1 /* 2 ** Copyright (c) 2019-2021 The Khronos Group Inc. 3 ** 4 ** SPDX-License-Identifier: Apache-2.0 5 */ 6 7 #ifndef VULKAN_VIDEO_CODEC_H264STD_H_ 8 #define VULKAN_VIDEO_CODEC_H264STD_H_ 1 9 10 #ifdef __cplusplus 11 extern "C" { 12 #endif 13 14 #include "vk_video/vulkan_video_codecs_common.h" 15 16 // Vulkan 0.9 provisional Vulkan video H.264 encode and decode std specification version number 17 #define VK_STD_VULKAN_VIDEO_CODEC_H264_API_VERSION_0_9_5 VK_MAKE_VIDEO_STD_VERSION(0, 9, 5) // Patch version should always be set to 0 18 19 // Format must be in the form XX.XX where the first two digits are the major and the second two, the minor. 20 #define VK_STD_VULKAN_VIDEO_CODEC_H264_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H264_API_VERSION_0_9_5 21 #define VK_STD_VULKAN_VIDEO_CODEC_H264_EXTENSION_NAME "VK_STD_vulkan_video_codec_h264" 22 23 // ************************************************* 24 // Video H.264 common definitions: 25 // ************************************************* 26 27 #define STD_VIDEO_H264_CPB_CNT_LIST_SIZE 32 28 #define STD_VIDEO_H264_SCALING_LIST_4X4_NUM_LISTS 6 29 #define STD_VIDEO_H264_SCALING_LIST_4X4_NUM_ELEMENTS 16 30 #define STD_VIDEO_H264_SCALING_LIST_8X8_NUM_LISTS 2 31 #define STD_VIDEO_H264_SCALING_LIST_8X8_NUM_ELEMENTS 64 32 33 typedef enum StdVideoH264ChromaFormatIdc { 34 STD_VIDEO_H264_CHROMA_FORMAT_IDC_MONOCHROME = 0, 35 STD_VIDEO_H264_CHROMA_FORMAT_IDC_420 = 1, 36 STD_VIDEO_H264_CHROMA_FORMAT_IDC_422 = 2, 37 STD_VIDEO_H264_CHROMA_FORMAT_IDC_444 = 3, 38 STD_VIDEO_H264_CHROMA_FORMAT_IDC_INVALID = 0x7FFFFFFF 39 } StdVideoH264ChromaFormatIdc; 40 41 typedef enum StdVideoH264ProfileIdc { 42 STD_VIDEO_H264_PROFILE_IDC_BASELINE = 66, /* Only constrained baseline is supported */ 43 STD_VIDEO_H264_PROFILE_IDC_MAIN = 77, 44 STD_VIDEO_H264_PROFILE_IDC_HIGH = 100, 45 STD_VIDEO_H264_PROFILE_IDC_HIGH_444_PREDICTIVE = 244, 46 STD_VIDEO_H264_PROFILE_IDC_INVALID = 0x7FFFFFFF 47 } StdVideoH264ProfileIdc; 48 49 typedef enum StdVideoH264Level { 50 STD_VIDEO_H264_LEVEL_1_0 = 0, 51 STD_VIDEO_H264_LEVEL_1_1 = 1, 52 STD_VIDEO_H264_LEVEL_1_2 = 2, 53 STD_VIDEO_H264_LEVEL_1_3 = 3, 54 STD_VIDEO_H264_LEVEL_2_0 = 4, 55 STD_VIDEO_H264_LEVEL_2_1 = 5, 56 STD_VIDEO_H264_LEVEL_2_2 = 6, 57 STD_VIDEO_H264_LEVEL_3_0 = 7, 58 STD_VIDEO_H264_LEVEL_3_1 = 8, 59 STD_VIDEO_H264_LEVEL_3_2 = 9, 60 STD_VIDEO_H264_LEVEL_4_0 = 10, 61 STD_VIDEO_H264_LEVEL_4_1 = 11, 62 STD_VIDEO_H264_LEVEL_4_2 = 12, 63 STD_VIDEO_H264_LEVEL_5_0 = 13, 64 STD_VIDEO_H264_LEVEL_5_1 = 14, 65 STD_VIDEO_H264_LEVEL_5_2 = 15, 66 STD_VIDEO_H264_LEVEL_6_0 = 16, 67 STD_VIDEO_H264_LEVEL_6_1 = 17, 68 STD_VIDEO_H264_LEVEL_6_2 = 18, 69 STD_VIDEO_H264_LEVEL_INVALID = 0x7FFFFFFF 70 } StdVideoH264Level; 71 72 typedef enum StdVideoH264PocType { 73 STD_VIDEO_H264_POC_TYPE_0 = 0, 74 STD_VIDEO_H264_POC_TYPE_1 = 1, 75 STD_VIDEO_H264_POC_TYPE_2 = 2, 76 STD_VIDEO_H264_POC_TYPE_INVALID = 0x7FFFFFFF 77 } StdVideoH264PocType; 78 79 typedef enum StdVideoH264AspectRatioIdc { 80 STD_VIDEO_H264_ASPECT_RATIO_IDC_UNSPECIFIED = 0, 81 STD_VIDEO_H264_ASPECT_RATIO_IDC_SQUARE = 1, 82 STD_VIDEO_H264_ASPECT_RATIO_IDC_12_11 = 2, 83 STD_VIDEO_H264_ASPECT_RATIO_IDC_10_11 = 3, 84 STD_VIDEO_H264_ASPECT_RATIO_IDC_16_11 = 4, 85 STD_VIDEO_H264_ASPECT_RATIO_IDC_40_33 = 5, 86 STD_VIDEO_H264_ASPECT_RATIO_IDC_24_11 = 6, 87 STD_VIDEO_H264_ASPECT_RATIO_IDC_20_11 = 7, 88 STD_VIDEO_H264_ASPECT_RATIO_IDC_32_11 = 8, 89 STD_VIDEO_H264_ASPECT_RATIO_IDC_80_33 = 9, 90 STD_VIDEO_H264_ASPECT_RATIO_IDC_18_11 = 10, 91 STD_VIDEO_H264_ASPECT_RATIO_IDC_15_11 = 11, 92 STD_VIDEO_H264_ASPECT_RATIO_IDC_64_33 = 12, 93 STD_VIDEO_H264_ASPECT_RATIO_IDC_160_99 = 13, 94 STD_VIDEO_H264_ASPECT_RATIO_IDC_4_3 = 14, 95 STD_VIDEO_H264_ASPECT_RATIO_IDC_3_2 = 15, 96 STD_VIDEO_H264_ASPECT_RATIO_IDC_2_1 = 16, 97 STD_VIDEO_H264_ASPECT_RATIO_IDC_EXTENDED_SAR = 255, 98 STD_VIDEO_H264_ASPECT_RATIO_IDC_INVALID = 0x7FFFFFFF 99 } StdVideoH264AspectRatioIdc; 100 101 typedef enum StdVideoH264WeightedBipredIdc { 102 STD_VIDEO_H264_WEIGHTED_BIPRED_IDC_DEFAULT = 0, 103 STD_VIDEO_H264_WEIGHTED_BIPRED_IDC_EXPLICIT = 1, 104 STD_VIDEO_H264_WEIGHTED_BIPRED_IDC_IMPLICIT = 2, 105 STD_VIDEO_H264_WEIGHTED_BIPRED_IDC_INVALID = 0x7FFFFFFF 106 } StdVideoH264WeightedBipredIdc; 107 108 typedef enum StdVideoH264ModificationOfPicNumsIdc { 109 STD_VIDEO_H264_MODIFICATION_OF_PIC_NUMS_IDC_SHORT_TERM_SUBTRACT = 0, 110 STD_VIDEO_H264_MODIFICATION_OF_PIC_NUMS_IDC_SHORT_TERM_ADD = 1, 111 STD_VIDEO_H264_MODIFICATION_OF_PIC_NUMS_IDC_LONG_TERM = 2, 112 STD_VIDEO_H264_MODIFICATION_OF_PIC_NUMS_IDC_END = 3, 113 STD_VIDEO_H264_MODIFICATION_OF_PIC_NUMS_IDC_INVALID = 0x7FFFFFFF 114 } StdVideoH264ModificationOfPicNumsIdc; 115 116 typedef enum StdVideoH264MemMgmtControlOp { 117 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_END = 0, 118 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_UNMARK_SHORT_TERM = 1, 119 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_UNMARK_LONG_TERM = 2, 120 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_MARK_LONG_TERM = 3, 121 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_SET_MAX_LONG_TERM_INDEX = 4, 122 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_UNMARK_ALL = 5, 123 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_MARK_CURRENT_AS_LONG_TERM = 6, 124 STD_VIDEO_H264_MEM_MGMT_CONTROL_OP_INVALID = 0x7FFFFFFF 125 } StdVideoH264MemMgmtControlOp; 126 127 typedef enum StdVideoH264CabacInitIdc { 128 STD_VIDEO_H264_CABAC_INIT_IDC_0 = 0, 129 STD_VIDEO_H264_CABAC_INIT_IDC_1 = 1, 130 STD_VIDEO_H264_CABAC_INIT_IDC_2 = 2, 131 STD_VIDEO_H264_CABAC_INIT_IDC_INVALID = 0x7FFFFFFF 132 } StdVideoH264CabacInitIdc; 133 134 typedef enum StdVideoH264DisableDeblockingFilterIdc { 135 STD_VIDEO_H264_DISABLE_DEBLOCKING_FILTER_IDC_DISABLED = 0, 136 STD_VIDEO_H264_DISABLE_DEBLOCKING_FILTER_IDC_ENABLED = 1, 137 STD_VIDEO_H264_DISABLE_DEBLOCKING_FILTER_IDC_PARTIAL = 2, 138 STD_VIDEO_H264_DISABLE_DEBLOCKING_FILTER_IDC_INVALID = 0x7FFFFFFF 139 } StdVideoH264DisableDeblockingFilterIdc; 140 141 typedef enum StdVideoH264SliceType { 142 STD_VIDEO_H264_SLICE_TYPE_P = 0, 143 STD_VIDEO_H264_SLICE_TYPE_B = 1, 144 STD_VIDEO_H264_SLICE_TYPE_I = 2, 145 // reserved STD_VIDEO_H264_SLICE_TYPE_SP = 3, 146 // reserved STD_VIDEO_H264_SLICE_TYPE_SI = 4, 147 STD_VIDEO_H264_SLICE_TYPE_INVALID = 0x7FFFFFFF 148 } StdVideoH264SliceType; 149 150 typedef enum StdVideoH264PictureType { 151 STD_VIDEO_H264_PICTURE_TYPE_P = 0, 152 STD_VIDEO_H264_PICTURE_TYPE_B = 1, 153 STD_VIDEO_H264_PICTURE_TYPE_I = 2, 154 // reserved STD_VIDEO_H264_PICTURE_TYPE_SP = 3, 155 // reserved STD_VIDEO_H264_PICTURE_TYPE_SI = 4, 156 STD_VIDEO_H264_PICTURE_TYPE_IDR = 5, 157 STD_VIDEO_H264_PICTURE_TYPE_INVALID = 0x7FFFFFFF 158 } StdVideoH264PictureType; 159 160 typedef enum StdVideoH264NonVclNaluType { 161 STD_VIDEO_H264_NON_VCL_NALU_TYPE_SPS = 0, 162 STD_VIDEO_H264_NON_VCL_NALU_TYPE_PPS = 1, 163 STD_VIDEO_H264_NON_VCL_NALU_TYPE_AUD = 2, 164 STD_VIDEO_H264_NON_VCL_NALU_TYPE_PREFIX = 3, 165 STD_VIDEO_H264_NON_VCL_NALU_TYPE_END_OF_SEQUENCE = 4, 166 STD_VIDEO_H264_NON_VCL_NALU_TYPE_END_OF_STREAM = 5, 167 STD_VIDEO_H264_NON_VCL_NALU_TYPE_PRECODED = 6, 168 STD_VIDEO_H264_NON_VCL_NALU_TYPE_INVALID = 0x7FFFFFFF 169 } StdVideoH264NonVclNaluType; 170 171 typedef struct StdVideoH264SpsVuiFlags { 172 uint32_t aspect_ratio_info_present_flag : 1; 173 uint32_t overscan_info_present_flag : 1; 174 uint32_t overscan_appropriate_flag : 1; 175 uint32_t video_signal_type_present_flag : 1; 176 uint32_t video_full_range_flag : 1; 177 uint32_t color_description_present_flag : 1; 178 uint32_t chroma_loc_info_present_flag : 1; 179 uint32_t timing_info_present_flag : 1; 180 uint32_t fixed_frame_rate_flag : 1; 181 uint32_t bitstream_restriction_flag : 1; 182 uint32_t nal_hrd_parameters_present_flag : 1; 183 uint32_t vcl_hrd_parameters_present_flag : 1; 184 } StdVideoH264SpsVuiFlags; 185 186 typedef struct StdVideoH264HrdParameters { // hrd_parameters 187 uint8_t cpb_cnt_minus1; 188 uint8_t bit_rate_scale; 189 uint8_t cpb_size_scale; 190 uint32_t bit_rate_value_minus1[STD_VIDEO_H264_CPB_CNT_LIST_SIZE]; // cpb_cnt_minus1 number of valid elements 191 uint32_t cpb_size_value_minus1[STD_VIDEO_H264_CPB_CNT_LIST_SIZE]; // cpb_cnt_minus1 number of valid elements 192 uint8_t cbr_flag[STD_VIDEO_H264_CPB_CNT_LIST_SIZE]; // cpb_cnt_minus1 number of valid elements 193 uint32_t initial_cpb_removal_delay_length_minus1; 194 uint32_t cpb_removal_delay_length_minus1; 195 uint32_t dpb_output_delay_length_minus1; 196 uint32_t time_offset_length; 197 } StdVideoH264HrdParameters; 198 199 typedef struct StdVideoH264SequenceParameterSetVui { 200 StdVideoH264AspectRatioIdc aspect_ratio_idc; 201 uint16_t sar_width; 202 uint16_t sar_height; 203 uint8_t video_format; 204 uint8_t color_primaries; 205 uint8_t transfer_characteristics; 206 uint8_t matrix_coefficients; 207 uint32_t num_units_in_tick; 208 uint32_t time_scale; 209 StdVideoH264HrdParameters* pHrdParameters; // must be a valid ptr to hrd_parameters, if nal_hrd_parameters_present_flag or vcl_hrd_parameters_present_flag are set 210 uint8_t max_num_reorder_frames; 211 uint8_t max_dec_frame_buffering; 212 StdVideoH264SpsVuiFlags flags; 213 } StdVideoH264SequenceParameterSetVui; 214 215 typedef struct StdVideoH264SpsFlags { 216 uint32_t constraint_set0_flag : 1; 217 uint32_t constraint_set1_flag : 1; 218 uint32_t constraint_set2_flag : 1; 219 uint32_t constraint_set3_flag : 1; 220 uint32_t constraint_set4_flag : 1; 221 uint32_t constraint_set5_flag : 1; 222 uint32_t direct_8x8_inference_flag : 1; 223 uint32_t mb_adaptive_frame_field_flag : 1; 224 uint32_t frame_mbs_only_flag : 1; 225 uint32_t delta_pic_order_always_zero_flag : 1; 226 uint32_t separate_colour_plane_flag : 1; 227 uint32_t gaps_in_frame_num_value_allowed_flag : 1; 228 uint32_t qpprime_y_zero_transform_bypass_flag : 1; 229 uint32_t frame_cropping_flag : 1; 230 uint32_t seq_scaling_matrix_present_flag : 1; 231 uint32_t vui_parameters_present_flag : 1; 232 } StdVideoH264SpsFlags; 233 234 typedef struct StdVideoH264ScalingLists 235 { 236 // scaling_list_present_mask has one bit for each 237 // seq_scaling_list_present_flag[i] for SPS OR 238 // pic_scaling_list_present_flag[i] for PPS, 239 // bit 0 - 5 are for each entry of ScalingList4x4 240 // bit 6 - 7 are for each entry plus 6 for ScalingList8x8 241 uint8_t scaling_list_present_mask; 242 // use_default_scaling_matrix_mask has one bit for each 243 // UseDefaultScalingMatrix4x4Flag[ i ] and 244 // UseDefaultScalingMatrix8x8Flag[ i - 6 ] for SPS OR PPS 245 // bit 0 - 5 are for each entry of ScalingList4x4 246 // bit 6 - 7 are for each entry plus 6 for ScalingList8x8 247 uint8_t use_default_scaling_matrix_mask; 248 uint8_t ScalingList4x4[STD_VIDEO_H264_SCALING_LIST_4X4_NUM_LISTS][STD_VIDEO_H264_SCALING_LIST_4X4_NUM_ELEMENTS]; 249 uint8_t ScalingList8x8[STD_VIDEO_H264_SCALING_LIST_8X8_NUM_LISTS][STD_VIDEO_H264_SCALING_LIST_8X8_NUM_ELEMENTS]; 250 } StdVideoH264ScalingLists; 251 252 typedef struct StdVideoH264SequenceParameterSet 253 { 254 StdVideoH264ProfileIdc profile_idc; 255 StdVideoH264Level level_idc; 256 uint8_t seq_parameter_set_id; 257 StdVideoH264ChromaFormatIdc chroma_format_idc; 258 uint8_t bit_depth_luma_minus8; 259 uint8_t bit_depth_chroma_minus8; 260 uint8_t log2_max_frame_num_minus4; 261 StdVideoH264PocType pic_order_cnt_type; 262 uint8_t log2_max_pic_order_cnt_lsb_minus4; 263 int32_t offset_for_non_ref_pic; 264 int32_t offset_for_top_to_bottom_field; 265 uint8_t num_ref_frames_in_pic_order_cnt_cycle; 266 uint8_t max_num_ref_frames; 267 uint32_t pic_width_in_mbs_minus1; 268 uint32_t pic_height_in_map_units_minus1; 269 uint32_t frame_crop_left_offset; 270 uint32_t frame_crop_right_offset; 271 uint32_t frame_crop_top_offset; 272 uint32_t frame_crop_bottom_offset; 273 StdVideoH264SpsFlags flags; 274 // pOffsetForRefFrame is a pointer representing the offset_for_ref_frame array with num_ref_frames_in_pic_order_cnt_cycle number of elements 275 // If pOffsetForRefFrame has nullptr value, then num_ref_frames_in_pic_order_cnt_cycle must also be "0". 276 int32_t* pOffsetForRefFrame; 277 StdVideoH264ScalingLists* pScalingLists; // Must be a valid pointer if seq_scaling_matrix_present_flag is set 278 StdVideoH264SequenceParameterSetVui* pSequenceParameterSetVui; // Must be a valid pointer if StdVideoH264SpsFlags:vui_parameters_present_flag is set 279 } StdVideoH264SequenceParameterSet; 280 281 typedef struct StdVideoH264PpsFlags { 282 uint32_t transform_8x8_mode_flag : 1; 283 uint32_t redundant_pic_cnt_present_flag : 1; 284 uint32_t constrained_intra_pred_flag : 1; 285 uint32_t deblocking_filter_control_present_flag : 1; 286 uint32_t weighted_bipred_idc_flag : 1; 287 uint32_t weighted_pred_flag : 1; 288 uint32_t pic_order_present_flag : 1; 289 uint32_t entropy_coding_mode_flag : 1; 290 uint32_t pic_scaling_matrix_present_flag : 1; 291 } StdVideoH264PpsFlags; 292 293 typedef struct StdVideoH264PictureParameterSet 294 { 295 uint8_t seq_parameter_set_id; 296 uint8_t pic_parameter_set_id; 297 uint8_t num_ref_idx_l0_default_active_minus1; 298 uint8_t num_ref_idx_l1_default_active_minus1; 299 StdVideoH264WeightedBipredIdc weighted_bipred_idc; 300 int8_t pic_init_qp_minus26; 301 int8_t pic_init_qs_minus26; 302 int8_t chroma_qp_index_offset; 303 int8_t second_chroma_qp_index_offset; 304 StdVideoH264PpsFlags flags; 305 StdVideoH264ScalingLists* pScalingLists; // Must be a valid pointer if StdVideoH264PpsFlags::pic_scaling_matrix_present_flag is set. 306 } StdVideoH264PictureParameterSet; 307 308 #ifdef __cplusplus 309 } 310 #endif 311 312 #endif // VULKAN_VIDEO_CODEC_H264STD_H_ 313