ASoC: Cold start latency reduction
This patch parallelizes FE (front end) and BE (back end)
when playback/capture path is set up. FE's and BE's which can
be parallelized should be indicated in corresponding machine file.
All the operations can be started asynchronously provided the usecase
supports asynchronous operation.
Parallelizing reduces the startup latency as both FE and BE operations
run in parallel.
Change-Id: I17e7e1c3d406713cc728ec262cfe388f7251f7de
Signed-off-by: Walter Yang <yandongy@codeaurora.org>
Signed-off-by: Fred Oh <fred@codeaurora.org>
Signed-off-by: Banajit Goswami <bgoswami@codeaurora.org>
diff --git a/sound/soc/soc-pcm.c b/sound/soc/soc-pcm.c
index 2bc3135..008cf28 100644
--- a/sound/soc/soc-pcm.c
+++ b/sound/soc/soc-pcm.c
@@ -2008,6 +2008,81 @@
return 0;
}
+int dpcm_fe_dai_hw_params_be(struct snd_soc_pcm_runtime *fe,
+ struct snd_soc_pcm_runtime *be,
+ struct snd_pcm_hw_params *params, int stream)
+{
+ int ret;
+ struct snd_soc_dpcm *dpcm;
+ struct snd_pcm_substream *be_substream =
+ snd_soc_dpcm_get_substream(be, stream);
+
+ /* is this op for this BE ? */
+ if (!snd_soc_dpcm_be_can_update(fe, be, stream))
+ return 0;
+
+ /* only allow hw_params() if no connected FEs are running */
+ if (!snd_soc_dpcm_can_be_params(fe, be, stream))
+ return 0;
+
+ if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_OPEN) &&
+ (be->dpcm[stream].state !=
+ SND_SOC_DPCM_STATE_HW_PARAMS) &&
+ (be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_FREE))
+ return 0;
+
+ dev_dbg(be->dev, "ASoC: hw_params BE %s\n",
+ fe->dai_link->name);
+
+ /* perform any hw_params fixups */
+ if (be->dai_link->be_hw_params_fixup) {
+ ret = be->dai_link->be_hw_params_fixup(be,
+ params);
+ if (ret < 0) {
+ dev_err(be->dev,
+ "ASoC: hw_params BE fixup failed %d\n",
+ ret);
+ goto unwind;
+ }
+ }
+
+ ret = soc_pcm_hw_params(be_substream, params);
+ if (ret < 0) {
+ dev_err(be->dev, "ASoC: hw_params BE failed %d\n", ret);
+ goto unwind;
+ }
+
+ be->dpcm[stream].state = SND_SOC_DPCM_STATE_HW_PARAMS;
+ return 0;
+
+unwind:
+ /* disable any enabled and non active backends */
+ list_for_each_entry(dpcm, &fe->dpcm[stream].be_clients, list_be) {
+ struct snd_soc_pcm_runtime *be = dpcm->be;
+ struct snd_pcm_substream *be_substream =
+ snd_soc_dpcm_get_substream(be, stream);
+
+ if (!snd_soc_dpcm_be_can_update(fe, be, stream))
+ continue;
+
+ /* only allow hw_free() if no connected FEs are running */
+ if (!snd_soc_dpcm_can_be_free_stop(fe, be, stream))
+ continue;
+
+ if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_OPEN) &&
+ (be->dpcm[stream].state
+ != SND_SOC_DPCM_STATE_HW_PARAMS) &&
+ (be->dpcm[stream].state
+ != SND_SOC_DPCM_STATE_HW_FREE) &&
+ (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP))
+ continue;
+
+ soc_pcm_hw_free(be_substream);
+ }
+
+ return ret;
+}
+
int dpcm_be_dai_hw_params(struct snd_soc_pcm_runtime *fe, int stream)
{
struct snd_soc_dpcm *dpcm;
@@ -2308,6 +2383,35 @@
return ret;
}
+int dpcm_fe_dai_prepare_be(struct snd_soc_pcm_runtime *fe,
+ struct snd_soc_pcm_runtime *be, int stream)
+{
+ struct snd_pcm_substream *be_substream =
+ snd_soc_dpcm_get_substream(be, stream);
+ int ret = 0;
+
+ /* is this op for this BE ? */
+ if (!snd_soc_dpcm_be_can_update(fe, be, stream))
+ return 0;
+
+ if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_PARAMS) &&
+ (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP))
+ return 0;
+
+ dev_dbg(be->dev, "ASoC: prepare BE %s\n",
+ fe->dai_link->name);
+
+ ret = soc_pcm_prepare(be_substream);
+ if (ret < 0) {
+ dev_err(be->dev, "ASoC: backend prepare failed %d\n",
+ ret);
+ return ret;
+ }
+
+ be->dpcm[stream].state = SND_SOC_DPCM_STATE_PREPARE;
+ return ret;
+}
+
static int dpcm_fe_dai_trigger(struct snd_pcm_substream *substream, int cmd)
{
struct snd_soc_pcm_runtime *fe = substream->private_data;
@@ -2360,13 +2464,90 @@
return ret;
}
+static void dpcm_be_async_prepare(void *data, async_cookie_t cookie)
+{
+ struct snd_soc_dpcm *dpcm = data;
+ struct snd_soc_pcm_runtime *be = dpcm->be;
+ int stream = dpcm->stream;
+ struct snd_pcm_substream *be_substream =
+ snd_soc_dpcm_get_substream(be, stream);
+ int ret;
+
+ dev_dbg(be->dev, "%s ASoC: prepare BE %s\n", __func__,
+ dpcm->fe->dai_link->name);
+ ret = soc_pcm_prepare(be_substream);
+ if (ret < 0) {
+ be->err_ops = ret;
+ dev_err(be->dev, "ASoC: backend prepare failed %d\n",
+ ret);
+ return;
+ }
+ be->dpcm[stream].state = SND_SOC_DPCM_STATE_PREPARE;
+}
+
+void dpcm_be_dai_prepare_async(struct snd_soc_pcm_runtime *fe, int stream,
+ struct async_domain *domain)
+{
+ struct snd_soc_dpcm *dpcm;
+ struct snd_soc_dpcm *dpcm_async[DPCM_MAX_BE_USERS];
+ int i = 0, j;
+
+ list_for_each_entry(dpcm, &fe->dpcm[stream].be_clients, list_be) {
+ struct snd_soc_pcm_runtime *be = dpcm->be;
+
+ be->err_ops = 0;
+ /* is this op for this BE ? */
+ if (!snd_soc_dpcm_be_can_update(fe, be, stream))
+ continue;
+
+ if ((be->dpcm[stream].state != SND_SOC_DPCM_STATE_HW_PARAMS) &&
+ (be->dpcm[stream].state != SND_SOC_DPCM_STATE_STOP))
+ continue;
+
+ /* does this BE support async op ?*/
+ if ((fe->dai_link->async_ops & ASYNC_DPCM_SND_SOC_PREPARE) &&
+ (be->dai_link->async_ops & ASYNC_DPCM_SND_SOC_PREPARE)) {
+ dpcm->stream = stream;
+ async_schedule_domain(dpcm_be_async_prepare,
+ dpcm, domain);
+ } else {
+ dpcm_async[i++] = dpcm;
+ }
+ }
+
+ for (j = 0; j < i; j++) {
+ struct snd_soc_dpcm *dpcm = dpcm_async[j];
+ struct snd_soc_pcm_runtime *be = dpcm->be;
+ struct snd_pcm_substream *be_substream =
+ snd_soc_dpcm_get_substream(be, stream);
+ int ret;
+
+ dev_dbg(be->dev, "ASoC: prepare BE %s\n",
+ dpcm->fe->dai_link->name);
+
+ ret = soc_pcm_prepare(be_substream);
+ if (ret < 0) {
+ dev_err(be->dev, "ASoC: backend prepare failed %d\n",
+ ret);
+ be->err_ops = ret;
+ return;
+ }
+
+ be->dpcm[stream].state = SND_SOC_DPCM_STATE_PREPARE;
+ }
+}
+
static int dpcm_fe_dai_prepare(struct snd_pcm_substream *substream)
{
struct snd_soc_pcm_runtime *fe = substream->private_data;
+ struct snd_soc_dpcm *dpcm;
int stream = substream->stream, ret = 0;
+ ASYNC_DOMAIN_EXCLUSIVE(async_domain);
mutex_lock_nested(&fe->card->mutex, SND_SOC_CARD_CLASS_RUNTIME);
+ fe->err_ops = 0;
+
dev_dbg(fe->dev, "ASoC: prepare FE %s\n", fe->dai_link->name);
dpcm_set_fe_update_state(fe, stream, SND_SOC_DPCM_UPDATE_FE);
@@ -2379,16 +2560,47 @@
goto out;
}
- ret = dpcm_be_dai_prepare(fe, substream->stream);
- if (ret < 0)
- goto out;
+ if (!(fe->dai_link->async_ops & ASYNC_DPCM_SND_SOC_PREPARE)) {
+ ret = dpcm_be_dai_prepare(fe, substream->stream);
+ if (ret < 0)
+ goto out;
+ /* call prepare on the frontend */
+ ret = soc_pcm_prepare(substream);
+ if (ret < 0) {
+ dev_err(fe->dev, "ASoC: prepare FE %s failed\n",
+ fe->dai_link->name);
+ goto out;
+ }
+ } else {
+ dpcm_be_dai_prepare_async(fe, substream->stream,
+ &async_domain);
- /* call prepare on the frontend */
- ret = soc_pcm_prepare(substream);
- if (ret < 0) {
- dev_err(fe->dev,"ASoC: prepare FE %s failed\n",
- fe->dai_link->name);
- goto out;
+ /* call prepare on the frontend */
+ ret = soc_pcm_prepare(substream);
+ if (ret < 0) {
+ fe->err_ops = ret;
+ dev_err(fe->dev, "ASoC: prepare FE %s failed\n",
+ fe->dai_link->name);
+ }
+
+ async_synchronize_full_domain(&async_domain);
+
+ /* check if any BE failed */
+ list_for_each_entry(dpcm, &fe->dpcm[stream].be_clients,
+ list_be) {
+ struct snd_soc_pcm_runtime *be = dpcm->be;
+
+ if (be->err_ops < 0) {
+ ret = be->err_ops;
+ goto out;
+ }
+ }
+
+ /* check if FE failed */
+ if (fe->err_ops < 0) {
+ ret = fe->err_ops;
+ goto out;
+ }
}
/* run the stream event for each BE */