1 /******************************************************************************
2 *
3 * Copyright (C) 2018 The Android Open Source Project
4 *
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at:
8 *
9 * http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 *
17 *****************************************************************************
18 * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
19 */
20 /*!
21 ******************************************************************************
22 * \file ihevce_me_utils_instr_set_router.c
23 *
24 * \brief
25 * This file contains function pointer initialization of me utility
26 * functions
27 *
28 * \date
29 * 15/07/2013
30 *
31 * \author
32 * Ittiam
33 *
34 * List of Functions
35 * ihevce_me_utils_instr_set_router()
36 *
37 ******************************************************************************
38 */
39
40 /*****************************************************************************/
41 /* File Includes */
42 /*****************************************************************************/
43 /* System include files */
44 #include <stdio.h>
45 #include <string.h>
46 #include <assert.h>
47
48 /* User include files */
49 #include "ihevc_typedefs.h"
50 #include "itt_video_api.h"
51 #include "ihevc_chroma_itrans_recon.h"
52 #include "ihevc_chroma_intra_pred.h"
53 #include "ihevc_debug.h"
54 #include "ihevc_deblk.h"
55 #include "ihevc_defs.h"
56 #include "ihevc_itrans_recon.h"
57 #include "ihevc_intra_pred.h"
58 #include "ihevc_inter_pred.h"
59 #include "ihevc_macros.h"
60 #include "ihevc_mem_fns.h"
61 #include "ihevc_padding.h"
62 #include "ihevc_quant_iquant_ssd.h"
63 #include "ihevc_resi_trans.h"
64 #include "ihevc_sao.h"
65 #include "ihevc_structs.h"
66 #include "ihevc_weighted_pred.h"
67 #include "ihevc_platform_macros.h"
68
69 #include "rc_cntrl_param.h"
70 #include "rc_frame_info_collector.h"
71 #include "rc_look_ahead_params.h"
72
73 #include "ihevce_api.h"
74 #include "ihevce_defs.h"
75 #include "ihevce_lap_enc_structs.h"
76 #include "ihevce_multi_thrd_structs.h"
77 #include "ihevce_function_selector.h"
78 #include "ihevce_me_common_defs.h"
79 #include "ihevce_enc_structs.h"
80 #include "ihevce_had_satd.h"
81 #include "ihevce_cmn_utils_instr_set_router.h"
82
83 #include "hme_datatype.h"
84 #include "hme_common_defs.h"
85 #include "hme_common_utils.h"
86 #include "hme_interface.h"
87 #include "hme_defs.h"
88 #include "hme_err_compute.h"
89 #include "hme_globals.h"
90
91 #include "ihevce_me_instr_set_router.h"
92
93 /*****************************************************************************/
94 /* Globals */
95 /*****************************************************************************/
96 static FT_SAD_EVALUATOR *gapf_sad_pt_npu[NUM_BLK_SIZES];
97 static FT_PART_SADS_EVALUATOR_16X16CU *gpf_part_sads_evaluator_16x16CU;
98 static FT_PART_SADS_EVALUATOR *gpf_part_sads_evaluator_MxM;
99 static FT_SAD_EVALUATOR *gpf_sad_grid_mxn;
100 /* 9 => Number of function types */
101 /* 2 => Number of results to store */
102 static FT_CALC_SAD_AND_RESULT *gapf_calc_sad_and_result_fxn[9][2];
103
104 static U08 gau1_calc_sad_and_result[2][2][4][TOT_NUM_PARTS] = {
105 //grid flag = 0
106 { //noise = 0
107 { //NxN or NxN & SMP
108 { 1, 2, 2, 2, 2, 2, 2, 2, 2, 4, 4, 4, 4, 4, 4, 4, 4 },
109 //SMP only
110 { 1, 2, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4 },
111 //AMP
112 { 1, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4 },
113 //2Nx2N only, i.e. num_parts = 1
114 { 1, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4 } },
115 //noise = 1
116 { { 5, 7, 7, 7, 6, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8 },
117 { 5, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8 },
118 { 5, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8 },
119 { 5, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8 } } },
120
121 //grid flag = 1
122 { //noise = 0
123 { { 0, 2, 2, 2, 2, 2, 2, 2, 2, 4, 4, 4, 4, 4, 4, 4, 4 },
124 { 0, 2, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4 },
125 { 0, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4 },
126 { 0, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4 } },
127 //noise = 1
128 { { 0, 7, 7, 7, 6, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8 },
129 { 0, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8 },
130 { 0, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8 },
131 { 0, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8 } } }
132 };
133
134 /*****************************************************************************/
135 /* Function Definitions */
136 /*****************************************************************************/
137 /*!
138 ******************************************************************************
139 * \if Function name : ihevce_me_instr_set_router \endif
140 *
141 * \brief
142 * Function pointer initialization of me utils struct
143 *
144 *****************************************************************************
145 */
ihevce_me_instr_set_router(ihevce_me_optimised_function_list_t * ps_func_list,IV_ARCH_T e_arch)146 void ihevce_me_instr_set_router(ihevce_me_optimised_function_list_t *ps_func_list, IV_ARCH_T e_arch)
147 {
148 // clang-format off
149 #ifdef DISABLE_AVX2_INTR
150 e_arch = (e_arch == ARCH_X86_AVX2) ? ARCH_X86_AVX : e_arch;
151 #endif
152
153 switch(e_arch)
154 {
155 #ifdef ENABLE_NEON
156 case ARCH_ARM_A9Q:
157 case ARCH_ARM_V8_NEON:
158 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_8x8 = hme_calc_pt_sad_and_result_explicit;
159 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_8x8_4x4 = hme_calc_pt_sad_and_result_explicit;
160 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_8x8_for_grid = hme_calc_pt_sad_and_result_explicit;
161 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_generic = hme_calc_pt_sad_and_result_explicit;
162 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_8x8 = hme_calc_pt_sad_and_result_explicit;
163 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_8x8_4x4 = hme_calc_pt_sad_and_result_explicit;
164 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_8x8_for_grid = hme_calc_pt_sad_and_result_explicit;
165 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_generic = hme_calc_pt_sad_and_result_explicit;
166 ps_func_list->pf_calc_sad_and_1_best_result_generic = hme_calc_sad_and_1_best_result;
167 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_generic = hme_calc_stim_injected_sad_and_1_best_result;
168 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_eq_1 = hme_calc_stim_injected_sad_and_1_best_result;
169 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_square_parts = hme_calc_stim_injected_sad_and_1_best_result;
170 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_lt_9 = hme_calc_stim_injected_sad_and_1_best_result;
171 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_lt_17 = hme_calc_stim_injected_sad_and_1_best_result;
172 ps_func_list->pf_compute_variance_for_all_parts = hme_compute_variance_for_all_parts;
173 ps_func_list->pf_compute_stim_injected_distortion_for_all_parts = hme_compute_stim_injected_distortion_for_all_parts;
174 ps_func_list->pf_calc_sad_and_1_best_result_num_part_1_for_grid = hme_calc_sad_and_1_best_result_neon;
175 ps_func_list->pf_calc_sad_and_1_best_result_num_part_eq_1 = hme_calc_sad_and_1_best_result_neon;
176 ps_func_list->pf_calc_sad_and_1_best_result_num_part_lt_17 = hme_calc_sad_and_1_best_result_neon;
177 ps_func_list->pf_calc_sad_and_1_best_result_num_part_lt_9 = hme_calc_sad_and_1_best_result_neon;
178 ps_func_list->pf_calc_sad_and_1_best_result_num_square_parts = hme_calc_sad_and_1_best_result_neon;
179 ps_func_list->pf_calc_sad_and_1_best_result_subpel_generic = hme_calc_sad_and_1_best_result_subpel;
180 ps_func_list->pf_calc_sad_and_1_best_result_subpel_num_part_eq_1 = hme_calc_sad_and_1_best_result_subpel_neon;
181 ps_func_list->pf_calc_sad_and_1_best_result_subpel_num_part_lt_17 = hme_calc_sad_and_1_best_result_subpel_neon;
182 ps_func_list->pf_calc_sad_and_1_best_result_subpel_num_part_lt_9 = hme_calc_sad_and_1_best_result_subpel_neon;
183 ps_func_list->pf_calc_sad_and_1_best_result_subpel_square_parts = hme_calc_sad_and_1_best_result_subpel_neon;
184 ps_func_list->pf_combine_4x4_sads_and_compute_cost_high_quality = hme_combine_4x4_sads_and_compute_cost_high_quality_neon;
185 ps_func_list->pf_combine_4x4_sads_and_compute_cost_high_speed = hme_combine_4x4_sads_and_compute_cost_high_speed_neon;
186 ps_func_list->pf_compute_4x4_sads_for_16x16_blk = compute_4x4_sads_for_16x16_blk_neon;
187 ps_func_list->pf_evalsad_grid_npu_MxN = hme_evalsad_grid_npu_MxN_neon;
188 ps_func_list->pf_evalsad_grid_pu_MxM = compute_part_sads_for_MxM_blk_neon;
189 ps_func_list->pf_evalsad_pt_npu_12x16_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
190 ps_func_list->pf_evalsad_pt_npu_16x12_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
191 ps_func_list->pf_evalsad_pt_npu_16x4_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
192 ps_func_list->pf_evalsad_pt_npu_24x32_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
193 ps_func_list->pf_evalsad_pt_npu_8x4_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
194 ps_func_list->pf_evalsad_pt_npu_mxn_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
195 ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
196 ps_func_list->pf_evalsad_pt_npu_width_multiple_4_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
197 ps_func_list->pf_evalsad_pt_npu_width_multiple_8_8bit = hme_evalsad_pt_npu_MxN_8bit_neon;
198 ps_func_list->pf_get_wt_inp_8x8 = hme_get_wt_inp_8x8_neon;
199 ps_func_list->pf_get_wt_inp_ctb = hme_get_wt_inp_ctb_neon;
200 ps_func_list->pf_get_wt_inp_generic = hme_get_wt_inp;
201 ps_func_list->pf_mv_clipper = hme_mv_clipper;
202 ps_func_list->pf_qpel_interp_avg_1pt = hme_qpel_interp_avg_1pt_neon;
203 ps_func_list->pf_qpel_interp_avg_2pt_horz_with_reuse = hme_qpel_interp_avg_2pt_horz_with_reuse_neon;
204 ps_func_list->pf_qpel_interp_avg_2pt_vert_with_reuse = hme_qpel_interp_avg_2pt_vert_with_reuse_neon;
205 ps_func_list->pf_qpel_interp_avg_generic = hme_qpel_interp_avg_neon;
206 ps_func_list->pf_store_4x4_sads_high_quality = hme_store_4x4_sads_high_quality_neon;
207 ps_func_list->pf_store_4x4_sads_high_speed = hme_store_4x4_sads_high_speed_neon;
208 break;
209 #endif
210 default:
211 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_8x8 = hme_calc_pt_sad_and_result_explicit;
212 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_8x8_4x4 = hme_calc_pt_sad_and_result_explicit;
213 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_8x8_for_grid = hme_calc_pt_sad_and_result_explicit;
214 ps_func_list->pf_calc_pt_sad_and_1_best_result_explicit_generic = hme_calc_pt_sad_and_result_explicit;
215 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_8x8 = hme_calc_pt_sad_and_result_explicit;
216 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_8x8_4x4 = hme_calc_pt_sad_and_result_explicit;
217 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_8x8_for_grid = hme_calc_pt_sad_and_result_explicit;
218 ps_func_list->pf_calc_pt_sad_and_2_best_results_explicit_generic = hme_calc_pt_sad_and_result_explicit;
219 ps_func_list->pf_calc_sad_and_1_best_result_generic = hme_calc_sad_and_1_best_result;
220 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_generic = hme_calc_stim_injected_sad_and_1_best_result;
221 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_eq_1 = hme_calc_stim_injected_sad_and_1_best_result;
222 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_square_parts = hme_calc_stim_injected_sad_and_1_best_result;
223 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_lt_9 = hme_calc_stim_injected_sad_and_1_best_result;
224 ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_lt_17 = hme_calc_stim_injected_sad_and_1_best_result;
225 ps_func_list->pf_compute_variance_for_all_parts = hme_compute_variance_for_all_parts;
226 ps_func_list->pf_compute_stim_injected_distortion_for_all_parts = hme_compute_stim_injected_distortion_for_all_parts;
227 ps_func_list->pf_calc_sad_and_1_best_result_num_part_1_for_grid = hme_calc_sad_and_1_best_result;
228 ps_func_list->pf_calc_sad_and_1_best_result_num_part_eq_1 = hme_calc_sad_and_1_best_result;
229 ps_func_list->pf_calc_sad_and_1_best_result_num_part_lt_17 = hme_calc_sad_and_1_best_result;
230 ps_func_list->pf_calc_sad_and_1_best_result_num_part_lt_9 = hme_calc_sad_and_1_best_result;
231 ps_func_list->pf_calc_sad_and_1_best_result_num_square_parts = hme_calc_sad_and_1_best_result;
232 ps_func_list->pf_calc_sad_and_1_best_result_subpel_generic = hme_calc_sad_and_1_best_result_subpel;
233 ps_func_list->pf_calc_sad_and_1_best_result_subpel_num_part_eq_1 = hme_calc_sad_and_1_best_result_subpel;
234 ps_func_list->pf_calc_sad_and_1_best_result_subpel_num_part_lt_17 = hme_calc_sad_and_1_best_result_subpel;
235 ps_func_list->pf_calc_sad_and_1_best_result_subpel_num_part_lt_9 = hme_calc_sad_and_1_best_result_subpel;
236 ps_func_list->pf_calc_sad_and_1_best_result_subpel_square_parts = hme_calc_sad_and_1_best_result_subpel;
237 ps_func_list->pf_combine_4x4_sads_and_compute_cost_high_quality = hme_combine_4x4_sads_and_compute_cost_high_quality;
238 ps_func_list->pf_combine_4x4_sads_and_compute_cost_high_speed = hme_combine_4x4_sads_and_compute_cost_high_speed;
239 ps_func_list->pf_compute_4x4_sads_for_16x16_blk = compute_4x4_sads_for_16x16_blk;
240 ps_func_list->pf_evalsad_grid_npu_MxN = hme_evalsad_grid_npu_MxN;
241 ps_func_list->pf_evalsad_grid_pu_MxM = compute_part_sads_for_MxM_blk;
242 ps_func_list->pf_evalsad_pt_npu_12x16_8bit = hme_evalsad_pt_npu_MxN_8bit;
243 ps_func_list->pf_evalsad_pt_npu_16x12_8bit = hme_evalsad_pt_npu_MxN_8bit;
244 ps_func_list->pf_evalsad_pt_npu_16x4_8bit = hme_evalsad_pt_npu_MxN_8bit;
245 ps_func_list->pf_evalsad_pt_npu_24x32_8bit = hme_evalsad_pt_npu_MxN_8bit;
246 ps_func_list->pf_evalsad_pt_npu_8x4_8bit = hme_evalsad_pt_npu_MxN_8bit;
247 ps_func_list->pf_evalsad_pt_npu_mxn_8bit = hme_evalsad_pt_npu_MxN_8bit;
248 ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit = hme_evalsad_pt_npu_MxN_8bit;
249 ps_func_list->pf_evalsad_pt_npu_width_multiple_4_8bit = hme_evalsad_pt_npu_MxN_8bit;
250 ps_func_list->pf_evalsad_pt_npu_width_multiple_8_8bit = hme_evalsad_pt_npu_MxN_8bit;
251 ps_func_list->pf_get_wt_inp_8x8 = hme_get_wt_inp;
252 ps_func_list->pf_get_wt_inp_ctb = hme_get_wt_inp;
253 ps_func_list->pf_get_wt_inp_generic = hme_get_wt_inp;
254 ps_func_list->pf_mv_clipper = hme_mv_clipper;
255 ps_func_list->pf_qpel_interp_avg_1pt = hme_qpel_interp_avg_1pt;
256 ps_func_list->pf_qpel_interp_avg_2pt_horz_with_reuse = hme_qpel_interp_avg_2pt_horz_with_reuse;
257 ps_func_list->pf_qpel_interp_avg_2pt_vert_with_reuse = hme_qpel_interp_avg_2pt_vert_with_reuse;
258 ps_func_list->pf_qpel_interp_avg_generic = hme_qpel_interp_avg;
259 ps_func_list->pf_store_4x4_sads_high_quality = hme_store_4x4_sads_high_quality;
260 ps_func_list->pf_store_4x4_sads_high_speed = hme_store_4x4_sads_high_speed;
261 break;
262 }
263
264 gapf_sad_pt_npu[BLK_4x4] = ps_func_list->pf_evalsad_pt_npu_width_multiple_4_8bit;
265 gapf_sad_pt_npu[BLK_4x8] = ps_func_list->pf_evalsad_pt_npu_width_multiple_4_8bit;
266 gapf_sad_pt_npu[BLK_8x4] = ps_func_list->pf_evalsad_pt_npu_8x4_8bit;
267 gapf_sad_pt_npu[BLK_8x8] = ps_func_list->pf_evalsad_pt_npu_width_multiple_8_8bit;
268 gapf_sad_pt_npu[BLK_4x16] = ps_func_list->pf_evalsad_pt_npu_width_multiple_4_8bit;
269 gapf_sad_pt_npu[BLK_8x16] = ps_func_list->pf_evalsad_pt_npu_width_multiple_8_8bit;
270 gapf_sad_pt_npu[BLK_12x16] = ps_func_list->pf_evalsad_pt_npu_12x16_8bit;
271 gapf_sad_pt_npu[BLK_16x4] = ps_func_list->pf_evalsad_pt_npu_16x4_8bit;
272 gapf_sad_pt_npu[BLK_16x8] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
273 gapf_sad_pt_npu[BLK_16x12] = ps_func_list->pf_evalsad_pt_npu_16x12_8bit;
274 gapf_sad_pt_npu[BLK_16x16] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
275 gapf_sad_pt_npu[BLK_8x32] = ps_func_list->pf_evalsad_pt_npu_width_multiple_8_8bit;
276 gapf_sad_pt_npu[BLK_16x32] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
277 gapf_sad_pt_npu[BLK_24x32] = ps_func_list->pf_evalsad_pt_npu_24x32_8bit;
278 gapf_sad_pt_npu[BLK_32x8] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
279 gapf_sad_pt_npu[BLK_32x16] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
280 gapf_sad_pt_npu[BLK_32x8] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
281 gapf_sad_pt_npu[BLK_32x16] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
282 gapf_sad_pt_npu[BLK_32x24] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
283 gapf_sad_pt_npu[BLK_32x32] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
284 gapf_sad_pt_npu[BLK_16x64] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
285 gapf_sad_pt_npu[BLK_32x64] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
286 gapf_sad_pt_npu[BLK_48x64] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
287 gapf_sad_pt_npu[BLK_64x16] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
288 gapf_sad_pt_npu[BLK_64x32] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
289 gapf_sad_pt_npu[BLK_64x48] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
290 gapf_sad_pt_npu[BLK_64x64] = ps_func_list->pf_evalsad_pt_npu_width_multiple_16_8bit;
291
292 gpf_part_sads_evaluator_16x16CU = ps_func_list->pf_compute_4x4_sads_for_16x16_blk;
293 gpf_part_sads_evaluator_MxM = ps_func_list->pf_evalsad_grid_pu_MxM;
294
295 gpf_sad_grid_mxn = ps_func_list->pf_evalsad_grid_npu_MxN;
296
297 gapf_calc_sad_and_result_fxn[0][0] = ps_func_list->pf_calc_sad_and_1_best_result_num_part_1_for_grid;
298 gapf_calc_sad_and_result_fxn[1][0] = ps_func_list->pf_calc_sad_and_1_best_result_num_part_eq_1;
299 gapf_calc_sad_and_result_fxn[2][0] = ps_func_list->pf_calc_sad_and_1_best_result_num_square_parts;
300 gapf_calc_sad_and_result_fxn[3][0] = ps_func_list->pf_calc_sad_and_1_best_result_num_part_lt_9;
301 gapf_calc_sad_and_result_fxn[4][0] = ps_func_list->pf_calc_sad_and_1_best_result_num_part_lt_17;
302 gapf_calc_sad_and_result_fxn[5][0] = ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_eq_1;
303 gapf_calc_sad_and_result_fxn[6][0] = ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_square_parts;
304 gapf_calc_sad_and_result_fxn[7][0] = ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_lt_9;
305 gapf_calc_sad_and_result_fxn[8][0] = ps_func_list->pf_calc_stim_injected_sad_and_1_best_result_num_part_lt_17;
306 gapf_calc_sad_and_result_fxn[0][1] = ps_func_list->pf_calc_sad_and_2_best_results_num_part_1_for_grid;
307 gapf_calc_sad_and_result_fxn[1][1] = ps_func_list->pf_calc_sad_and_2_best_results_num_part_eq_1;
308 gapf_calc_sad_and_result_fxn[2][1] = ps_func_list->pf_calc_sad_and_2_best_results_num_square_parts;
309 gapf_calc_sad_and_result_fxn[3][1] = ps_func_list->pf_calc_sad_and_2_best_results_num_part_lt_9;
310 gapf_calc_sad_and_result_fxn[4][1] = ps_func_list->pf_calc_sad_and_2_best_results_num_part_lt_17;
311 gapf_calc_sad_and_result_fxn[5][1] = ps_func_list->pf_calc_stim_injected_sad_and_2_best_results_num_part_eq_1;
312 gapf_calc_sad_and_result_fxn[6][1] = ps_func_list->pf_calc_stim_injected_sad_and_2_best_results_num_square_parts;
313 gapf_calc_sad_and_result_fxn[7][1] = ps_func_list->pf_calc_stim_injected_sad_and_2_best_results_num_part_lt_9;
314 gapf_calc_sad_and_result_fxn[8][1] = ps_func_list->pf_calc_stim_injected_sad_and_2_best_results_num_part_lt_17;
315 }
316 // clang-format on
317
hme_get_calc_sad_and_result_fxn(S08 i1_grid_flag,U08 u1_is_cu_noisy,S32 i4_part_mask,S32 num_parts,S32 num_results)318 FT_CALC_SAD_AND_RESULT *hme_get_calc_sad_and_result_fxn(
319 S08 i1_grid_flag, U08 u1_is_cu_noisy, S32 i4_part_mask, S32 num_parts, S32 num_results)
320 {
321 U08 u1_index;
322
323 ASSERT((1 == num_results) || (2 == num_results));
324
325 u1_index =
326 gau1_calc_sad_and_result[i1_grid_flag][u1_is_cu_noisy]
327 [(!!(i4_part_mask & (ENABLE_SMP | ENABLE_NxN)) &&
328 !(i4_part_mask & ENABLE_AMP))
329 ? (!!(i4_part_mask & ENABLE_NxN) ? 0 : 1)
330 : (!!(i4_part_mask & ENABLE_AMP) ? 2 : 3)][num_parts - 1];
331
332 return gapf_calc_sad_and_result_fxn[u1_index][2 == num_results];
333 }
334
hme_evalsad_grid_pu_MxM(err_prms_t * ps_prms)335 void hme_evalsad_grid_pu_MxM(err_prms_t *ps_prms)
336 {
337 grid_ctxt_t s_grid;
338 cand_t as_candt[9];
339
340 S32 *api4_sad_grid[TOT_NUM_PARTS];
341
342 hme_mv_t s_mv = { 0, 0 };
343
344 CU_SIZE_T e_cu_size = (CU_SIZE_T)(hme_get_range(ps_prms->i4_blk_wd) - 4);
345
346 S32 i4_ref_idx = 0, i;
347 S32 num_candts = 0;
348
349 s_grid.num_grids = 1;
350 s_grid.ref_buf_stride = ps_prms->i4_ref_stride;
351 s_grid.grd_sz_y_x = ((ps_prms->i4_step << 16) | ps_prms->i4_step);
352 s_grid.ppu1_ref_ptr = &ps_prms->pu1_ref;
353 s_grid.pi4_grd_mask = &ps_prms->i4_grid_mask;
354 s_grid.p_mv = &s_mv;
355 s_grid.p_ref_idx = &i4_ref_idx;
356
357 for(i = 0; i < 9; i++)
358 {
359 if(s_grid.pi4_grd_mask[0] & (1 << i))
360 {
361 num_candts++;
362 }
363 }
364
365 for(i = 0; i < TOT_NUM_PARTS; i++)
366 {
367 api4_sad_grid[i] = &ps_prms->pi4_sad_grid[i * num_candts];
368 }
369
370 gpf_part_sads_evaluator_MxM(
371 &s_grid,
372 ps_prms->pu1_inp,
373 ps_prms->i4_inp_stride,
374 (WORD32 **)api4_sad_grid,
375 as_candt,
376 &num_candts,
377 e_cu_size);
378 }
379
hme_get_sad_fxn(BLK_SIZE_T e_blk_size,S32 i4_grid_mask,S32 i4_part_mask)380 PF_SAD_FXN_T hme_get_sad_fxn(BLK_SIZE_T e_blk_size, S32 i4_grid_mask, S32 i4_part_mask)
381 {
382 S32 i4_grid_en = ((i4_grid_mask & 0x1fe) != 0);
383
384 if(i4_grid_en)
385 {
386 if(i4_part_mask & (i4_part_mask - 1))
387 {
388 if(BLK_16x16 == e_blk_size)
389 {
390 return hme_evalsad_grid_pu_16x16;
391 }
392 else
393 {
394 return hme_evalsad_grid_pu_MxM;
395 }
396 }
397 else
398 {
399 return gpf_sad_grid_mxn;
400 }
401 }
402 else
403 {
404 if(i4_part_mask & (i4_part_mask - 1))
405 {
406 if(BLK_16x16 == e_blk_size)
407 {
408 return hme_evalsad_grid_pu_16x16;
409 }
410 else
411 {
412 return hme_evalsad_grid_pu_MxM;
413 }
414 }
415 else
416 {
417 return gapf_sad_pt_npu[e_blk_size];
418 }
419 }
420 }
421
ihevce_sifter_sad_fxn_assigner(FT_SAD_EVALUATOR ** ppf_evalsad_pt_npu_mxn,IV_ARCH_T e_arch)422 void ihevce_sifter_sad_fxn_assigner(FT_SAD_EVALUATOR **ppf_evalsad_pt_npu_mxn, IV_ARCH_T e_arch)
423 {
424 switch(e_arch)
425 {
426 #ifdef ENABLE_NEON
427 case ARCH_ARM_A9Q:
428 case ARCH_ARM_V8_NEON:
429 ppf_evalsad_pt_npu_mxn[0] = hme_evalsad_pt_npu_MxN_8bit_neon;
430 break;
431 #endif
432
433 default:
434 ppf_evalsad_pt_npu_mxn[0] = hme_evalsad_pt_npu_MxN_8bit;
435 break;
436 }
437 }
438