Hi Marek,
just as I thought, it was _set_seqtype() from seqdb_review that caused
the slowdown. It was reading every single sequence in the database,
instead of just one! Neither Titus nor I caught this very simple
error, but fortunately the megatest did... Here are the current
timings, which seem close to your original times. I have pushed the
fix to the public repository master.
leelab2 60% python protest.py annotation_dm2_megatest.py
...
============================================================
FINAL: 0 error(s) in 3 tests, 0 skipped in 4.428 sec
leelab2 61% python protest.py annotation_hg18_megatest.py
...
============================================================
FINAL: 0 error(s) in 3 tests, 0 skipped in 124.788 sec
leelab2 62% python protest.py nlmsa_hg18_megatest.py
..
============================================================
FINAL: 0 error(s) in 2 tests, 0 skipped in 132.195 sec
-- Chris
On Mar 26, 2009, at 1:14 PM, Marek Szuba wrote:
> On Wed, 25 Mar 2009 20:47:44 -0700
> Christopher Lee <l...@chem.ucla.edu> wrote:
>
>> let's see whether the megatests run successfully from the master
>> branch tonight, which now contains both metabase and seqdb-review,
>> plus my latest megatest fixes.
> Okay, we've got a problem - looks like some of the patches recently
> committed to master have seriously messed up our performance... To
> illustrate the problem, here are some running times compared between
> the latest master HEAD and my local development branch (last
> synchronised with the master on the 20th of March ("Merge: 23f4275...
> de2e35b..."), with the only thing making it different from then-HEAD
> being megatest portability patches):
>
> master ms
> annotation_dm2_megatest 207.705 s 4.736 s
> annotation_hg18_megatest 25801.196 s 92.859 s
> dm2_download_megatest 248.954 s 264.710 s
> nlmsa_dm2_megatest 53.610 s 49.196 s
> nlmsa_hg18_megatest (*) 149.119 s
> pairwise_hg18_megatest 21.974 s 32.166 s
>
> (*) at least about 30 minutes, after which I killed the test
>
> At first I thought bad results from the master were caused by a
> problem with leelab2, unfortunately this is not the case - the
> ms-branch numbers have all been produced within the last hour...
>
> The exact source of the problem is yet to be determined but I think it
> has been introduced within the last 48 hours - I ran nlmsa_dm2,
> nlmsa_hg18 and pairwise_hg18 on (slightly modified) master code
> on Tuesday, and the timings were reasonable. Then again, annotation
> tests didn't run at all back then and among the remaining three only
> nlmsa_hg18 has actually got worse since that run, so it's hard to be
> certain.
>
> --
> MS
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"pygr-dev" group.
To post to this group, send email to pygr-dev@googlegroups.com
To unsubscribe from this group, send email to
pygr-dev+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/pygr-dev?hl=en
-~----------~----~----~----~------~----~------~--~---
