/* * This file is part of the MicroPython project, http://micropython.org/ * * The MIT License (MIT) * * Copyright (c) 2021 Scott Shawcroft for Adafruit Industries * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to deal * in the Software without restriction, including without limitation the rights * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell * copies of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN * THE SOFTWARE. */ #include "audio_dma.h" #include "shared-bindings/audiocore/RawSample.h" #include "shared-bindings/audiocore/WaveFile.h" #include "shared-bindings/microcontroller/__init__.h" #include "bindings/rp2pio/StateMachine.h" #include "supervisor/background_callback.h" #include "py/mpstate.h" #include "py/runtime.h" #include "src/rp2_common/hardware_irq/include/hardware/irq.h" #if CIRCUITPY_AUDIOCORE void audio_dma_reset(void) { for (size_t channel = 0; channel < NUM_DMA_CHANNELS; channel++) { if (MP_STATE_PORT(playing_audio)[channel] == NULL) { continue; } audio_dma_stop(MP_STATE_PORT(playing_audio)[channel]); } } STATIC size_t audio_dma_convert_samples(audio_dma_t *dma, uint8_t *input, uint32_t input_length, uint8_t *output, uint32_t output_length) { #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-Wcast-align" uint32_t output_length_used = input_length / dma->sample_spacing; if (output_length_used > output_length) { mp_raise_RuntimeError(translate("Internal audio buffer too small")); } uint32_t out_i = 0; if (dma->sample_resolution <= 8 && dma->output_resolution > 8) { // reading bytes, writing 16-bit words, so output buffer will be bigger. output_length_used *= 2; if (output_length_used > output_length) { mp_raise_RuntimeError(translate("Internal audio buffer too small")); } // Correct "rail-to-rail" scaling of arbitrary-depth input to output // requires more operations than this, but at least the vital 8- to // 16-bit cases are correctly scaled now. Prior code was only // considering 8-to-16 anyway, but had a slight DC offset in the // result, so this is no worse off WRT supported resolutions. uint16_t mul = ((1 << dma->output_resolution) - 1) / ((1 << dma->sample_resolution) - 1); uint16_t offset = (1 << dma->output_resolution) / 2; for (uint32_t i = 0; i < input_length; i += dma->sample_spacing) { if (dma->signed_to_unsigned) { ((uint16_t *)output)[out_i] = (uint16_t)((((int8_t *)input)[i] + 0x80) * mul); } else if (dma->unsigned_to_signed) { ((int16_t *)output)[out_i] = (int16_t)(((uint8_t *)input)[i] * mul - offset); } else { ((uint16_t *)output)[out_i] = (uint16_t)(((uint8_t *)input)[i] * mul); } out_i += 1; } } else if (dma->sample_resolution <= 8 && dma->output_resolution <= 8) { for (uint32_t i = 0; i < input_length; i += dma->sample_spacing) { if (dma->signed_to_unsigned) { ((uint8_t *)output)[out_i] = ((int8_t *)input)[i] + 0x80; } else if (dma->unsigned_to_signed) { ((int8_t *)output)[out_i] = ((uint8_t *)input)[i] - 0x80; } else { ((uint8_t *)output)[out_i] = ((uint8_t *)input)[i]; } out_i += 1; } } else if (dma->sample_resolution > 8 && dma->output_resolution > 8) { size_t shift = 16 - dma->output_resolution; for (uint32_t i = 0; i < input_length / 2; i += dma->sample_spacing) { if (dma->signed_to_unsigned) { ((uint16_t *)output)[out_i] = ((int16_t *)input)[i] + 0x8000; } else if (dma->unsigned_to_signed) { ((int16_t *)output)[out_i] = ((uint16_t *)input)[i] - 0x8000; } else { ((uint16_t *)output)[out_i] = ((uint16_t *)input)[i]; } if (dma->output_resolution < 16) { if (dma->output_signed) { ((int16_t *)output)[out_i] = ((int16_t *)output)[out_i] >> shift; } else { ((uint16_t *)output)[out_i] = ((uint16_t *)output)[out_i] >> shift; } } out_i += 1; } } else { // (dma->sample_resolution > 8 && dma->output_resolution <= 8) // Not currently used, but might be in the future. mp_raise_RuntimeError(translate("Audio conversion not implemented")); } #pragma GCC diagnostic pop return output_length_used; } // buffer_idx is 0 or 1. STATIC void audio_dma_load_next_block(audio_dma_t *dma, size_t buffer_idx) { size_t dma_channel = dma->channel[buffer_idx]; audioio_get_buffer_result_t get_buffer_result; uint8_t *sample_buffer; uint32_t sample_buffer_length; get_buffer_result = audiosample_get_buffer(dma->sample, dma->single_channel_output, dma->audio_channel, &sample_buffer, &sample_buffer_length); if (get_buffer_result == GET_BUFFER_ERROR) { audio_dma_stop(dma); return; } // Convert the sample format resolution and signedness, as necessary. // The input sample buffer is what was read from a file, Mixer, or a raw sample buffer. // The output buffer is one of the DMA buffers (passed in). size_t output_length_used = audio_dma_convert_samples( dma, sample_buffer, sample_buffer_length, dma->buffer[buffer_idx], dma->buffer_length[buffer_idx]); dma_channel_set_read_addr(dma_channel, dma->buffer[buffer_idx], false /* trigger */); dma_channel_set_trans_count(dma_channel, output_length_used / dma->output_size, false /* trigger */); if (get_buffer_result == GET_BUFFER_DONE) { if (dma->loop) { audiosample_reset_buffer(dma->sample, dma->single_channel_output, dma->audio_channel); } else { // Set channel trigger to ourselves so we don't keep going. dma_channel_hw_t *c = &dma_hw->ch[dma_channel]; c->al1_ctrl = (c->al1_ctrl & ~DMA_CH0_CTRL_TRIG_CHAIN_TO_BITS) | (dma_channel << DMA_CH0_CTRL_TRIG_CHAIN_TO_LSB); if (output_length_used == 0 && !dma_channel_is_busy(dma->channel[0]) && !dma_channel_is_busy(dma->channel[1])) { // No data has been read, and both DMA channels have now finished, so it's safe to stop. audio_dma_stop(dma); dma->playing_in_progress = false; } } } } // Playback should be shutdown before calling this. audio_dma_result audio_dma_setup_playback( audio_dma_t *dma, mp_obj_t sample, bool loop, bool single_channel_output, uint8_t audio_channel, bool output_signed, uint8_t output_resolution, uint32_t output_register_address, uint8_t dma_trigger_source) { // Use two DMA channels to play because the DMA can't wrap to itself without the // buffer being power of two aligned. int dma_channel_0_maybe = dma_claim_unused_channel(false); if (dma_channel_0_maybe < 0) { return AUDIO_DMA_DMA_BUSY; } int dma_channel_1_maybe = dma_claim_unused_channel(false); if (dma_channel_1_maybe < 0) { dma_channel_unclaim((uint)dma_channel_0_maybe); return AUDIO_DMA_DMA_BUSY; } dma->channel[0] = (uint8_t)dma_channel_0_maybe; dma->channel[1] = (uint8_t)dma_channel_1_maybe; dma->sample = sample; dma->loop = loop; dma->single_channel_output = single_channel_output; dma->audio_channel = audio_channel; dma->signed_to_unsigned = false; dma->unsigned_to_signed = false; dma->output_signed = output_signed; dma->sample_spacing = 1; dma->output_resolution = output_resolution; dma->sample_resolution = audiosample_bits_per_sample(sample); dma->output_register_address = output_register_address; audiosample_reset_buffer(sample, single_channel_output, audio_channel); bool single_buffer; // True if data fits in one single buffer. bool samples_signed; uint32_t max_buffer_length; audiosample_get_buffer_structure(sample, single_channel_output, &single_buffer, &samples_signed, &max_buffer_length, &dma->sample_spacing); // Check to see if we have to scale the resolution up. if (dma->sample_resolution <= 8 && dma->output_resolution > 8) { max_buffer_length *= 2; } if (output_signed != samples_signed || dma->sample_spacing > 1 || (dma->sample_resolution != dma->output_resolution)) { max_buffer_length /= dma->sample_spacing; } dma->buffer[0] = (uint8_t *)m_realloc(dma->buffer[0], max_buffer_length); dma->buffer_length[0] = max_buffer_length; if (dma->buffer[0] == NULL) { return AUDIO_DMA_MEMORY_ERROR; } if (!single_buffer) { dma->buffer[1] = (uint8_t *)m_realloc(dma->buffer[1], max_buffer_length); dma->buffer_length[1] = max_buffer_length; if (dma->buffer[1] == NULL) { return AUDIO_DMA_MEMORY_ERROR; } } dma->signed_to_unsigned = !output_signed && samples_signed; dma->unsigned_to_signed = output_signed && !samples_signed; if (output_resolution > 8) { dma->output_size = 2; } else { dma->output_size = 1; } // Transfer both channels at once. if (!single_channel_output && audiosample_channel_count(sample) == 2) { dma->output_size *= 2; } enum dma_channel_transfer_size dma_size = DMA_SIZE_8; if (dma->output_size == 2) { dma_size = DMA_SIZE_16; } else if (dma->output_size == 4) { dma_size = DMA_SIZE_32; } for (size_t i = 0; i < 2; i++) { dma_channel_config c = dma_channel_get_default_config(dma->channel[i]); channel_config_set_transfer_data_size(&c, dma_size); channel_config_set_dreq(&c, dma_trigger_source); channel_config_set_read_increment(&c, true); channel_config_set_write_increment(&c, false); // Chain to the other channel by default. channel_config_set_chain_to(&c, dma->channel[(i + 1) % 2]); dma_channel_set_config(dma->channel[i], &c, false /* trigger */); dma_channel_set_write_addr(dma->channel[i], (void *)output_register_address, false /* trigger */); } // We keep the audio_dma_t for internal use and the sample as a root pointer because it // contains the audiodma structure. MP_STATE_PORT(playing_audio)[dma->channel[0]] = dma; MP_STATE_PORT(playing_audio)[dma->channel[1]] = dma; // Load the first two blocks up front. audio_dma_load_next_block(dma, 0); if (!single_buffer) { audio_dma_load_next_block(dma, 1); } // Special case the DMA for a single buffer. It's commonly used for a single wave length of sound // and may be short. Therefore, we use DMA chaining to loop quickly without involving interrupts. // On the RP2040 we chain by having a second DMA writing to the config registers of the first. // Read and write addresses change with DMA so we need to reset the read address back to the // start of the sample. if (single_buffer) { dma_channel_config c = dma_channel_get_default_config(dma->channel[1]); channel_config_set_transfer_data_size(&c, DMA_SIZE_32); channel_config_set_dreq(&c, 0x3f); // dma as fast as possible channel_config_set_read_increment(&c, false); channel_config_set_write_increment(&c, false); channel_config_set_chain_to(&c, dma->channel[1]); // Chain to ourselves so we stop. dma_channel_configure(dma->channel[1], &c, &dma_hw->ch[dma->channel[0]].al3_read_addr_trig, // write address &dma->buffer[0], // read address 1, // transaction count false); // trigger } else { // Enable our DMA channels on DMA_IRQ_0 to the CPU. This will wake us up when // we're WFI. dma_hw->inte0 |= (1 << dma->channel[0]) | (1 << dma->channel[1]); irq_set_mask_enabled(1 << DMA_IRQ_0, true); } dma->playing_in_progress = true; dma_channel_start(dma->channel[0]); return AUDIO_DMA_OK; } void audio_dma_stop(audio_dma_t *dma) { // Disable our interrupts. uint32_t channel_mask = 0; if (dma->channel[0] < NUM_DMA_CHANNELS) { channel_mask |= 1 << dma->channel[0]; } if (dma->channel[1] < NUM_DMA_CHANNELS) { channel_mask |= 1 << dma->channel[1]; } dma_hw->inte0 &= ~channel_mask; if (!dma_hw->inte0) { irq_set_mask_enabled(1 << DMA_IRQ_0, false); } // Run any remaining audio tasks because we remove ourselves from // playing_audio. RUN_BACKGROUND_TASKS; for (size_t i = 0; i < 2; i++) { size_t channel = dma->channel[i]; if (channel == NUM_DMA_CHANNELS) { // Channel not in use. continue; } dma_channel_config c = dma_channel_get_default_config(dma->channel[i]); channel_config_set_enable(&c, false); dma_channel_set_config(channel, &c, false /* trigger */); if (dma_channel_is_busy(channel)) { dma_channel_abort(channel); } dma_channel_set_read_addr(channel, NULL, false /* trigger */); dma_channel_set_write_addr(channel, NULL, false /* trigger */); dma_channel_set_trans_count(channel, 0, false /* trigger */); dma_channel_unclaim(channel); MP_STATE_PORT(playing_audio)[channel] = NULL; dma->channel[i] = NUM_DMA_CHANNELS; } dma->playing_in_progress = false; // Hold onto our buffers. } // To pause we simply stop the DMA. It is the responsibility of the output peripheral // to hold the previous value. void audio_dma_pause(audio_dma_t *dma) { dma_hw->ch[dma->channel[0]].al1_ctrl &= ~DMA_CH0_CTRL_TRIG_EN_BITS; dma_hw->ch[dma->channel[1]].al1_ctrl &= ~DMA_CH1_CTRL_TRIG_EN_BITS; } void audio_dma_resume(audio_dma_t *dma) { // Always re-enable the non-busy channel first so it's ready to continue when the busy channel // finishes and chains to it. (An interrupt could make the time between enables long.) if (dma_channel_is_busy(dma->channel[0])) { dma_hw->ch[dma->channel[1]].al1_ctrl |= DMA_CH1_CTRL_TRIG_EN_BITS; dma_hw->ch[dma->channel[0]].al1_ctrl |= DMA_CH0_CTRL_TRIG_EN_BITS; } else { dma_hw->ch[dma->channel[0]].al1_ctrl |= DMA_CH0_CTRL_TRIG_EN_BITS; dma_hw->ch[dma->channel[1]].al1_ctrl |= DMA_CH1_CTRL_TRIG_EN_BITS; } } bool audio_dma_get_paused(audio_dma_t *dma) { if (dma->channel[0] >= NUM_DMA_CHANNELS) { return false; } uint32_t control = dma_hw->ch[dma->channel[0]].ctrl_trig; return (control & DMA_CH0_CTRL_TRIG_EN_BITS) == 0; } uint32_t audio_dma_pause_all(void) { uint32_t result = 0; for (size_t channel = 0; channel < NUM_DMA_CHANNELS; channel++) { audio_dma_t *dma = MP_STATE_PORT(playing_audio)[channel]; if (dma != NULL && !audio_dma_get_paused(dma)) { audio_dma_pause(dma); result |= (1 << channel); } } return result; } void audio_dma_unpause_mask(uint32_t channel_mask) { for (size_t channel = 0; channel < NUM_DMA_CHANNELS; channel++) { audio_dma_t *dma = MP_STATE_PORT(playing_audio)[channel]; if (dma != NULL && (channel_mask & (1 << channel))) { audio_dma_resume(dma); } } } void audio_dma_init(audio_dma_t *dma) { dma->buffer[0] = NULL; dma->buffer[1] = NULL; dma->channel[0] = NUM_DMA_CHANNELS; dma->channel[1] = NUM_DMA_CHANNELS; } void audio_dma_deinit(audio_dma_t *dma) { m_free(dma->buffer[0]); dma->buffer[0] = NULL; m_free(dma->buffer[1]); dma->buffer[1] = NULL; } bool audio_dma_get_playing(audio_dma_t *dma) { if (dma->channel[0] == NUM_DMA_CHANNELS) { return false; } return dma->playing_in_progress; } // WARN(tannewt): DO NOT print from here, or anything it calls. Printing calls // background tasks such as this and causes a stack overflow. // NOTE(dhalbert): I successfully printed from here while debugging. // So it's possible, but be careful. STATIC void dma_callback_fun(void *arg) { audio_dma_t *dma = arg; if (dma == NULL) { return; } common_hal_mcu_disable_interrupts(); uint32_t channels_to_load_mask = dma->channels_to_load_mask; dma->channels_to_load_mask = 0; common_hal_mcu_enable_interrupts(); // Load the blocks for the requested channels. uint32_t channel = 0; while (channels_to_load_mask) { if (channels_to_load_mask & 1) { if (dma->channel[0] == channel) { audio_dma_load_next_block(dma, 0); } if (dma->channel[1] == channel) { audio_dma_load_next_block(dma, 1); } } channels_to_load_mask >>= 1; channel++; } } void isr_dma_0(void) { for (size_t i = 0; i < NUM_DMA_CHANNELS; i++) { uint32_t mask = 1 << i; if ((dma_hw->intr & mask) == 0) { continue; } // acknowledge interrupt early. Doing so late means that you could lose an // interrupt if the buffer is very small and the DMA operation // completed by the time callback_add() / dma_complete() returned. This // affected PIO continuous write more than audio. dma_hw->ints0 = mask; if (MP_STATE_PORT(playing_audio)[i] != NULL) { audio_dma_t *dma = MP_STATE_PORT(playing_audio)[i]; // Record all channels whose DMA has completed; they need loading. dma->channels_to_load_mask |= mask; background_callback_add(&dma->callback, dma_callback_fun, (void *)dma); } if (MP_STATE_PORT(background_pio)[i] != NULL) { rp2pio_statemachine_obj_t *pio = MP_STATE_PORT(background_pio)[i]; rp2pio_statemachine_dma_complete(pio, i); } } } #endif