Horizon dev - #34
Open
paresh-bhagat wants to merge 5 commits into
Open
Conversation
- Extract pipeline files: audio_enhancement, stft_istft, tvm, audio_utils, pipeline_common - Rename GenericTaskClient -> DspTaskClient; remove interactive CLI mode - Add pipeline_type field for dispatch; remove pipeline_id and model_config - Move model params (hop_size, model_elems, total_frames, batch_n) to stage parameters - Add dsp_config (proc_id, endpoint) to JSON; remove all hardcoded RPMsg constants - Size STFT/ISTFT DMA buffers independently from their respective stage parameters Signed-off-by: Paresh Bhagat <p-bhagat@ti.com>
Replace non-overlapping chunk processing with overlap-save approach: - OVERLAP_FRAMES=100, HOP_FRAMES=301, T_FRAMES=50 - Consecutive chunks overlap by 100 frames (1600 samples at 160 hop) - trim_reconstruct: first chunk keeps [0..end-T], middle chunks keep [T..end-T], last chunk keeps [T..real_end] Signed-off-by: Paresh Bhagat <p-bhagat@ti.com>
Introduces a Unix domain socket daemon (demo-manager) that manages the lifecycle of DSP demos on the board. The daemon accepts JSON commands (list/run/stop/status/preload/quit) over /var/run/demo-manager.sock and handles two demo classes: - edge-ai: automatically preloads the TVM model before launch to eliminate cold-start latency from Module::LoadFromFile - dsp-compute (2dfft, audio-offload, sigchain-biquad): invalidates the TVM model cache after completion, since these demos load their own DSP firmware and overwrite TVM runtime state Also adds: - demo-ctl: CLI client for sending commands to the daemon - tvm_model_daemon: persistent service that keeps the TVM model loaded and serves inference requests over a socket, enabling the fast-path in TvmInferenceClient - systemd service files for demo-manager and tvm-model-daemon - TvmInferenceClient: fast-path to delegate to daemon, avoiding per-run model reload; synchronize_dma_buffer now takes an explicit fd Signed-off-by: Vishnu Singh <v-singh1@ti.com> Signed-off-by: Paresh Bhagat <p-bhagat@ti.com>
- Rename stft_process_msg -> stft_istft_msg, remove redundant graph_id with selected_model. - Add selected_model to STFT/ISTFT stage parameters in JSON files Signed-off-by: Paresh Bhagat <p-bhagat@ti.com>
Add a new pipeline type 'audio_classification' for analysis-only models (YAMNet, VGGish) that run STFT log-mel extraction without ISTFT synthesis. Pipeline details: selected_model=4, hop_size=160, model_elems=64, total_frames=96 Signed-off-by: Paresh Bhagat <p-bhagat@ti.com>
paresh-bhagat
force-pushed
the
horizon-dev
branch
from
August 24, 2026 12:38
a84fa79 to
67576bb
Compare
v-singh1
reviewed
Aug 25, 2026
| @@ -0,0 +1,14 @@ | |||
| #ifndef AUDIO_ENHANCEMENT_PIPELINE_H | |||
Collaborator
There was a problem hiding this comment.
change file name to Speech Enhancement pipeline
| @@ -1,26 +1,62 @@ | |||
| { | |||
Collaborator
There was a problem hiding this comment.
rename speech enhancement
| @@ -1,26 +1,62 @@ | |||
| { | |||
| "pipeline_id": "audio_gcrn_pipeline", | |||
| "pipeline_type": "audio_enhancement", | |||
Collaborator
There was a problem hiding this comment.
rename speech enhancement
| @@ -0,0 +1,413 @@ | |||
| #include "audio_enhancement_pipeline.h" | |||
| }; | ||
|
|
||
| struct stft_process_msg { | ||
| struct stft_istft_msg { |
|
|
||
| static_assert(sizeof(c7x_msg_hdr) == 16); | ||
| static_assert(sizeof(stft_process_msg) == 36); | ||
| static_assert(sizeof(stft_istft_msg) == 36); |
Collaborator
There was a problem hiding this comment.
no stft/istft, rename it
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.