[Scons-users] Not able to make full use of CPUs for processing speed

[Bill Deegan]

Markus,

Any chance you'd contribute such via a PR?

-Bill
SCons Project Co-Manager

On Mon, Sep 27, 2021 at 8:57 AM Markus B. Schenkel <mbschenkel at gmail.com>
wrote:

> Hi Don,
>
> A colleague of mine (Edo) has implemented the debug option
> --action_timestamps with 3.1.2.
> It prints time-stamps for each build step and we post-processed this (with
> some proprietary scripts) to analyze parallelism issues.
> Maybe that helps...
>
> Regards,
> Markus
>
>
> Mats Wichmann <mats at wichmann.us> schrieb am Do., 16. Sept. 2021, 21:51:
>
>>
>> On 9/16/21 10:08, Bill Deegan wrote:
>> > you can also use the --taskmastertrace=FILE scons option and look at
>> the
>> > output file, that may give you hints why you're not getting the
>> > parallelism you suspect, though that's a deeper dive.
>>
>>
>> I guess a question to answer is if there's enough parallelism available
>> in your setup that a larger jobs pool helps.  If there rarely are
>> situations when more than five builds are ready to run at once,
>> increasing the jobs pool way beyond five is unlikely to help much.  Not
>> sure there's a really good way to get SCons to tell you that... a lot of
>> the information is in the taskmaster trace, sure, but my experience is
>> using it is kind of like trying to drink from a firehose most times.
>>
>> There should probably be some kind of sampling profile setup to help
>> figure out if things aren't being kept busy just because there's not
>> unblocked work available, or if SCons is blowing too much time
>> calculating things... it's my impression (don't shoot me, Bill!) that
>> the performance characterizations that were made of SCons and its
>> taskmaster/jobs setup were done on computing technologies several
>> generations removed from today's and could stand a revisit.
>> _______________________________________________
>> Scons-users mailing list
>> Scons-users at scons.org
>> https://pairlist4.pair.net/mailman/listinfo/scons-users
>>
> _______________________________________________
> Scons-users mailing list
> Scons-users at scons.org
> https://pairlist4.pair.net/mailman/listinfo/scons-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://pairlist4.pair.net/pipermail/scons-users/attachments/20210927/e143144f/attachment-0001.htm>