iframe-proxy

jsarha · 2026-06-04T22:11:50Z

Refactor SRC and SRC-lite so that filter stage setup and delay line allocation happen during the module init callback instead of prepare. This ensures the bulk of SRC memory allocation occurs while the vregion allocator is still in its lifetime phase, before the interim heap is created. The allocations then persist across prepare/reset cycles without needing to be re-allocated each time.

A new setup_stages() callback is added to struct comp_data, set by each variant (src.c, src_lite.c) to point at its own coefficient tables. The common src_allocate_delay_lines() is factored out of the old prepare path into src_common.c.

For IPC4, src_init_stages() calls setup_stages() and src_allocate_delay_lines() at init time. The prepare path (src_prepare_do) only validates rates and sets downstream params.

For IPC3, src_init_stages() is a no-op and src_prepare_do() retains the original behavior of doing full setup at prepare time, since IPC3 cannot be tested at this time.

Copilot

Pull request overview

This PR refactors the SRC and SRC-lite modules so that filter-stage setup and delay-line allocation can happen during module init (IPC4) rather than during prepare, allowing the bulk of SRC allocations to occur earlier (while the vregion allocator is still in its lifetime phase) and persist across prepare/reset cycles.

Changes:

Add a setup_stages() callback on struct comp_data so each SRC variant can bind its own coefficient tables during init.
Factor delay-line allocation and polyphase initialization into src_allocate_delay_lines() in src_common.c.
Split IPC behavior: IPC4 allocates at init (src_init_stages()), while IPC3 keeps full setup in prepare (src_prepare_do()).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

jsarha · 2026-07-02T17:35:45Z

+		uint32_t frame_bytes = cd->channels_count * cd->sample_container_bytes;
+
+		if (frame_bytes && cd->sink_rate) {
+			dev->period = 1000000ULL *
+				(cd->ipc_config.base.obs / frame_bytes) /
+				cd->sink_rate;
+			dev->period /= LL_TIMER_PERIOD_US;
+			dev->period *= LL_TIMER_PERIOD_US;
+			component_set_nearest_period_frames(dev, cd->sink_rate);
+		}


jsarha · 2026-07-02T17:36:18Z

+	struct comp_data *cd = module_get_private_data(mod);
+	struct comp_dev *dev = mod->dev;
+	int ret;
+
+	if (cd->source_rate != cd->ipc_config.base.audio_fmt.sampling_frequency ||
+	    cd->sink_rate != cd->ipc_config.sink_rate) {
+		comp_err(mod->dev, "rate mismatch: source %u/%u sink %u/%u",
+			 cd->source_rate,
+			 cd->ipc_config.base.audio_fmt.sampling_frequency,
+			 cd->sink_rate, cd->ipc_config.sink_rate);
+		return -EINVAL;
+	}
+
+	ret = src_set_params(mod, sink);
+	if (ret < 0) {
+		comp_err(mod->dev, "set params failed.");
+		return ret;
+	}
+
+	/* Update frame counts with final dev->frames from src_set_params */
+	cd->source_frames = dev->frames * cd->source_rate / cd->sink_rate;
+	cd->sink_frames = dev->frames;
+
+	return src_prepare_general(mod, source, sink);
+}


Two followup commit to address this and more complex issues arising from it. But in the way out Linux driver uses SOF, thoise two followup commits, would not be strictly needed.

lyakh · 2026-06-05T07:10:03Z

+}
+
+/* IPC3: Full filter setup at prepare time */
+int src_prepare_do(struct processing_module *mod,


let's make up our mind - "do_prepare" or "prepare_do?" :-)

...to match src_do_init()

lyakh · 2026-06-05T07:13:54Z

+ * and stage pointers (stage1, stage2) are already set up via
+ * cd->setup_stages().
+ */
+int src_allocate_delay_lines(struct processing_module *mod)


why in src_common.c if it's only used with IPC4? Looks like this is duplicating code now for IPC3 and IPC4. Can we not extract and reuse some common functions?

Common is more correct place for this, but I did not dare to touch ipc3 as I do not know how to test that code.

lgirdwood · 2026-06-08T11:47:05Z

@singalsu pls take a look.

Refactor SRC and SRC-lite so that filter stage setup and delay line allocation happen during the module init callback instead of prepare. This ensures the bulk of SRC memory allocation occurs while the vregion allocator is still in its lifetime phase, before the interim heap is created. The allocations then persist across prepare/reset cycles without needing to be re-allocated each time. A new setup_stages() callback is added to struct comp_data, set by each variant (src.c, src_lite.c) to point at its own coefficient tables. The common src_allocate_delay_lines() is factored out of the old prepare path into src_common.c. For IPC4, src_init_stages() calls setup_stages() and src_allocate_delay_lines() at init time. The prepare path (src_prepare_do) only validates rates and sets downstream params. For IPC3, src_init_stages() is a no-op and src_prepare_do() retains the original behavior of doing full setup at prepare time, since IPC3 cannot be tested at this time. Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Now that filter stage setup and delay line allocation happen at init time, src_reset() must not destroy that state. Replace the call to src_polyphase_reset() (which NULLs stage pointers and zeroes sizes) with the new src_polyphase_reset_state() that only: - zeroes the delay line memory contents (clearing filter history), - resets fir_wp / out_rp pointers to their initial positions, - resets the inter-stage buffer pointers and counter. The filter stage descriptors, delay line pointers, and their sizes are preserved so the next prepare/trigger cycle can resume without re-allocating or re-initializing the filters. Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Check that the source stream rate and channel count at prepare match what was used to initialize the filter stages and delay lines. Since these allocations now happen at init and persist across prepare/reset cycles, a mismatch would cause incorrect processing or buffer overruns. Also compare rates against the actual source stream rather than only the IPC config, catching cases where the upstream pipeline delivers a different rate than originally configured. Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

jsarha · 2026-07-02T18:07:12Z

lyakh

sorry, I think this needs a very thorough review, in particular from @singalsu but I'd also like to try and better understand where various parts come from. I previously commented about code duplication, and the current stats for the first commit, which is supposed to (mostly) move code around are:

 6 files changed, 239 insertions(+), 38 deletions(-)

which looks like a fair bit more than moving around. Let me press the brake on this one for now until we can review it in better detail.

Copilot AI review requested due to automatic review settings June 4, 2026 22:11

jsarha requested review from abonislawski and singalsu as code owners June 4, 2026 22:11

Copilot started reviewing on behalf of jsarha June 4, 2026 22:12 View session

Copilot AI reviewed Jun 4, 2026

View reviewed changes

lyakh reviewed Jun 5, 2026

View reviewed changes

Jyri Sarha added 3 commits July 2, 2026 20:33

jsarha force-pushed the src_move_allocs_to_init branch from 9def2d6 to 056152e Compare July 2, 2026 17:39

lyakh requested changes Jul 3, 2026

View reviewed changes

File	Description
src/audio/src/src.c	Introduces `src_setup_stages()` and an init wrapper to bind coefficient tables and run init-time stage setup.
src/audio/src/src_lite.c	Mirrors SRC changes for SRC-lite via `src_lite_setup_stages()` and init wrapper.
src/audio/src/src_ipc4.c	Adds IPC4 `src_init_stages()` (init-time allocation) and new IPC4 `src_prepare_do()` path.
src/audio/src/src_ipc3.c	Adds IPC3 `src_init_stages()` no-op and keeps full setup/allocation in prepare via `src_prepare_do()`.
src/audio/src/src_common.h	Extends `struct comp_data` with `setup_stages` callback and adds new common function prototypes.
src/audio/src/src_common.c	Adds `src_allocate_delay_lines()` to centralize delay-line allocation and polyphase initialization.

Sunbelt Computer Software

PL/B Language Development and Support

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

audio: src: move filter and delay line allocation to init phase#10846

audio: src: move filter and delay line allocation to init phase#10846
jsarha wants to merge 3 commits into
thesofproject:mainfrom
jsarha:src_move_allocs_to_init

jsarha commented Jun 4, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

jsarha Jul 2, 2026

Uh oh!

jsarha Jul 2, 2026

Uh oh!

lyakh Jun 5, 2026

Uh oh!

lyakh Jun 9, 2026

Uh oh!

lyakh Jun 5, 2026

Uh oh!

jsarha Jul 2, 2026

Uh oh!

lgirdwood commented Jun 8, 2026

Uh oh!

jsarha commented Jul 2, 2026

Uh oh!

lyakh left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Sunbelt Computer Software

PL/B Language Development and Support

Uh oh!

Conversation

jsarha commented Jun 4, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

jsarha Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

jsarha Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

lyakh Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

lyakh Jun 9, 2026

Choose a reason for hiding this comment

Uh oh!

lyakh Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

jsarha Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

lgirdwood commented Jun 8, 2026

Uh oh!

jsarha commented Jul 2, 2026

Uh oh!

lyakh left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants