/*************************************************************************** * * Copyright 2015-2019 BES. * All rights reserved. All unpublished rights reserved. * * No part of this work may be used or reproduced in any form or by any * means, or stored in a database or retrieval system, without prior written * permission of BES. * * Use of this work is governed by a license granted by BES. * This work contains confidential and proprietary information of * BES. which is protected by copyright, trade secret, * trademark and other intellectual property rights. * ****************************************************************************/ // Standard C Included Files #include "a2dp_decoder_internal.h" #include "app_audio.h" #include "cmsis.h" #include "hal_location.h" #include "heap_api.h" #include "plat_types.h" #include #if defined(A2DP_SCALABLE_ON) #include "heap_api.h" #include "ssc.h" #define SCALABLE_MTU_LIMITER (32) #define SCALABLE_MEMPOOL_SIZE 1024 #define SCALABLE_OUTPUT_FRAME_SAMPLES (SCALABLE_FRAME_SIZE) typedef void *HANDLE_DECODER; static A2DP_AUDIO_CONTEXT_T *a2dp_audio_context_p = NULL; static A2DP_AUDIO_OUTPUT_CONFIG_T output_config; static unsigned char *scalable_decoder_place = NULL; static unsigned char *scalable_decoder_temp_buf = NULL; static short ss_pcm_buff[SCALABLE_FRAME_SIZE * 4]; static int scalable_uhq_flag __attribute__((unused)) = 0; static HANDLE_DECODER scalableDec_handle = NULL; static A2DP_AUDIO_DECODER_LASTFRAME_INFO_T lastframe_info; typedef struct { uint16_t sequenceNumber; uint32_t timestamp; uint8_t *buffer; int buffer_len; } a2dp_audio_scalable_decoder_frame_t; static void ss_to_24bit_buf(int32_t *out, int32_t *in, int size) { for (int i = 0; i < size; i++) { out[i] = in[i]; } } static void a2dp_audio_scalable_decoder_init(void) { if (scalableDec_handle == NULL) { scalableDec_handle = scalable_decoder_place; ssc_decoder_init(scalableDec_handle, output_config.num_channels, output_config.sample_rate); } } static int scalableDecoder_Close(HANDLE_DECODER handle) { if (handle) { a2dp_audio_heap_free(handle); a2dp_audio_heap_free(scalable_decoder_temp_buf); } return 0; } static void a2dp_audio_scalable_decoder_deinit(void) { if (scalableDec_handle) { scalableDecoder_Close(scalableDec_handle); scalableDec_handle = NULL; } } static void a2dp_audio_scalable_decoder_reinit(void) { if (scalableDec_handle) { a2dp_audio_scalable_decoder_deinit(); } a2dp_audio_scalable_decoder_init(); } static bool is_valid_frame(a2dp_audio_scalable_decoder_frame_t *decoder_frame_p) { int hw_tmp, len, bitrate_bps, frame_len, frame_size; int sampling_rate = 44100; unsigned char *input_buf = decoder_frame_p->buffer; if (decoder_frame_p->buffer_len < SCALABLE_HEAD_SIZE) { TRACE_A2DP_DECODER_E( "invalid scalable a2dp frame, length < SCALABLE_HEAD_SIZE !!!!!!!"); return false; } scalable_uhq_flag = 0; switch ((input_buf[3] & 0xf7)) { case 0xF0: bitrate_bps = 88000; break; case 0xF1: bitrate_bps = 96000; break; case 0xF2: bitrate_bps = 128000; break; case 0xF3: bitrate_bps = 192000; break; case 0xF4: bitrate_bps = 229000; break; case 0xF5: scalable_uhq_flag = 1; bitrate_bps = 328000; sampling_rate = 96000; break; default: bitrate_bps = 192000; break; } frame_size = SCALABLE_FRAME_SIZE; len = bitrate_bps * frame_size / sampling_rate / 8; if (scalable_uhq_flag == 0) { hw_tmp = (len * 3) >> 7; len = hw_tmp + len; len = len + ((len & 1) ^ 1); } else { len = 369; // 744/2-4+1 } TRACE_A2DP_DECODER_D( "scalable a2dp frame, length:%d bitrate:%d sampling_rate:%d", decoder_frame_p->buffer_len, bitrate_bps, sampling_rate); frame_len = SCALABLE_HEAD_SIZE + len - 1; if (decoder_frame_p->buffer_len < frame_len) { TRACE_A2DP_DECODER_E( "invalid scalable a2dp frame, length:%d < %d !!!!!!!", decoder_frame_p->buffer_len, frame_len); return false; } return true; } #ifdef A2DP_CP_ACCEL struct A2DP_CP_scalable_IN_FRM_INFO_T { uint16_t sequenceNumber; uint32_t timestamp; }; struct A2DP_CP_scalable_OUT_FRM_INFO_T { struct A2DP_CP_scalable_IN_FRM_INFO_T in_info; uint16_t frame_samples; uint16_t decoded_frames; uint16_t frame_idx; uint16_t pcm_len; }; static bool cp_codec_reset; extern uint32_t app_bt_stream_get_dma_buffer_samples(void); TEXT_SSC_LOC static int a2dp_cp_scalable_mcu_decode(uint8_t *buffer, uint32_t buffer_bytes) { a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; list_node_t *node = NULL; list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; int ret, dec_ret; struct A2DP_CP_scalable_IN_FRM_INFO_T in_info; struct A2DP_CP_scalable_OUT_FRM_INFO_T *p_out_info; uint8_t *out; uint32_t out_len; uint32_t out_frame_len; uint32_t cp_buffer_frames_max = 0; cp_buffer_frames_max = app_bt_stream_get_dma_buffer_samples() / 2; if (cp_buffer_frames_max % (lastframe_info.frame_samples)) { cp_buffer_frames_max = cp_buffer_frames_max / (lastframe_info.frame_samples) + 1; } else { cp_buffer_frames_max = cp_buffer_frames_max / (lastframe_info.frame_samples); } out_frame_len = sizeof(*p_out_info) + buffer_bytes; ret = a2dp_cp_decoder_init(out_frame_len, cp_buffer_frames_max * 2); ASSERT(ret == 0, "%s: a2dp_cp_decoder_init() failed: ret=%d", __func__, ret); while ((node = a2dp_audio_list_begin(list)) != NULL) { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); if (false == is_valid_frame(decoder_frame_p)) { return A2DP_DECODER_DECODE_ERROR; } in_info.sequenceNumber = decoder_frame_p->sequenceNumber; in_info.timestamp = decoder_frame_p->timestamp; ret = a2dp_cp_put_in_frame(&in_info, sizeof(in_info), decoder_frame_p->buffer, decoder_frame_p->buffer_len); if (ret) { break; } a2dp_audio_list_remove(list, decoder_frame_p); } ret = a2dp_cp_get_full_out_frame((void **)&out, &out_len); if (ret) { return A2DP_DECODER_DECODE_ERROR; } if (out_len == 0) { memset(buffer, 0, buffer_bytes); a2dp_cp_consume_full_out_frame(); return A2DP_DECODER_NO_ERROR; } ASSERT(out_len == out_frame_len, "%s: Bad out len %u (should be %u)", __func__, out_len, out_frame_len); p_out_info = (struct A2DP_CP_scalable_OUT_FRM_INFO_T *)out; if (p_out_info->pcm_len) { lastframe_info.sequenceNumber = p_out_info->in_info.sequenceNumber; lastframe_info.timestamp = p_out_info->in_info.timestamp; lastframe_info.curSubSequenceNumber = 0; lastframe_info.totalSubSequenceNumber = 0; lastframe_info.frame_samples = p_out_info->frame_samples; lastframe_info.decoded_frames += p_out_info->decoded_frames; lastframe_info.undecode_frames = a2dp_audio_list_length(list) + a2dp_cp_get_in_frame_cnt_by_index(p_out_info->frame_idx) - 1; a2dp_audio_decoder_internal_lastframe_info_set(&lastframe_info); } if (p_out_info->pcm_len == buffer_bytes) { memcpy(buffer, p_out_info + 1, p_out_info->pcm_len); dec_ret = A2DP_DECODER_NO_ERROR; } else { dec_ret = A2DP_DECODER_DECODE_ERROR; } ret = a2dp_cp_consume_full_out_frame(); ASSERT(ret == 0, "%s: a2dp_cp_consume_full_out_frame() failed: ret=%d", __func__, ret); return dec_ret; } TEXT_SSC_LOC int a2dp_cp_scalable_cp_decode(void) { int ret; enum CP_EMPTY_OUT_FRM_T out_frm_st; uint8_t *out; uint32_t out_len; uint8_t *dec_start; uint32_t dec_len; struct A2DP_CP_scalable_IN_FRM_INFO_T *p_in_info; struct A2DP_CP_scalable_OUT_FRM_INFO_T *p_out_info; uint8_t *in_buf; uint32_t in_len; uint32_t dec_sum; int error, output_samples = 0, output_byte = 0; if (cp_codec_reset) { cp_codec_reset = false; a2dp_audio_scalable_decoder_init(); } out_frm_st = a2dp_cp_get_emtpy_out_frame((void **)&out, &out_len); if (out_frm_st != CP_EMPTY_OUT_FRM_OK && out_frm_st != CP_EMPTY_OUT_FRM_WORKING) { return 1; } ASSERT(out_len > sizeof(*p_out_info), "%s: Bad out_len %u (should > %u)", __func__, out_len, sizeof(*p_out_info)); p_out_info = (struct A2DP_CP_scalable_OUT_FRM_INFO_T *)out; if (out_frm_st == CP_EMPTY_OUT_FRM_OK) { p_out_info->pcm_len = 0; p_out_info->decoded_frames = 0; } ASSERT(out_len > sizeof(*p_out_info) + p_out_info->pcm_len, "%s: Bad out_len %u (should > %u + %u)", __func__, out_len, sizeof(*p_out_info), p_out_info->pcm_len); dec_start = (uint8_t *)(p_out_info + 1) + p_out_info->pcm_len; dec_len = out_len - (dec_start - (uint8_t *)out); if (!scalableDec_handle) { TRACE(0, "scalable_decode not ready"); return 3; } dec_sum = 0; error = 0; while (dec_sum < dec_len && error == 0) { ret = a2dp_cp_get_in_frame((void **)&in_buf, &in_len); if (ret) { break; } ASSERT(in_len > sizeof(*p_in_info), "%s: Bad in_len %u (should > %u)", __func__, in_len, sizeof(*p_in_info)); p_in_info = (struct A2DP_CP_scalable_IN_FRM_INFO_T *)in_buf; in_buf += sizeof(*p_in_info); in_len -= sizeof(*p_in_info); /* decode one SSC frame */ output_samples = ssc_decode(scalableDec_handle, (const unsigned char *)in_buf, ss_pcm_buff, SCALABLE_FRAME_SIZE, 0, 2); if (0 == output_samples) { a2dp_audio_scalable_decoder_reinit(); TRACE(0, "scalable_decode reinin codec \n"); error = A2DP_DECODER_DECODE_ERROR; goto exit; } ss_to_24bit_buf((int32_t *)(dec_start + dec_sum), (int32_t *)ss_pcm_buff, output_samples); output_byte = output_samples * 4; dec_sum += output_byte; exit: memcpy(&p_out_info->in_info, ss_pcm_buff, sizeof(*p_in_info)); p_out_info->decoded_frames++; p_out_info->frame_samples = SCALABLE_OUTPUT_FRAME_SAMPLES; p_out_info->frame_idx = a2dp_cp_get_in_frame_index(); ret = a2dp_cp_consume_in_frame(); ASSERT(ret == 0, "%s: a2dp_cp_consume_in_frame() failed: ret=%d", __func__, ret); } p_out_info->pcm_len += dec_sum; if (error || out_len <= sizeof(*p_out_info) + p_out_info->pcm_len) { ret = a2dp_cp_consume_emtpy_out_frame(); ASSERT(ret == 0, "%s: a2dp_cp_consume_emtpy_out_frame() failed: ret=%d", __func__, ret); } return error; } #endif static int a2dp_audio_scalable_init(A2DP_AUDIO_OUTPUT_CONFIG_T *config, void *context) { TRACE_A2DP_DECODER_D("%s", __func__); TRACE(0, "\n\nA2DP SSC-LC INIT\n"); a2dp_audio_context_p = (A2DP_AUDIO_CONTEXT_T *)context; memcpy(&output_config, config, sizeof(A2DP_AUDIO_OUTPUT_CONFIG_T)); memset(&lastframe_info, 0, sizeof(A2DP_AUDIO_DECODER_LASTFRAME_INFO_T)); lastframe_info.stream_info = output_config; lastframe_info.frame_samples = SCALABLE_FRAME_SIZE; lastframe_info.list_samples = SCALABLE_FRAME_SIZE; a2dp_audio_decoder_internal_lastframe_info_set(&lastframe_info); ASSERT(a2dp_audio_context_p->dest_packet_mut < SCALABLE_MTU_LIMITER, "%s MTU OVERFLOW:%u/%u", __func__, a2dp_audio_context_p->dest_packet_mut, SCALABLE_MTU_LIMITER); int decoder_size; decoder_size = ssc_decoder_get_size( output_config.num_channels); // todo: get size with codec capability TRACE(0, "decoder size %d", decoder_size); scalable_decoder_place = (unsigned char *)a2dp_audio_heap_malloc(decoder_size); ASSERT_A2DP_DECODER(scalable_decoder_place, "no memory resource for scalable_decoder_place"); scalable_decoder_temp_buf = (unsigned char *)a2dp_audio_heap_malloc(SCALABLE_FRAME_SIZE * 16); ASSERT_A2DP_DECODER(scalable_decoder_temp_buf, "no memory resource for scalable_decoder_temp_buf"); #ifdef A2DP_CP_ACCEL int ret; cp_codec_reset = true; ret = a2dp_cp_init(a2dp_cp_scalable_cp_decode, CP_PROC_DELAY_1_FRAME); ASSERT(ret == 0, "%s: a2dp_cp_init() failed: ret=%d", __func__, ret); #else a2dp_audio_scalable_decoder_init(); #endif return A2DP_DECODER_NO_ERROR; } static int a2dp_audio_scalable_deinit(void) { #ifdef A2DP_CP_ACCEL a2dp_cp_deinit(); #endif a2dp_audio_scalable_decoder_deinit(); TRACE(0, "\n\nA2DP SCALABLE DEINIT\n"); return A2DP_DECODER_NO_ERROR; } int a2dp_audio_scalable_mcu_decode_frame(uint8_t *buffer, uint32_t buffer_bytes) { list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; list_node_t *node = NULL; a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; int ret = A2DP_DECODER_NO_ERROR; bool cache_underflow = false; int output_byte = 0, output_samples = 0; uint8_t *output = buffer; // TRACE(1,"bbd %d",buffer_bytes ); if (buffer_bytes < (SCALABLE_FRAME_SIZE * output_config.num_channels * output_config.bits_depth / 8)) { TRACE(1, "scalable_decode pcm_len = %d \n", buffer_bytes); return A2DP_DECODER_NO_ERROR; } if (!scalableDec_handle) { TRACE(0, "scalable_decode not ready"); return A2DP_DECODER_NO_ERROR; } while (output < buffer + buffer_bytes) { node = a2dp_audio_list_begin(list); if (!node) { TRACE(0, "scalable_decode cache underflow"); cache_underflow = true; goto exit; } else { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); if (false == is_valid_frame(decoder_frame_p)) { TRACE_A2DP_DECODER_E("%s %d invalid a2dp frame", __func__, __LINE__); ret = A2DP_DECODER_DECODE_ERROR; goto exit; } /* decode one SSC frame */ output_samples = ssc_decode(scalableDec_handle, decoder_frame_p->buffer, ss_pcm_buff, SCALABLE_FRAME_SIZE, 0, 2); TRACE_A2DP_DECODER_D("scalable_decode seq:%d len:%d output_samples:%d", decoder_frame_p->sequenceNumber, decoder_frame_p->buffer_len, output_samples); if (0 == output_samples) { TRACE(0, "scalable_decode failed !!!!!!"); // if failed reopen it again a2dp_audio_scalable_decoder_reinit(); TRACE(0, "scalable_decode reinin codec \n"); ret = A2DP_DECODER_DECODE_ERROR; goto exit; } ss_to_24bit_buf((int32_t *)output, (int32_t *)ss_pcm_buff, output_samples * 2); output_byte = output_samples * 8; output += output_byte; ASSERT(SCALABLE_FRAME_SIZE == output_samples, "scalable_decode output mismatch samples:%d", output_samples); lastframe_info.sequenceNumber = decoder_frame_p->sequenceNumber; lastframe_info.timestamp = decoder_frame_p->timestamp; lastframe_info.curSubSequenceNumber = 0; lastframe_info.totalSubSequenceNumber = 0; lastframe_info.frame_samples = SCALABLE_FRAME_SIZE; lastframe_info.decoded_frames++; lastframe_info.undecode_frames = a2dp_audio_list_length(list) - 1; a2dp_audio_decoder_internal_lastframe_info_set(&lastframe_info); a2dp_audio_list_remove(list, decoder_frame_p); } } exit: if (cache_underflow) { lastframe_info.undecode_frames = 0; a2dp_audio_decoder_internal_lastframe_info_set(&lastframe_info); ret = A2DP_DECODER_CACHE_UNDERFLOW_ERROR; } // TRACE(0,"abd"); return ret; } static int a2dp_audio_scalable_decode_frame(uint8_t *buffer, uint32_t buffer_bytes) { #ifdef A2DP_CP_ACCEL return a2dp_cp_scalable_mcu_decode(buffer, buffer_bytes); #else return a2dp_audio_scalable_mcu_decode_frame(buffer, buffer_bytes); #endif } static int a2dp_audio_scalable_preparse_packet(btif_media_header_t *header, uint8_t *buffer, uint32_t buffer_bytes) { lastframe_info.sequenceNumber = header->sequenceNumber; lastframe_info.timestamp = header->timestamp; lastframe_info.curSubSequenceNumber = 0; lastframe_info.totalSubSequenceNumber = 0; lastframe_info.frame_samples = SCALABLE_FRAME_SIZE; lastframe_info.list_samples = SCALABLE_FRAME_SIZE; lastframe_info.decoded_frames = 0; lastframe_info.undecode_frames = 0; a2dp_audio_decoder_internal_lastframe_info_set(&lastframe_info); TRACE(3, "%s seq:%d timestamp:%08x", __func__, header->sequenceNumber, header->timestamp); return A2DP_DECODER_NO_ERROR; } static void *a2dp_audio_scalable_frame_malloc(uint32_t packet_len) { a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; uint8_t *buffer = NULL; buffer = (uint8_t *)a2dp_audio_heap_malloc(SCALABLE_READBUF_SIZE); decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_heap_malloc( sizeof(a2dp_audio_scalable_decoder_frame_t)); decoder_frame_p->buffer = buffer; decoder_frame_p->buffer_len = packet_len; return (void *)decoder_frame_p; } static void a2dp_audio_scalable_free(void *packet) { a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)packet; a2dp_audio_heap_free(decoder_frame_p->buffer); a2dp_audio_heap_free(decoder_frame_p); } static int a2dp_audio_scalable_store_packet(btif_media_header_t *header, uint8_t *buffer, uint32_t buffer_bytes) { list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; int nRet = A2DP_DECODER_NO_ERROR; if (a2dp_audio_list_length(list) < SCALABLE_MTU_LIMITER) { a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_scalable_frame_malloc( buffer_bytes); TRACE_A2DP_DECODER_D("%s seq:%d len:%d", __func__, header->sequenceNumber, buffer_bytes); decoder_frame_p->sequenceNumber = header->sequenceNumber; decoder_frame_p->timestamp = header->timestamp; memcpy(decoder_frame_p->buffer, buffer, buffer_bytes); decoder_frame_p->buffer_len = buffer_bytes; a2dp_audio_list_append(list, decoder_frame_p); nRet = A2DP_DECODER_NO_ERROR; } else { TRACE(2, "%s list full current len:%d", __func__, a2dp_audio_list_length(list)); nRet = A2DP_DECODER_MTU_LIMTER_ERROR; } return nRet; } static int a2dp_audio_scalable_discards_packet(uint32_t packets) { int nRet = A2DP_DECODER_MEMORY_ERROR; list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; list_node_t *node = NULL; a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; #ifdef A2DP_CP_ACCEL a2dp_cp_reset_frame(); #endif if (packets <= a2dp_audio_list_length(list)) { for (uint8_t i = 0; i < packets; i++) { if ((node = a2dp_audio_list_begin(list)) != NULL) { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); a2dp_audio_list_remove(list, decoder_frame_p); } } nRet = A2DP_DECODER_NO_ERROR; } TRACE(3, "%s packets:%d nRet:%d", __func__, packets, nRet); return nRet; } static int a2dp_audio_scalable_headframe_info_get( A2DP_AUDIO_HEADFRAME_INFO_T *headframe_info) { list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; list_node_t *node = NULL; a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; if (a2dp_audio_list_length(list)) { if ((node = a2dp_audio_list_begin(list)) != NULL) { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); headframe_info->sequenceNumber = decoder_frame_p->sequenceNumber; headframe_info->timestamp = decoder_frame_p->timestamp; headframe_info->curSubSequenceNumber = 0; headframe_info->totalSubSequenceNumber = 0; } } else { memset(headframe_info, 0, sizeof(A2DP_AUDIO_HEADFRAME_INFO_T)); } return A2DP_DECODER_NO_ERROR; } static int a2dp_audio_scalable_info_get(void *info) { return A2DP_DECODER_NO_ERROR; } static int a2dp_audio_scalable_synchronize_packet(A2DP_AUDIO_SYNCFRAME_INFO_T *sync_info, uint32_t mask) { int nRet = A2DP_DECODER_SYNC_ERROR; list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; list_node_t *node = NULL; int list_len; a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; #ifdef A2DP_CP_ACCEL a2dp_cp_reset_frame(); #endif list_len = a2dp_audio_list_length(list); for (uint16_t i = 0; i < list_len; i++) { if ((node = a2dp_audio_list_begin(list)) != NULL) { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); if (A2DP_AUDIO_SYNCFRAME_CHK(decoder_frame_p->sequenceNumber == sync_info->sequenceNumber, A2DP_AUDIO_SYNCFRAME_MASK_SEQ, mask) && A2DP_AUDIO_SYNCFRAME_CHK(decoder_frame_p->timestamp == sync_info->timestamp, A2DP_AUDIO_SYNCFRAME_MASK_TIMESTAMP, mask)) { nRet = A2DP_DECODER_NO_ERROR; break; } a2dp_audio_list_remove(list, decoder_frame_p); } } node = a2dp_audio_list_begin(list); if (node) { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); TRACE(4, "%s nRet:%d SEQ:%d timestamp:%d", __func__, nRet, decoder_frame_p->sequenceNumber, decoder_frame_p->timestamp); } else { TRACE(2, "%s nRet:%d", __func__, nRet); } return nRet; } static int a2dp_audio_scalable_synchronize_dest_packet_mut(uint16_t packet_mut) { list_node_t *node = NULL; uint32_t list_len = 0; list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; a2dp_audio_scalable_decoder_frame_t *decoder_frame_p = NULL; list_len = a2dp_audio_list_length(list); if (list_len > packet_mut) { do { node = a2dp_audio_list_begin(list); if (node) { decoder_frame_p = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); a2dp_audio_list_remove(list, decoder_frame_p); } } while (a2dp_audio_list_length(list) > packet_mut); } TRACE(2, "%s list:%d", __func__, a2dp_audio_list_length(list)); return A2DP_DECODER_NO_ERROR; } int a2dp_audio_scalable_convert_list_to_samples(uint32_t *samples) { uint32_t list_len = 0; list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; list_len = a2dp_audio_list_length(list); *samples = SCALABLE_FRAME_SIZE * list_len; TRACE(3, "%s list:%d samples:%d", __func__, list_len, *samples); return A2DP_DECODER_NO_ERROR; } int a2dp_audio_scalable_discards_samples(uint32_t samples) { int nRet = A2DP_DECODER_SYNC_ERROR; list_t *list = a2dp_audio_context_p->audio_datapath.input_raw_packet_list; a2dp_audio_scalable_decoder_frame_t *scalable_decoder_frame = NULL; list_node_t *node = NULL; int need_remove_list = 0; uint32_t list_samples = 0; ASSERT(!(samples % SCALABLE_FRAME_SIZE), "%s samples err:%d", __func__, samples); a2dp_audio_scalable_convert_list_to_samples(&list_samples); if (list_samples >= samples) { need_remove_list = samples / SCALABLE_FRAME_SIZE; for (int i = 0; i < need_remove_list; i++) { node = a2dp_audio_list_begin(list); if (node) { scalable_decoder_frame = (a2dp_audio_scalable_decoder_frame_t *)a2dp_audio_list_node(node); a2dp_audio_list_remove(list, scalable_decoder_frame); } } nRet = A2DP_DECODER_NO_ERROR; } return nRet; } A2DP_AUDIO_DECODER_T a2dp_audio_scalable_decoder_config = { {44100, 2, 16}, 1, a2dp_audio_scalable_init, a2dp_audio_scalable_deinit, a2dp_audio_scalable_decode_frame, a2dp_audio_scalable_preparse_packet, a2dp_audio_scalable_store_packet, a2dp_audio_scalable_discards_packet, a2dp_audio_scalable_synchronize_packet, a2dp_audio_scalable_synchronize_dest_packet_mut, a2dp_audio_scalable_convert_list_to_samples, a2dp_audio_scalable_discards_samples, a2dp_audio_scalable_headframe_info_get, a2dp_audio_scalable_info_get, a2dp_audio_scalable_free, }; #else A2DP_AUDIO_DECODER_T a2dp_audio_scalable_decoder_config = { 0, }; #endif