mt8195 fixups #8691
mt8195 fixups #8691
Conversation
andyross
requested review from
ranj063,
yaochunhung,
kuanhsuncheng,
a team,
lgirdwood,
plbossart,
mmaka1,
lbetlej,
dbaluta and
kv2019i
as code owners
src/audio/rtnr/rtnr.c
Outdated
| @@ -843,7 +843,6 @@ static int rtnr_prepare(struct comp_dev *dev) | |||
| { | |||
| struct comp_data *cd = comp_get_drvdata(dev); | |||
| struct comp_buffer *sinkb; | |||
| struct comp_buffer *sink_c; | |||
There was a problem hiding this comment.
(Which... implies that RTNR isn't being built anywhere in CI?)
This code is compiled in the IPC3 stub but I just tested that again and clang version 16 (and others) does not complain about unused variables in that function. Mysteries of compiler warnings...
RTNR is not compiled in app/stub_build_all_ipc4.conf, feel free to add there.
He is from Google, not MTK. We need @thesofproject/mediatek to respond, we will ping through other channels next week if we don't get a response by monday. |
| #define PLATFORM_TASK_DEFAULT_STACK_SIZE 3072 | ||
| #endif |
There was a problem hiding this comment.
wow, that's a bit heavy-handed, but well, if it works...
There was a problem hiding this comment.
It's actually the only place to put a stack size in XTOS SOF. FWIW the same kind of issue exists with the DP scheduler too, which shares a single worker thread for a core and thus has to be sized to the largest stack consumer that it might run, which is likewise done with a global setting, though there AEC was the first consumer so it got an 8k stack in the first version.
|
@andyross can we rebase this? |
AEC initialization is stack-hungry, needing ~6kb during initialization. That happens under the IPC task, not a component-local stack. So bump the default for all tasks as that's the only way to make sure IPC gets it. (Really this shouldn't be a platform choice, but that's where the symbol exists) Signed-off-by: Andy Ross <andyross@google.com>
Add the .dot file header/footer macros. These are benign if unused, and it seems like most other devices have them defined. Signed-off-by: Andy Ross <andyross@google.com>
The mocking layer is an important debugging tool. It shouldn't be on by default lest we accidentally build without the real feature. Signed-off-by: Andy Ross <andyross@google.com>
|
Prune some stale patches that merged via other means. The stack patch is here is still needed to have working AEC. The other two are just hygine. |
Various fixups required to resurrect builds for mt8195 in main. Split out from #8571 as they should be uncontroversial.