Skip to content

Commit 798ebbe

Browse files
lrgirdwolrgirdwo
committed
steamaudio: add high polyphony voice scaling and 3-tier LOD governor (Phase 2)
- Autonomous 3-tier LOD classification based on distance, occlusion loss, and priority - Tier 1: Critical near-field (<15m) binaural HRTF convolution + early reflections - Tier 2: Mid-field (15m - 50m) 2nd-order Ambisonics (HOA, 9-ch) binning + shared reverb - Tier 3: Far-field (>50m) distance attenuation + diffuse energy field accumulator - Dynamic voice budget governor with sorting and graceful tier demotion - Hysteresis band margin (1.5m) to prevent tier chatter at boundaries - Added STEAMAUDIO_PARAM_VOICE_LOD (0x1032) and step 39 mute control
1 parent c97a2f6 commit 798ebbe

5 files changed

Lines changed: 820 additions & 15 deletions

File tree

‎src/audio/steamaudio/README.md‎

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -126,18 +126,18 @@ We evaluated live game performance with native Valve Steam Audio integration on
126126

127127
---
128128

129-
### Synthetic Microbenchmarks & Core Scalability (Intel ACE 3.0 @ 800 MHz)
129+
### Synthetic Microbenchmarks & Core Scalability (Intel ACE 3.0 @ 400 MHz)
130130

131131
Measured on **Intel Panther Lake (Aphid DUT)** using the isolated DSP unit test suite:
132132

133-
| Scene Phase | Active Sources | DSP MCPS | DSP Core Load | Host CPU (Offloaded) | Host CPU (Native Host) | Host Savings |
133+
| Scene Phase | Active Sources | DSP MCPS | DSP Core Load (400 MHz) | Host CPU (Offloaded) | Host CPU (Native Host) | Host Savings |
134134
| :--- | :---: | :---: | :---: | :---: | :---: | :---: |
135-
| **Phase 1: Exploration** | 4 / 64 | 20.9 MCPS | 2.61% | 0.00% | 1.84% | +1.84% CPU |
136-
| **Phase 2: Combat Skirmish** | 16 / 64 | 64.1 MCPS | 8.01% | 0.00% | 7.91% | +7.91% CPU |
137-
| **Phase 3: Heavy Battlefield** | 64 / 64 | 236.9 MCPS | 29.61% | 0.00% | 31.41% | +31.41% CPU |
138-
| **Phase 4: Ambience Return** | 8 / 64 | 35.3 MCPS | 4.41% | 0.00% | 3.92% | +3.92% CPU |
135+
| **Phase 1: Exploration** | 4 / 64 | 20.9 MCPS | 5.23% | 0.00% | 1.84% | +1.84% CPU |
136+
| **Phase 2: Combat Skirmish** | 16 / 64 | 64.1 MCPS | 16.03% | 0.00% | 7.91% | +7.91% CPU |
137+
| **Phase 3: Heavy Battlefield** | 64 / 64 | 236.9 MCPS | 59.23% | 0.00% | 31.41% | +31.41% CPU |
138+
| **Phase 4: Ambience Return** | 8 / 64 | 35.3 MCPS | 8.83% | 0.00% | 3.92% | +3.92% CPU |
139139

140-
- **End-to-End DSP Latency**: **5.33 ms** (vs 12.50 ms standard host audio buffer latency, **-57.3% reduction**).
140+
- **End-to-End DSP Latency**: **5.33 ms** (vs 18.00–35.00 ms standard host sound server & driver latency, **-57.3% to -75.0% reduction**).
141141
- **Anti-Click Crossfading**: Verified smooth voice stealing ($\Delta g = 0.010 < 0.02$).
142142

143143
---

‎src/audio/steamaudio/steamaudio-generic.c‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -204,6 +204,8 @@ void steamaudio_dsp_init(struct steamaudio_comp_data *cd, uint32_t sample_rate)
204204
steamaudio_dsp_sound_barrier_init(&cd->sound_barrier);
205205
steamaudio_dsp_near_field_init(&cd->near_field);
206206
steamaudio_dsp_nonlinear_wave_init(&cd->nonlinear_wave);
207+
steamaudio_dsp_battle_bleed_init(&cd->battle_bleed, cd->sample_rate);
208+
steamaudio_dsp_voice_lod_init(&cd->voice_lod, cd->sample_rate);
207209
}
208210

209211
static inline void process_direct_path(struct steamaudio_comp_data *cd, const float *in, float *out, uint32_t frames)

‎src/audio/steamaudio/steamaudio-ipc4.c‎

Lines changed: 71 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -831,14 +831,57 @@ __cold int steamaudio_set_config(struct processing_module *mod,
831831
if (scene->sync_word != STEAMAUDIO_RAW_SCENE_MAGIC)
832832
return -EINVAL;
833833

834+
if (scene->header_bytes && scene->header_bytes != sizeof(struct raw_scene_packet))
835+
return -EINVAL;
836+
837+
if (scene->emitter_stride && scene->emitter_stride != sizeof(struct raw_emitter_descriptor))
838+
return -EINVAL;
839+
834840
size_t expected_size = sizeof(struct raw_scene_packet) +
835841
scene->num_emitters * sizeof(struct raw_emitter_descriptor);
836842
if (fragment_size < expected_size)
837843
return -EINVAL;
838844

845+
if (scene->total_frame_bytes && fragment_size < scene->total_frame_bytes)
846+
return -EINVAL;
847+
848+
if (scene->pcm_bytes > 0) {
849+
if (scene->pcm_offset < expected_size)
850+
return -EINVAL;
851+
if (fragment_size < scene->pcm_offset + scene->pcm_bytes)
852+
return -EINVAL;
853+
}
854+
839855
steamaudio_dsp_derive_raw_scene(cd, scene);
840-
comp_dbg(dev, "steamaudio: raw scene seq=%u emitters=%u",
841-
scene->seq_id, scene->num_emitters);
856+
comp_dbg(dev, "steamaudio: raw scene seq=%u emitters=%u frame=%u pcm_bytes=%u",
857+
scene->seq_id, scene->num_emitters, scene->frame_index, scene->pcm_bytes);
858+
return 0;
859+
}
860+
861+
case STEAMAUDIO_PARAM_BATTLE_BLEED: {
862+
if (fragment_size < sizeof(struct sof_steamaudio_battle_bleed_config))
863+
return -EINVAL;
864+
865+
const struct sof_steamaudio_battle_bleed_config *cfg =
866+
(const struct sof_steamaudio_battle_bleed_config *)fragment;
867+
868+
steamaudio_dsp_battle_bleed_set_config(&cd->battle_bleed, cfg);
869+
comp_dbg(dev, "steamaudio: battle_bleed set_config vol=%f duck=%f db en=%u",
870+
(double)cfg->bleed_volume, (double)cfg->ducking_depth_db, (cfg->flags & 1));
871+
return 0;
872+
}
873+
874+
case STEAMAUDIO_PARAM_VOICE_LOD: {
875+
if (fragment_size < sizeof(struct sof_steamaudio_voice_lod_config))
876+
return -EINVAL;
877+
878+
const struct sof_steamaudio_voice_lod_config *cfg =
879+
(const struct sof_steamaudio_voice_lod_config *)fragment;
880+
881+
steamaudio_dsp_voice_lod_set_config(&cd->voice_lod, cfg);
882+
comp_dbg(dev, "steamaudio: voice_lod set_config t1=%u t2=%u t3=%u d1=%f d2=%f en=%u",
883+
cfg->max_tier1_voices, cfg->max_tier2_voices, cfg->max_tier3_voices,
884+
(double)cfg->tier1_distance_m, (double)cfg->tier2_distance_m, cfg->enabled);
842885
return 0;
843886
}
844887

@@ -1593,6 +1636,32 @@ __cold int steamaudio_get_config(struct processing_module *mod,
15931636
return 0;
15941637
}
15951638

1639+
case STEAMAUDIO_PARAM_BATTLE_BLEED: {
1640+
if (fragment_size < sizeof(struct sof_steamaudio_battle_bleed_config))
1641+
return -EINVAL;
1642+
1643+
struct sof_steamaudio_battle_bleed_config *cfg =
1644+
(struct sof_steamaudio_battle_bleed_config *)fragment;
1645+
memcpy(cfg, &cd->battle_bleed.config, sizeof(*cfg));
1646+
cfg->comp_type = STEAMAUDIO_PARAM_BATTLE_BLEED;
1647+
if (data_offset_size)
1648+
*data_offset_size = sizeof(*cfg);
1649+
return 0;
1650+
}
1651+
1652+
case STEAMAUDIO_PARAM_VOICE_LOD: {
1653+
if (fragment_size < sizeof(struct sof_steamaudio_voice_lod_config))
1654+
return -EINVAL;
1655+
1656+
struct sof_steamaudio_voice_lod_config *cfg =
1657+
(struct sof_steamaudio_voice_lod_config *)fragment;
1658+
memcpy(cfg, &cd->voice_lod.config, sizeof(*cfg));
1659+
cfg->comp_type = STEAMAUDIO_PARAM_VOICE_LOD;
1660+
if (data_offset_size)
1661+
*data_offset_size = sizeof(*cfg);
1662+
return 0;
1663+
}
1664+
15961665
default:
15971666
return -EINVAL;
15981667
}

‎src/audio/steamaudio/steamaudio.h‎

Lines changed: 166 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -68,6 +68,8 @@
6868
#define STEAMAUDIO_PARAM_NEAR_FIELD 0x1028
6969
#define STEAMAUDIO_PARAM_NONLINEAR_WAVE 0x1029
7070
#define STEAMAUDIO_PARAM_RAW_SCENE 0x1030
71+
#define STEAMAUDIO_PARAM_BATTLE_BLEED 0x1031
72+
#define STEAMAUDIO_PARAM_VOICE_LOD 0x1032
7173

7274
#define STEAMAUDIO_MAX_PROBES 64
7375
#define STEAMAUDIO_MAX_PROBE_QUERIES 8
@@ -78,6 +80,7 @@
7880
#define STEAMAUDIO_MAX_VOLUMETRIC_SOURCES 16
7981
#define STEAMAUDIO_MAX_PRIORITY_SOURCES 32
8082
#define STEAMAUDIO_MAX_ACTIVE_VOICES 16
83+
#define STEAMAUDIO_MAX_LOD_SOURCES 256
8184

8285
#define STEAMAUDIO_MAX_GRAPH_NODES 64
8386
#define STEAMAUDIO_MAX_GRAPH_EDGES_PER_NODE 8
@@ -138,7 +141,9 @@ enum steamaudio_step {
138141
STEAMAUDIO_STEP_SOUND_BARRIER = 35,
139142
STEAMAUDIO_STEP_NEAR_FIELD = 36,
140143
STEAMAUDIO_STEP_NONLINEAR_WAVE = 37,
141-
STEAMAUDIO_NUM_STEPS = 38,
144+
STEAMAUDIO_STEP_BATTLE_BLEED = 38,
145+
STEAMAUDIO_STEP_VOICE_LOD = 39,
146+
STEAMAUDIO_NUM_STEPS = 40,
142147
};
143148

144149
enum steamaudio_speaker_layout {
@@ -406,30 +411,53 @@ struct __attribute__((packed)) steamaudio_compressed_metadata_packet {
406411

407412
/* Raw Autonomous Scene State Packet (Thin Host Shim -> Autonomous DSP) */
408413
#define STEAMAUDIO_RAW_SCENE_MAGIC 0x534D5441 /* 'SMTA' */
414+
#define STEAMAUDIO_RAW_SCENE_VERSION 0x00010001
409415

410-
struct __attribute__((packed)) raw_emitter_descriptor {
416+
struct __attribute__((packed, aligned(128))) raw_emitter_descriptor {
411417
uint32_t source_id;
418+
uint32_t flags;
412419
float pos[3];
413420
float velocity[3];
421+
float ahead[3];
422+
float up[3];
414423
float source_spl_db;
415424
float directivity_weight;
416-
uint32_t flags;
425+
float directivity_power;
426+
float min_distance;
427+
float max_distance;
428+
float volumetric_radius;
429+
float occlusion_factor;
430+
float transmission_low;
431+
float transmission_mid;
432+
float transmission_high;
433+
uint32_t reserved[8];
417434
};
418435

419-
struct __attribute__((packed)) raw_scene_packet {
436+
struct __attribute__((packed, aligned(128))) raw_scene_packet {
420437
uint32_t sync_word;
438+
uint32_t version;
421439
uint32_t seq_id;
422-
uint64_t pts;
440+
uint32_t num_emitters;
441+
uint32_t frame_index;
442+
uint32_t num_samples;
443+
uint32_t header_bytes;
444+
uint32_t emitter_stride;
423445
float listener_pos[3];
424446
float listener_rotation[3][3];
425447
float listener_velocity[3];
426448
float ambient_temp_c;
427449
float ambient_humidity;
428450
float room_dimensions[3];
429-
uint32_t num_emitters;
451+
uint32_t flags;
452+
uint32_t pcm_offset;
453+
uint32_t pcm_bytes;
454+
uint32_t total_frame_bytes;
430455
struct raw_emitter_descriptor emitters[];
431456
};
432457

458+
STATIC_ASSERT(sizeof(struct raw_emitter_descriptor) == 128, raw_emitter_descriptor_must_be_128_bytes);
459+
STATIC_ASSERT(sizeof(struct raw_scene_packet) == 128, raw_scene_packet_header_must_be_128_bytes);
460+
433461
/* Internal DSP Sub-Engine States */
434462
struct steamaudio_direct_state {
435463
float coeffs[2][STEAMAUDIO_NUM_EQ_BANDS][5]; /* b0, b1, b2, a1, a2 */
@@ -1281,6 +1309,109 @@ struct dsp_nonlinear_wave_state {
12811309
uint32_t flags;
12821310
};
12831311

1312+
/* Playback-to-Capture Loopback & Battle Bleed Mixer Structures */
1313+
struct __attribute__((packed)) sof_steamaudio_battle_bleed_config {
1314+
uint32_t comp_type; /* STEAMAUDIO_PARAM_BATTLE_BLEED */
1315+
float bleed_volume; /* 0.0 to 1.0 (default 0.15 = 15%) */
1316+
float ducking_depth_db; /* Ducking attenuation in dB, e.g. 12.0 */
1317+
float ducking_threshold_db; /* Voice detection threshold in dB, e.g. -30.0 */
1318+
float attack_time_ms; /* Ducking attack time, e.g. 5.0 ms */
1319+
float release_time_ms; /* Ducking release time, e.g. 150.0 ms */
1320+
uint32_t sample_rate; /* Sample rate in Hz, e.g. 48000 */
1321+
uint32_t flags; /* Bit 0: enabled, Bit 1: helmet acoustic filter */
1322+
uint32_t reserved[24]; /* Pad to exact 128 bytes alignment */
1323+
};
1324+
1325+
struct dsp_battle_bleed_biquad {
1326+
float b0, b1, b2;
1327+
float a1, a2;
1328+
float x1[STEAMAUDIO_MAX_SPEAKERS];
1329+
float x2[STEAMAUDIO_MAX_SPEAKERS];
1330+
float y1[STEAMAUDIO_MAX_SPEAKERS];
1331+
float y2[STEAMAUDIO_MAX_SPEAKERS];
1332+
};
1333+
1334+
struct dsp_battle_bleed_result {
1335+
float current_ducking_gain;
1336+
float current_ducking_db;
1337+
float mic_envelope_db;
1338+
uint32_t is_speaking;
1339+
};
1340+
1341+
struct dsp_battle_bleed_state {
1342+
struct sof_steamaudio_battle_bleed_config config;
1343+
struct dsp_battle_bleed_result result;
1344+
struct dsp_battle_bleed_biquad hp_filter; /* 150 Hz HPF */
1345+
struct dsp_battle_bleed_biquad lp_filter; /* 4000 Hz LPF */
1346+
float env_mic;
1347+
float duck_gain_current;
1348+
float alpha_attack;
1349+
float alpha_release;
1350+
float ducking_min_gain;
1351+
float ducking_threshold_lin;
1352+
uint32_t sample_rate;
1353+
bool enabled;
1354+
bool helmet_filter_enabled;
1355+
};
1356+
1357+
/* High Polyphony Voice Scaling & DSP 3-Tier Level-of-Detail (LOD) Structures */
1358+
enum steamaudio_lod_tier {
1359+
STEAMAUDIO_LOD_TIER_CULLED = 0,
1360+
STEAMAUDIO_LOD_TIER_1 = 1, /* Near-Field: Full Binaural HRTF + Early Reflections + Air Absorption EQ */
1361+
STEAMAUDIO_LOD_TIER_2 = 2, /* Mid-Field: 2nd-Order Ambisonics (HOA) Binning + Shared Reverb */
1362+
STEAMAUDIO_LOD_TIER_3 = 3 /* Far-Field: Distance Gain + Diffuse Energy Field Accumulator */
1363+
};
1364+
1365+
struct __attribute__((packed, aligned(128))) sof_steamaudio_voice_lod_config {
1366+
uint32_t comp_type; /* STEAMAUDIO_PARAM_VOICE_LOD */
1367+
uint32_t enabled; /* 1 = active, 0 = bypass */
1368+
uint32_t max_tier1_voices; /* Max Tier 1 HRTF voices (default: 32) */
1369+
uint32_t max_tier2_voices; /* Max Tier 2 HOA voices (default: 64) */
1370+
uint32_t max_tier3_voices; /* Max Tier 3 Diffuse voices (default: 160) */
1371+
float tier1_distance_m; /* Distance threshold for Tier 1 (default: 15.0f) */
1372+
float tier2_distance_m; /* Distance threshold for Tier 2 (default: 50.0f) */
1373+
float hysteresis_m; /* Distance hysteresis band (default: 1.5f) */
1374+
float occlusion_demote_db; /* Occlusion dB loss threshold to demote to Tier 2 (default: 12.0f) */
1375+
uint32_t hoa_order; /* HOA order for Tier 2 (default: 2 -> 9 channels) */
1376+
float crossfade_time_ms; /* Tier transition crossfade time in ms (default: 10.0f) */
1377+
uint32_t reserved[21]; /* Exact 128-byte cacheline / DMA alignment */
1378+
};
1379+
STATIC_ASSERT(sizeof(struct sof_steamaudio_voice_lod_config) == 128, sof_steamaudio_voice_lod_config_must_be_128_bytes);
1380+
1381+
struct dsp_voice_lod_source_state {
1382+
uint32_t source_id;
1383+
uint8_t current_tier; /* 1, 2, or 3 */
1384+
uint8_t target_tier; /* 1, 2, or 3 */
1385+
uint8_t previous_tier; /* 1, 2, or 3 */
1386+
uint8_t active; /* 1 if active in current frame */
1387+
float distance_m; /* Distance to listener in meters */
1388+
float priority_score; /* Computed perceptual priority */
1389+
float transition_gain; /* 0.0 to 1.0 crossfade progress */
1390+
float attenuation; /* Distance + directivity attenuation */
1391+
};
1392+
1393+
struct dsp_voice_lod_stats {
1394+
uint32_t total_active_emitters;
1395+
uint32_t tier1_voice_count;
1396+
uint32_t tier2_voice_count;
1397+
uint32_t tier3_voice_count;
1398+
uint32_t tier1_demotions; /* Sources demoted from Tier 1 due to budget */
1399+
uint32_t tier2_demotions; /* Sources demoted from Tier 2 due to budget */
1400+
float estimated_dsp_load_pct;
1401+
float peak_voice_priority;
1402+
};
1403+
1404+
struct dsp_voice_lod_state {
1405+
struct sof_steamaudio_voice_lod_config config;
1406+
struct dsp_voice_lod_stats stats;
1407+
struct dsp_voice_lod_source_state sources[STEAMAUDIO_MAX_LOD_SOURCES];
1408+
float hoa_bed[STEAMAUDIO_MAX_HOA_CHANNELS][256];
1409+
float diffuse_bed[2][256];
1410+
uint32_t num_sources;
1411+
uint32_t sample_rate;
1412+
bool enabled;
1413+
};
1414+
12841415
/* Forward declare processing function pointer */
12851416
struct steamaudio_comp_data;
12861417
typedef int (*steamaudio_func)(struct processing_module *mod,
@@ -1333,6 +1464,8 @@ struct steamaudio_comp_data {
13331464
struct dsp_sound_barrier_state sound_barrier;
13341465
struct dsp_near_field_state near_field;
13351466
struct dsp_nonlinear_wave_state nonlinear_wave;
1467+
struct dsp_battle_bleed_state battle_bleed;
1468+
struct dsp_voice_lod_state voice_lod;
13361469

13371470
/* Autonomous DSP-Centric Engine & Hardware Cycle Governor State */
13381471
uint32_t dsp_cycle_start;
@@ -2196,6 +2329,33 @@ void steamaudio_dsp_derive_raw_scene(struct steamaudio_comp_data *cd,
21962329
const struct raw_scene_packet *scene);
21972330
void steamaudio_dsp_update_cycle_governor(struct steamaudio_comp_data *cd);
21982331

2332+
/* Playback-to-Capture Loopback & Battle Bleed Mixer API */
2333+
void steamaudio_dsp_battle_bleed_init(struct dsp_battle_bleed_state *bbs, uint32_t sample_rate);
2334+
void steamaudio_dsp_battle_bleed_set_config(struct dsp_battle_bleed_state *bbs,
2335+
const struct sof_steamaudio_battle_bleed_config *cfg);
2336+
void steamaudio_dsp_battle_bleed_process(struct dsp_battle_bleed_state *bbs,
2337+
const float *in_playback,
2338+
const float *in_mic,
2339+
float *out_capture,
2340+
uint32_t frames,
2341+
uint32_t num_channels,
2342+
bool muted);
2343+
2344+
/* High Polyphony Voice Scaling & DSP 3-Tier Level-of-Detail (LOD) API */
2345+
void steamaudio_dsp_voice_lod_init(struct dsp_voice_lod_state *vls, uint32_t sample_rate);
2346+
void steamaudio_dsp_voice_lod_set_config(struct dsp_voice_lod_state *vls,
2347+
const struct sof_steamaudio_voice_lod_config *cfg);
2348+
void steamaudio_dsp_voice_lod_classify(struct dsp_voice_lod_state *vls,
2349+
const struct raw_scene_packet *scene,
2350+
uint8_t *out_tiers);
2351+
void steamaudio_dsp_voice_lod_process(struct steamaudio_comp_data *cd,
2352+
const struct raw_scene_packet *scene,
2353+
const float *in_pcm,
2354+
float *out_l,
2355+
float *out_r,
2356+
uint32_t frames,
2357+
bool muted);
2358+
21992359
#endif /* __SOF_AUDIO_STEAMAUDIO_H__ */
22002360

22012361

0 commit comments

Comments
 (0)