I'd guess people might be curious on the actual parsing/cache impacts
of this series.

To be honest I'm having trouble isolating a parsing time change for it,
which is quite sad. Someone with a slower system or limiting it to a
small set of cores might get some better results. I can see slight
reductions in the generate_dependencies function times. I was also
profiling the "bitbake -S" cache mode, better results might be obtained
without enabling that.

It does reduce the overall "bitbake -S none" cache file size by around
420k (on a 74MB file). Much more significantly, it also reduces the
do_package sigdata file from 224k to 159k uncompressed or 54k -> 36k
compressed (which the files are).

Whilst those improvements don't sound much, this is a significant
reduction in the amount of data bitbake is chucking around pipes
internally and the reduction in size of the sigdata will mount up over
many recipes/tasks in an sstate share.

I do think we can end up with cleaner more understandable/maintainable
code if we do this correctly so whilst I had hoped for more, I still
think this direction is the right one and the changes are worth
pursuing.

Cheers,

Richard




-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#175500): 
https://lists.openembedded.org/g/openembedded-core/message/175500
Mute This Topic: https://lists.openembedded.org/mt/96052504/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to