[jajanusz]

For #1198 #1199 + some more things described in commit messages.

Looks like it has not been working for some time.
I guess that since switching to cmake we were just checking if it builds and if we want to support testbench we cannot continue like this. Aligning to changes introduced over months of development at once is a bit painful (CI builds it anyway, so it can run at least 1 test also, it takes microseconds).

[jajanusz]

@singalsu
Please check if parse_topology still makes sense.

For command:

../../../build_host/install//bin/testbench -d -r 8000 -R 8000 -i chirp_test_in.raw -o chirp_test_out.raw -t ../../build_tools/test/topology/test-playback-ssp2-mclk-0-I2S-src-s16le-s16le-48k-24576k-nocodec.tplg -a src=libsof_src.so -b S16_LE

I was getting:

debug: topology parsing start
debug: type: 8, size: 0x70 count: 1 index: 0
debug: type: 5, size: 0x838 count: 9 index: 1
debug: number of DAPM widgets 9
debug: loading widget codec_out0 id 0
info: Widget type not supported 1
debug: loading widget PCM0P id 1
debug: registered file comp driver
debug: loading widget SRC1.0 id 2
debug: registered comp driver for src
debug: opening shared lib libsof_src.so
SRC src_new()debug: loading widget PGA1.0 id 3
debug: registered comp driver for vol
debug: opening shared lib libsof_volume.so
VOLUME volume_new()debug: loading widget BUF1.0 id 4
BUFFER buffer_new()debug: loading widget BUF1.1 id 5
BUFFER buffer_new()debug: loading widget BUF1.2 id 6
BUFFER buffer_new()debug: loading widget SSP2.OUT id 7
debug: loading widget PIPELINE.1.SSP2.OUT id 8
PIPE pipeline_new()debug: type: 5, size: 0x84 count: 1 index: 2
debug: number of DAPM widgets 1
debug: loading widget codec_out1 id 9
info: Widget type not supported 1
debug: type: 5, size: 0x84 count: 1 index: 3
debug: number of DAPM widgets 1
debug: loading widget codec_in0 id 10
info: Widget type not supported 0
debug: type: 5, size: 0x84 count: 1 index: 4
debug: number of DAPM widgets 1
debug: loading widget codec_in1 id 11
info: Widget type not supported 0
debug: type: 7, size: 0x390 count: 1 index: 0
debug: type: a, size: 0x6f0 count: 1 index: 0
debug: type: 4, size: 0x318 count: 6 index: 1

[...]

So final graph were trying to create pipeline just for last DAPM widgets group that had only 1 widget. Now I made it to aggregate all comps (+ removed leak btw), but I'm not sure that it's correct approach, cos I'm not expert in terms of tplg files.

[singalsu]

This works for me, good work. Thanks @jajanusz !

[singalsu]

I was able to run sof-apl-src-pcm512x.tplg with this version. The actual test cases failed because volume component after SRC was not set up properly by testbench. So the volume loader & control needs further work.

The test topologies fail but they have been not maintained for a while (I tested with my SRC PR, it might also cause something). It would be better to create new ones, e.g. from sof-apl-pcm512x.m4 for volume test case and then other processing components by replacing volume. The test topologies were for DAI parameter variants initially anyway.

[lgirdwood]

I was able to run sof-apl-src-pcm512x.tplg with this version. The actual test cases failed because volume component after SRC was not set up properly by testbench. So the volume loader & control needs further work.

ok, can you add that as an issue and work with @jajanusz on this for the testbench. If it's a quick fix we can include in v1.3.

The test topologies fail but they have been not maintained for a while (I tested with my SRC PR, it might also cause something). It would be better to create new ones, e.g. from sof-apl-pcm512x.m4 for volume test case and then other processing components by replacing volume. The test topologies were for DAI parameter variants initially anyway.

OK, we will need new one(s) for v1.3. Can you add this as a bug with milestone 1.3

[ranj063]

@jajanusz sorry for the delayed response. although harmless, this step is unnecessary because there will only ever be one DAPM_WIDGET header in topology containing all the widgets, if I am not mistaken. So the num_comps should just be equal to the hdr->count.