[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
gprof --line taking _forever_
From: |
David Mathog |
Subject: |
gprof --line taking _forever_ |
Date: |
Wed, 14 Apr 2004 08:55:34 -0700 |
I'm trying to profile the "blastall" program from the NCBI toolbox.
(ftp://ftp.ncbi.nih.gov/toolbox/ncbi_tools/CURRENT/ncbi.tar.gz). The
programs were all compiled with "gcc -g -pg -O3" on a RH 7.3 system
with gcc 3.3.3. One of the programs was run:
time /usr/common/ncbi_new/ncbi/build/blastall -p blastp -d nr_lcl -i
a1hu.pfa -b 10 -v 10 >testout.txt
to produce a gmon.out file. The default gprof command works:
gprof /usr/common/ncbi_new/ncbi/build/blastall
However, when I try this:
grpof --line /usr/common/ncbi_new/ncbi/build/blastall
the CPU pegs at 99% and stays there for at least 45 minutes
(it's still running as I write this.)
This seems like an excessive amount of time for a line profile.
The same method works fine on other (smaller) programs. The
toolbox though, contains a humongous number of files (694 *.c files
containing 1.2M lines of code - not all of which are included in
blastall). It also creates some static libraries which are linked
into the executable.
Is this a bug in gprof or does it just take a hideously long time
on this sort of source code?
Is there some trick that is required to make gprof --line work
more quickly with a program like this?
I already know which routines the program spends most of its time
in (from the first gprof). Is there a way to run gprof so
that it only reports line usage from those routines?
Any other suggestions?
Thanks,
David Mathog
address@hidden
Manager, Sequence Analysis Facility, Biology Division, Caltech
- gprof --line taking _forever_,
David Mathog <=