/* * Copyright (C) 2020 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. */ #define LOG_TAG "audio_hw_fir_filter" //#define LOG_NDEBUG 0 #include #include #include #include #include #include #include #include "fir_filter.h" #ifdef __ARM_NEON #include "arm_neon.h" #endif /* #ifdef __ARM_NEON */ fir_filter_t* fir_init(uint32_t channels, fir_filter_mode_t mode, uint32_t filter_length, uint32_t input_length, int16_t* coeffs) { if ((channels == 0) || (filter_length == 0) || (coeffs == NULL)) { ALOGE("%s: Invalid channel count, filter length or coefficient array.", __func__); return NULL; } fir_filter_t* fir = (fir_filter_t*)calloc(1, sizeof(fir_filter_t)); if (fir == NULL) { ALOGE("%s: Unable to allocate memory for fir_filter.", __func__); return NULL; } fir->channels = channels; fir->filter_length = filter_length; /* Default: same filter coeffs for all channels */ fir->mode = FIR_SINGLE_FILTER; uint32_t coeff_bytes = fir->filter_length * sizeof(int16_t); if (mode == FIR_PER_CHANNEL_FILTER) { fir->mode = FIR_PER_CHANNEL_FILTER; coeff_bytes = fir->filter_length * fir->channels * sizeof(int16_t); } fir->coeffs = (int16_t*)malloc(coeff_bytes); if (fir->coeffs == NULL) { ALOGE("%s: Unable to allocate memory for FIR coeffs", __func__); goto exit_1; } memcpy(fir->coeffs, coeffs, coeff_bytes); fir->buffer_size = (input_length + fir->filter_length) * fir->channels; fir->state = (int16_t*)malloc(fir->buffer_size * sizeof(int16_t)); if (fir->state == NULL) { ALOGE("%s: Unable to allocate memory for FIR state", __func__); goto exit_2; } #ifdef __ARM_NEON ALOGI("%s: Using ARM Neon", __func__); #endif /* #ifdef __ARM_NEON */ fir_reset(fir); return fir; exit_2: free(fir->coeffs); exit_1: free(fir); return NULL; } void fir_release(fir_filter_t* fir) { if (fir == NULL) { return; } free(fir->state); free(fir->coeffs); free(fir); } void fir_reset(fir_filter_t* fir) { if (fir == NULL) { return; } memset(fir->state, 0, fir->buffer_size * sizeof(int16_t)); } void fir_process_interleaved(fir_filter_t* fir, int16_t* input, int16_t* output, uint32_t samples) { assert(fir != NULL); int start_offset = (fir->filter_length - 1) * fir->channels; memcpy(&fir->state[start_offset], input, samples * fir->channels * sizeof(int16_t)); // int ch; bool use_2nd_set_coeffs = (fir->channels > 1) && (fir->mode == FIR_PER_CHANNEL_FILTER); int16_t* p_coeff_A = &fir->coeffs[0]; int16_t* p_coeff_B = use_2nd_set_coeffs ? &fir->coeffs[fir->filter_length] : &fir->coeffs[0]; int16_t* p_output; for (int ch = 0; ch < fir->channels; ch += 2) { p_output = &output[ch]; int offset = start_offset + ch; for (int s = 0; s < samples; s++) { int32_t acc_A = 0; int32_t acc_B = 0; #ifdef __ARM_NEON int32x4_t acc_vec = vdupq_n_s32(0); for (int k = 0; k < fir->filter_length; k++, offset -= fir->channels) { int16x4_t coeff_vec = vdup_n_s16(p_coeff_A[k]); coeff_vec = vset_lane_s16(p_coeff_B[k], coeff_vec, 1); int16x4_t input_vec = vld1_s16(&fir->state[offset]); acc_vec = vmlal_s16(acc_vec, coeff_vec, input_vec); } acc_A = vgetq_lane_s32(acc_vec, 0); acc_B = vgetq_lane_s32(acc_vec, 1); #else for (int k = 0; k < fir->filter_length; k++, offset -= fir->channels) { int32_t input_A = (int32_t)(fir->state[offset]); int32_t coeff_A = (int32_t)(p_coeff_A[k]); int32_t input_B = (int32_t)(fir->state[offset + 1]); int32_t coeff_B = (int32_t)(p_coeff_B[k]); acc_A += (input_A * coeff_A); acc_B += (input_B * coeff_B); } #endif /* #ifdef __ARM_NEON */ *p_output = clamp16(acc_A >> 15); if (ch < fir->channels - 1) { *(p_output + 1) = clamp16(acc_B >> 15); } /* Move to next sample */ p_output += fir->channels; offset += (fir->filter_length + 1) * fir->channels; } if (use_2nd_set_coeffs) { p_coeff_A += (fir->filter_length << 1); p_coeff_B += (fir->filter_length << 1); } } memmove(fir->state, &fir->state[samples * fir->channels], (fir->filter_length - 1) * fir->channels * sizeof(int16_t)); }