T Rec H.264.*/ /* */ /* Outputs : None */ /* */ /* Returns : None */ /* */ /* Issues : None */ /* */ /* Revision History: */ /* */ /* DD MM YYYY Author(s) Changes (Describe the changes made) */ /* 12 02 2015 Naveen Kumar P Initial version */ /* */ /*****************************************************************************/ void ih264_deblk_luma_vert_bslt4_mbaff_ssse3(UWORD8 *pu1_src, WORD32 src_strd, WORD32 alpha, WORD32 beta, UWORD32 u4_bs, const UWORD8 *pu1_cliptab) { __m128i zero = _mm_setzero_si128(); __m128i bs_flag_16x8b, C0_16x8, C0_8x16, C_8x16; __m128i q0_16x8, q1_16x8, q2_16x8, q3_16x8; __m128i p0_16x8, p1_16x8, p2_16x8, p3_16x8; __m128i temp1, temp2, temp3, temp4; __m128i Alpha_8x16, Beta_8x16, flag1_16x8, flag2_16x8, flag3_16x8; __m128i in_macro_16x8; __m128i const_val4_8x16; UWORD8 u1_Bs0, u1_Bs1, u1_Bs2, u1_Bs3; UWORD8 clip0, clip1, clip2, clip3; __m128i line1, line2, line3, line4, line5, line6, line7, line8; __m128i q0_16x8_1, q1_16x8_1, q0_16x8_2; __m128i p0_16x8_1, p1_16x8_1, p0_16x8_2; line1 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 0 * src_strd)); line2 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 1 * src_strd)); line3 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 2 * src_strd)); line4 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 3 * src_strd)); line5 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 4 * src_strd)); line6 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 5 * src_strd)); line7 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 6 * src_strd)); line8 = _mm_loadl_epi64((__m128i *)(pu1_src - 4 + 7 * src_strd)); temp1 = _mm_unpacklo_epi8(line1, line2); temp2 = _mm_unpacklo_epi8(line3, line4); temp3 = _mm_unpacklo_epi8(line5, line6); temp4 = _mm_unpacklo_epi8(line7, line8); line1 = _mm_unpacklo_epi16(temp1, temp2); line2 = _mm_unpackhi_epi16(temp1, temp2); line3 = _mm_unpacklo_epi16(temp3, temp4); line4 = _mm_unpackhi_epi16(temp3, temp4); temp1 = _mm_unpacklo_epi32(line1, line3); temp2 = _mm_unpackhi_epi32(line1, line3); temp3 = _mm_unpacklo_epi32(line2, line4); temp4 = _mm_unpackhi_epi32(line2, line4); p3_16x8 = _mm_unpacklo_epi64(temp1, zero); p2_16x8 = _mm_unpackhi_epi64(temp1, zero); q2_16x8 = _mm_unpacklo_epi64(temp4, zero); q3_16x8 = _mm_unpackhi_epi64(temp4, zero); p1_16x8 = _mm_unpacklo_epi64(temp2, zero); p0_16x8 = _mm_unpackhi_epi64(temp2, zero); q0_16x8 = _mm_unpacklo_epi64(temp3, zero); q1_16x8 = _mm_unpackhi_epi64(temp3, zero); u1_Bs0 = (u4_bs >> 24) & 0xff; u1_Bs1 = (u4_bs >> 16) & 0xff; u1_Bs2 = (u4_bs >> 8) & 0xff; u1_Bs3 = (u4_bs >> 0) & 0xff; clip0 = pu1_cliptab[u1_Bs0]; clip1 = pu1_cliptab[u1_Bs1]; clip2 = pu1_cliptab[u1_Bs2]; clip3 = pu1_cliptab[u1_Bs3]; Alpha_8x16 = _mm_set1_epi16(alpha); Beta_8x16 = _mm_set1_epi16(beta); bs_flag_16x8b = _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0, u1_Bs3, u1_Bs3, u1_Bs2, u1_Bs2, u1_Bs1, u1_Bs1, u1_Bs0, u1_Bs0); C0_16x8 = _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0, clip3, clip3, clip2, clip2, clip1, clip1, clip0, clip0); bs_flag_16x8b = _mm_cmpeq_epi8(bs_flag_16x8b, zero); bs_flag_16x8b = _mm_xor_si128(bs_flag_16x8b, _mm_set1_epi8(0xFF)); //Invert for required mask C0_8x16 = _mm_unpacklo_epi8(C0_16x8, zero); //Cond1 (ABS(p0 - q0) < alpha) temp1 = _mm_subs_epu8(q0_16x8, p0_16x8); temp2 = _mm_subs_epu8(p0_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Alpha_8x16, temp2); flag1_16x8 = _mm_packs_epi16(temp2, zero); flag1_16x8 = _mm_and_si128(flag1_16x8, bs_flag_16x8b); //Cond2 (ABS(q1 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q1_16x8); temp2 = _mm_subs_epu8(q1_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); flag2_16x8 = _mm_packs_epi16(temp2, zero); flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); //Cond3 (ABS(p1 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p1_16x8); temp2 = _mm_subs_epu8(p1_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); flag2_16x8 = _mm_packs_epi16(temp2, zero); // !((ABS(p0 - q0) < alpha) || (ABS(q1 - q0) < beta) || (ABS(p1 - p0) < beta)) flag1_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); // (ABS(p2 - p0) < beta) temp1 = _mm_subs_epu8(p0_16x8, p2_16x8); temp2 = _mm_subs_epu8(p2_16x8, p0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); flag2_16x8 = _mm_packs_epi16(temp2, zero); flag2_16x8 = _mm_and_si128(flag1_16x8, flag2_16x8); temp2 = _mm_subs_epi16(zero, temp2); C_8x16 = _mm_add_epi16(C0_8x16, temp2); // (ABS(q2 - q0) < beta) temp1 = _mm_subs_epu8(q0_16x8, q2_16x8); temp2 = _mm_subs_epu8(q2_16x8, q0_16x8); temp1 = _mm_add_epi8(temp1, temp2); temp2 = _mm_unpacklo_epi8(temp1, zero); temp2 = _mm_cmpgt_epi16(Beta_8x16, temp2); flag3_16x8 = _mm_packs_epi16(temp2, zero); flag3_16x8 = _mm_and_si128(flag1_16x8, flag3_16x8); temp2 = _mm_subs_epi16(zero, temp2); C_8x16 = _mm_add_epi16(C_8x16, temp2); const_val4_8x16 = _mm_set1_epi16(4); temp1 = _mm_subs_epi16(_mm_unpacklo_epi8(q0_16x8, zero), _mm_unpacklo_epi8(p0_16x8, zero)); temp2 = _mm_subs_epi16(_mm_unpacklo_epi8(p1_16x8, zero), _mm_unpacklo_epi8(q1_16x8, zero)); temp1 = _mm_slli_epi16(temp1, 2); temp1 = _mm_add_epi16(temp1, temp2); temp1 = _mm_add_epi16(temp1, const_val4_8x16); in_macro_16x8 = _mm_srai_epi16(temp1, 3); in_macro_16x8 = _mm_min_epi16(C_8x16, in_macro_16x8); //CLIP3 C_8x16 = _mm_subs_epi16(zero, C_8x16); in_macro_16x8 = _mm_max_epi16(C_8x16, in_macro_16x8); //CLIP3 // p0 temp1 = _mm_add_epi16(_mm_unpacklo_epi8(p0_16x8, zero), in_macro_16x8); temp1 = _mm_packus_epi16(temp1, zero); p0_16x8_1 = _mm_and_si128(temp1, flag1_16x8); p0_16x8_2 = _mm_and_si128( p0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi16(0xFFFF))); p0_16x8_1 = _mm_add_epi8(p0_16x8_1, p0_16x8_2); // q0 temp1 = _mm_sub_epi16(_mm_unpacklo_epi8(q0_16x8, zero), in_macro_16x8); temp1 = _mm_packus_epi16(temp1, zero); q0_16x8_1 = _mm_and_si128(temp1, flag1_16x8); q0_16x8_2 = _mm_and_si128( q0_16x8, _mm_xor_si128(flag1_16x8, _mm_set1_epi16(0xFFFF))); q0_16x8_1 = _mm_add_epi8(q0_16x8_1, q0_16x8_2); //if(Ap < Beta) temp1 = _mm_avg_epu16(_mm_unpacklo_epi8(q0_16x8, zero), _mm_unpacklo_epi8(p0_16x8, zero)); temp2 = _mm_slli_epi16(_mm_unpacklo_epi8(p1_16x8, zero), 1); //temp2 = _mm_subs_epi16(zero,temp2); temp2 = _mm_subs_epi16(_mm_unpacklo_epi8(p2_16x8, zero), temp2); temp2 = _mm_add_epi16(temp1, temp2); in_macro_16x8 = _mm_srai_epi16(temp2, 1); in_macro_16x8 = _mm_min_epi16(C0_8x16, in_macro_16x8); //CLIP3 C0_8x16 = _mm_subs_epi16(zero, C0_8x16); in_macro_16x8 = _mm_max_epi16(C0_8x16, in_macro_16x8); //CLIP3 // p1 temp1 = _mm_add_epi16(_mm_unpacklo_epi8(p1_16x8, zero), in_macro_16x8); temp1 = _mm_packus_epi16(temp1, zero); p1_16x8_1 = _mm_and_si128(temp1, flag2_16x8); p1_16x8 = _mm_and_si128(p1_16x8, _mm_xor_si128(flag2_16x8, _mm_set1_epi16(0xFFFF))); p1_16x8 = _mm_add_epi8(p1_16x8, p1_16x8_1); //if(Aq < Beta) temp1 = _mm_avg_epu16(_mm_unpacklo_epi8(q0_16x8, zero), _mm_unpacklo_epi8(p0_16x8, zero)); temp2 = _mm_slli_epi16(_mm_unpacklo_epi8(q1_16x8, zero), 1); //temp2 = _mm_slli_epi16 (temp2, 1); temp2 = _mm_subs_epi16(_mm_unpacklo_epi8(q2_16x8, zero), temp2); temp2 = _mm_add_epi16(temp1, temp2); in_macro_16x8 = _mm_srai_epi16(temp2, 1); in_macro_16x8 = _mm_max_epi16(C0_8x16, in_macro_16x8); //CLIP3 C0_8x16 = _mm_subs_epi16(zero, C0_8x16); in_macro_16x8 = _mm_min_epi16(C0_8x16, in_macro_16x8); //CLIP3 temp1 = _mm_add_epi16(_mm_unpacklo_epi8(q1_16x8, zero), in_macro_16x8); // q1 temp1 = _mm_packus_epi16(temp1, zero); q1_16x8_1 = _mm_and_si128(temp1, flag3_16x8); q1_16x8 = _mm_and_si128(q1_16x8, _mm_xor_si128(flag3_16x8, _mm_set1_epi16(0xFFFF))); q1_16x8 = _mm_add_epi8(q1_16x8, q1_16x8_1); temp1 = _mm_unpacklo_epi8(p3_16x8, p2_16x8); temp2 = _mm_unpacklo_epi8(p1_16x8, p0_16x8_1); temp3 = _mm_unpacklo_epi8(q0_16x8_1, q1_16x8); temp4 = _mm_unpacklo_epi8(q2_16x8, q3_16x8); line7 = _mm_unpacklo_epi16(temp1, temp2); temp1 = _mm_unpackhi_epi16(temp1, temp2); line8 = _mm_unpacklo_epi16(temp3, temp4); temp2 = _mm_unpackhi_epi16(temp3, temp4); line1 = _mm_unpacklo_epi32(line7, line8); line2 = _mm_srli_si128(line1, 8); line3 = _mm_unpackhi_epi32(line7, line8); line4 = _mm_srli_si128(line3, 8); line5 = _mm_unpacklo_epi32(temp1, temp2); line6 = _mm_srli_si128(line5, 8); line7 = _mm_unpackhi_epi32(temp1, temp2); line8 = _mm_srli_si128(line7, 8); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 0 * src_strd), line1); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 1 * src_strd), line2); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 2 * src_strd), line3); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 3 * src_strd), line4); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 4 * src_strd), line5); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 5 * src_strd), line6); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 6 * src_strd), line7); _mm_storel_epi64((__m128i *)(pu1_src - 4 + 7 * src_strd), line8); }