The kernel renders two inputs now: the Rosella .personalized_headphone model and an already compiled .jochrtf field. The SOFA reader, its compile cache and the field fit leave the kernel, and the direct SOFA runtime built on them goes with them: the point-source distance reference, the shoebox image-source model, the generated FDN tables and the asset archives are all gone. Fitting a SOFA into a .jochrtf belongs to the offline toolchain. The compiled-field runtime is the one that predates that work: 64-QMF analysis, 77 hybrid bands, per-object path delays and the project room (image sources plus the shared unitary FDN) computed in code. It keeps its behaviour and takes the .jochrtf name throughout: sofa_binaural_renderer.cpp -> jochrtf_renderer.cpp, binaural_runtime.* -> jochrtf_runtime.*, SofaBinauralRuntime -> JochrtfRuntime, ejoc_sofa_binaural_* -> ejoc_jochrtf_*. The room becomes a switch. --jochrtf-room on|off (default on) drives the enable_early_reflections and enable_late_room flags of the existing room configuration; off renders the field's direct binaural sound only. joc_task_config and joc_stream_config drop hrtf_sofa_path, hrtf_cache_dir, hrtf_cache_policy and hrtf_radius_m, and gain jochrtf_room. The CLI no longer accepts --sofa-hrtf, --hrtf-cache-policy, --hrtf-cache-dir or --hrtf-radius-m; without an explicit input the renderer uses HRTF/binaural.personalized_headphone, and a .jochrtf has to be named through --compiled-hrtf-cache. The SOFA-only unit tests go with the reader. Rosella parsing, filterbank, room and timing are untouched. Both .jochrtf modes reproduce the Python reference sample for sample, and ctest passes.
256 lines
9.5 KiB
C
256 lines
9.5 KiB
C
#pragma once
|
|
|
|
#include <stdint.h>
|
|
|
|
/* EJOC_STATIC: this branch exists for building the sources directly into an application, where nothing is imported or exported. */
|
|
#if defined(EJOC_STATIC)
|
|
#define EJOC_API
|
|
#define EJOC_CALL __cdecl
|
|
#elif defined(_WIN32)
|
|
#if defined(EJOC_BUILD_DLL)
|
|
#define EJOC_API __declspec(dllexport)
|
|
#else
|
|
#define EJOC_API __declspec(dllimport)
|
|
#endif
|
|
#define EJOC_CALL __cdecl
|
|
#else
|
|
#define EJOC_API __attribute__((visibility("default")))
|
|
#define EJOC_CALL
|
|
#endif
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
enum {
|
|
EJOC_ABI_VERSION = 2,
|
|
EJOC_FRAME_SAMPLES = 1536,
|
|
EJOC_TIMESLOTS = 24,
|
|
EJOC_SUBBANDS = 64,
|
|
EJOC_CORE_CHANNELS = 5, /* downmix configurations 0 and 3 */
|
|
EJOC_MAX_CORE_CHANNELS = 7, /* downmix configurations 1, 2 and 4 */
|
|
EJOC_OUTPUT_CHANNELS = 16,
|
|
EJOC_MAX_OBJECTS = 15,
|
|
EJOC_MAX_DPOINTS = 2,
|
|
EJOC_MAX_PARAMETER_BANDS = 23,
|
|
EJOC_SPEAKER_BLOCK_SAMPLES = 32,
|
|
EJOC_SPEAKER_COORDINATES = 3,
|
|
EJOC_BINAURAL_BLOCK_SAMPLES = 512,
|
|
EJOC_BINAURAL_INPUT_CHANNELS = 16,
|
|
EJOC_BINAURAL_OUTPUT_CHANNELS = 2,
|
|
EJOC_BINAURAL_QMF_BANDS = 64,
|
|
EJOC_BINAURAL_HYBRID_BANDS = 77
|
|
};
|
|
|
|
typedef void* ejoc_renderer_handle;
|
|
typedef void* ejoc_speaker_renderer_handle;
|
|
typedef void* ejoc_binaural_renderer_handle;
|
|
|
|
/*
|
|
Fixed array layouts used by ejoc_renderer_process():
|
|
bed_planar [core_channel_count][1536]
|
|
lfe [1536] or NULL
|
|
n_bands [15]
|
|
n_dpoints [15]
|
|
slope_idx [15]
|
|
offset_ts [15][2]
|
|
dq [15][2][7][23]
|
|
output16 [16][1536]
|
|
|
|
core_channel_count comes from ejoc_renderer_set_core_channels() and defaults to
|
|
EJOC_CORE_CHANNELS; the dq stride is always EJOC_MAX_CORE_CHANNELS.
|
|
|
|
Only objects selected by object_mask are read from the descriptor arrays.
|
|
dq carries already dequantized matrix coefficients in double precision; the
|
|
caller performs the JOC bitstream differential decoding for both dense and
|
|
sparse objects, so this ABI is identical for both syntaxes.
|
|
*/
|
|
|
|
EJOC_API uint32_t EJOC_CALL ejoc_abi_version(void);
|
|
EJOC_API const char* EJOC_CALL ejoc_build_info(void);
|
|
EJOC_API ejoc_renderer_handle EJOC_CALL ejoc_renderer_create(void);
|
|
EJOC_API void EJOC_CALL ejoc_renderer_destroy(ejoc_renderer_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_renderer_reset(ejoc_renderer_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_renderer_set_threads(ejoc_renderer_handle handle, uint32_t total_threads);
|
|
/*
|
|
core_channel_count comes from joc_dmx_config_idx: 5 for configurations 0 and 3,
|
|
7 for 1, 2 and 4. bed_planar is then [core_channel_count][1536] and only channels
|
|
below core_channel_count are read from dq, whose stride stays [15][2][7][23].
|
|
|
|
phase_shift is sigma(k) = (k in {3,4}): it enables the Ls/Rs -j rotation of
|
|
subbands above 0 together with the band-0 21-tap DC compensation. With the switch
|
|
off, channels 3 and 4 only take the common ten-slot surround delay. The delay
|
|
line and DC history advance either way, so the flag may change between frames.
|
|
*/
|
|
EJOC_API uint32_t EJOC_CALL ejoc_renderer_thread_count(ejoc_renderer_handle handle);
|
|
EJOC_API const char* EJOC_CALL ejoc_renderer_last_error(ejoc_renderer_handle handle);
|
|
|
|
EJOC_API int EJOC_CALL ejoc_renderer_process(
|
|
ejoc_renderer_handle handle,
|
|
const float* bed_planar,
|
|
const float* lfe,
|
|
uint32_t core_channel_count,
|
|
uint32_t phase_shift,
|
|
uint32_t object_mask,
|
|
const uint8_t* n_bands,
|
|
const uint8_t* n_dpoints,
|
|
const uint8_t* slope_idx,
|
|
const uint8_t* offset_ts,
|
|
const double* dq,
|
|
double clipgain,
|
|
float phase_new,
|
|
float output_scale,
|
|
float* output16_planar);
|
|
|
|
/*
|
|
High-precision object-to-speaker renderer.
|
|
|
|
The renderer consumes interleaved float32 input PCM arranged as:
|
|
objects16_interleaved [sample_count][16]
|
|
where channel 0 is LFE and channels 1..15 are point objects. All spatial
|
|
calculations, gain ramps, and accumulation use double. Output is interleaved:
|
|
output_interleaved [sample_count][layout_channel_count]
|
|
|
|
Each metadata entry is a complete object-state snapshot:
|
|
metadata_offsets [metadata_count], relative to this process call
|
|
ramp_durations [metadata_count], in samples
|
|
positions_q15 [metadata_count][15][3]
|
|
region_indices [metadata_count][15] or NULL (all region 0)
|
|
height_enabled [metadata_count][15] or NULL (all enabled)
|
|
object_gains [metadata_count][15] or NULL (all 1.0)
|
|
|
|
metadata_offsets must be nondecreasing and <= sample_count. sample_count must
|
|
be a multiple of EJOC_SPEAKER_BLOCK_SAMPLES. State and unfinished ramps are
|
|
preserved across calls.
|
|
*/
|
|
EJOC_API uint32_t EJOC_CALL ejoc_speaker_layout_channel_count(uint32_t speaker_bitfield);
|
|
EJOC_API ejoc_speaker_renderer_handle EJOC_CALL ejoc_speaker_renderer_create(uint32_t speaker_bitfield);
|
|
EJOC_API void EJOC_CALL ejoc_speaker_renderer_destroy(ejoc_speaker_renderer_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_speaker_renderer_reset(ejoc_speaker_renderer_handle handle);
|
|
EJOC_API const char* EJOC_CALL ejoc_speaker_renderer_last_error(ejoc_speaker_renderer_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_speaker_renderer_process(
|
|
ejoc_speaker_renderer_handle handle,
|
|
const float* objects16_interleaved,
|
|
uint32_t sample_count,
|
|
uint32_t metadata_count,
|
|
const uint32_t* metadata_offsets,
|
|
const uint32_t* ramp_durations,
|
|
const uint16_t* positions_q15,
|
|
const uint8_t* region_indices,
|
|
const uint8_t* height_enabled,
|
|
const double* object_gains,
|
|
double* output_interleaved);
|
|
|
|
EJOC_API ejoc_binaural_renderer_handle EJOC_CALL ejoc_binaural_renderer_create(void);
|
|
EJOC_API void EJOC_CALL ejoc_binaural_renderer_destroy(ejoc_binaural_renderer_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_binaural_renderer_reset(ejoc_binaural_renderer_handle handle);
|
|
EJOC_API const char* EJOC_CALL ejoc_binaural_renderer_last_error(
|
|
ejoc_binaural_renderer_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_binaural_renderer_configure_kernels(
|
|
ejoc_binaural_renderer_handle handle,
|
|
const double* qmf_analysis,
|
|
const double* hybrid_low,
|
|
const int16_t* hybrid_indices,
|
|
const double* hybrid_values,
|
|
uint32_t hybrid_count,
|
|
const double* qmf_basis,
|
|
const double* qmf_taps);
|
|
EJOC_API int EJOC_CALL ejoc_binaural_renderer_configure_room(
|
|
ejoc_binaural_renderer_handle handle,
|
|
uint32_t bands,
|
|
uint32_t allpass_count,
|
|
const uint32_t* allpass_delays,
|
|
const double* allpass_gains,
|
|
const uint32_t* fdn_delays,
|
|
const double* fdn_matrix,
|
|
uint32_t output_tap_delay,
|
|
const double* feedback_complex,
|
|
const double* output_taps,
|
|
const double* output_complex,
|
|
uint32_t extra_count,
|
|
const uint32_t* extra_delays,
|
|
const double* extra_fields_complex,
|
|
const double* extra_matrices);
|
|
EJOC_API int EJOC_CALL ejoc_binaural_renderer_process(
|
|
ejoc_binaural_renderer_handle handle,
|
|
const double* input16_interleaved,
|
|
const double* gains_complex,
|
|
const double* room_sends,
|
|
double output_gain,
|
|
double* output_stereo_interleaved);
|
|
|
|
/*
|
|
Native SOFA binaural renderer.
|
|
|
|
The handle owns the complete runtime: 64-QMF/77-hybrid analysis and synthesis,
|
|
fifth-order ACN/N3D real spherical-harmonic direction-field evaluation,
|
|
per-object whole-QMF-slot delay histories, six first-order image-source early
|
|
reflections, the shared unitary-FDN late room, the 120-180 Hz LFE low-pass and
|
|
the 961-sample latency compensation. The caller configures the filterbank
|
|
tables, the compiled HRTF field and the room constants once, then per 512-sample
|
|
block updates every source with ejoc_jochrtf_set_source() and calls
|
|
ejoc_jochrtf_process(). Process returns the number of trimmed stereo
|
|
samples written; the first 961 processed samples across calls are discarded.
|
|
*/
|
|
typedef void* ejoc_jochrtf_handle;
|
|
|
|
EJOC_API ejoc_jochrtf_handle EJOC_CALL ejoc_jochrtf_create(void);
|
|
EJOC_API void EJOC_CALL ejoc_jochrtf_destroy(ejoc_jochrtf_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_reset(ejoc_jochrtf_handle handle);
|
|
EJOC_API const char* EJOC_CALL ejoc_jochrtf_last_error(
|
|
ejoc_jochrtf_handle handle);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_configure_kernels(
|
|
ejoc_jochrtf_handle handle,
|
|
const double* qmf_analysis,
|
|
const double* hybrid_low,
|
|
const int16_t* hybrid_indices,
|
|
const double* hybrid_values,
|
|
uint32_t hybrid_count,
|
|
const double* qmf_basis,
|
|
const double* qmf_taps);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_configure_field(
|
|
ejoc_jochrtf_handle handle,
|
|
const double* coefficients,
|
|
const double* delay_coefficients,
|
|
const double* delay_bounds,
|
|
const double* band_centers,
|
|
double measurement_radius_m);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_configure_room(
|
|
ejoc_jochrtf_handle handle,
|
|
const double* room_dims,
|
|
const double* listener_pos,
|
|
const double* wall_gains,
|
|
double speed_of_sound,
|
|
const uint32_t* fdn_delays,
|
|
const double* fdn_feedback,
|
|
double damping,
|
|
double fdn_output_gain,
|
|
const uint32_t* allpass_delays,
|
|
const double* allpass_gains,
|
|
uint32_t enable_early_reflections,
|
|
uint32_t enable_late_room);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_set_source(
|
|
ejoc_jochrtf_handle handle,
|
|
uint32_t source,
|
|
const double* position_adm,
|
|
uint32_t profile,
|
|
double gain,
|
|
uint32_t enabled,
|
|
uint32_t special_lfe,
|
|
uint32_t fade);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_process(
|
|
ejoc_jochrtf_handle handle,
|
|
const double* input16_interleaved,
|
|
uint32_t sample_count,
|
|
double output_gain,
|
|
double* output_stereo_interleaved);
|
|
EJOC_API int EJOC_CALL ejoc_jochrtf_finish(
|
|
ejoc_jochrtf_handle handle,
|
|
uint32_t flush_samples,
|
|
double* output_stereo_interleaved,
|
|
uint32_t capacity);
|
|
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|