#401 Progress message from tide-search inaccurate with threading

[William S Noble]

The user-level progress report message that tide-search reports is inaccurate:

INFO: Running search
INFO: 1000 spectrum-charge combinations searched, 1% complete
INFO: 2000 spectrum-charge combinations searched, 2% complete
INFO: 3000 spectrum-charge combinations searched, 4% complete
INFO: 4000 spectrum-charge combinations searched, 5% complete
INFO: 5000 spectrum-charge combinations searched, 6% complete
INFO: Time per spectrum-charge combination: 0.001589 s.
INFO: Average number of candidates per spectrum-charge combination: 0.158543
INFO: Elapsed time: 129 s

Note that the percentage only goes up to 6% before the search is done. The problem here is two-fold. First, the numerator in the percentage calculation is not considering spectra that were skipped for various reasons (not enough peaks, zero candidates, etc.). This problem can be fixed with the attached patch.

However, that leaves the second problem, which is that when more than one thread is running, then the numerator only includes the searches for the current thread. I don't know if there is an easy way around this.

Alice, can you comment on this?

[Alice Cheng]

Would the patch attached work? Instead of introducing a separate variable, we could just move the reporting to the top of the function so that we get the skipped sc_combos, and start sc_index from -1 instead of 0. Since sc_index was already under a lock, it should be updated across all threads.

[William S Noble]

Yes, I tried this out, and it works fine. The changes look OK to me.
Assuming you run it through the smoke tests, please go ahead and check in
this change. You can also resolve this issue.

Thanks!
Bill

On Wed, Jun 22, 2016 at 4:03 PM, Alice Cheng acheng94@users.sf.net wrote:

Would the patch attached work? Instead of introducing a separate variable,
we could just move the reporting to the top of the function so that we get
the skipped sc_combos, and start sc_index from -1 instead of 0. Since
sc_index was already under a lock, it should be updated across all threads.

Attachments:

• carppatch.diff
  https://sourceforge.net/p/cruxtoolkit/issues/_discuss/thread/db5e33be/aafc/attachment/carppatch.diff
  (3.4 kB; application/octet-stream)

---

• [issues:#401] https://sourceforge.net/p/cruxtoolkit/issues/401/
  Progress message from tide-search inaccurate with threading*

Status: open
Milestone: Crux v2.0
Created: Wed Jun 22, 2016 10:27 PM UTC by William S Noble
Last Updated: Wed Jun 22, 2016 10:27 PM UTC
Owner: Alice Cheng
Attachments:

• progress.diff
  https://sourceforge.net/p/cruxtoolkit/issues/401/attachment/progress.diff
  (1.1 kB; text/x-patch)

The user-level progress report message that tide-search reports is
inaccurate:

INFO: Running search
INFO: 1000 spectrum-charge combinations searched, 1% complete
INFO: 2000 spectrum-charge combinations searched, 2% complete
INFO: 3000 spectrum-charge combinations searched, 4% complete
INFO: 4000 spectrum-charge combinations searched, 5% complete
INFO: 5000 spectrum-charge combinations searched, 6% complete
INFO: Time per spectrum-charge combination: 0.001589 s.
INFO: Average number of candidates per spectrum-charge combination:
0.158543
INFO: Elapsed time: 129 s

Note that the percentage only goes up to 6% before the search is done. The
problem here is two-fold. First, the numerator in the percentage
calculation is not considering spectra that were skipped for various
reasons (not enough peaks, zero candidates, etc.). This problem can be
fixed with the attached patch.

However, that leaves the second problem, which is that when more than one
thread is running, then the numerator only includes the searches for the
current thread. I don't know if there is an easy way around this.

Alice, can you comment on this?

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/cruxtoolkit/issues/401/

To unsubscribe from further messages, please visit
https://sourceforge.net/auth/subscriptions/