[Pan-users] Re: Filtering unusable slow

[walt]

On Fri, 17 Aug 2007 01:57:26 +0000, Greg Lee wrote:

> On Thu, 16 Aug 2007 23:44:19 +0000, walt wrote:
> 
>> Did you both build your own pan using a 64-bit compiler, or are you
>> running 32-bit pan?
> 
> 64 bit compiler compiling 64 bit binaries.
> 
>> Is your swapspace filling up during the stalls?  Disk thrashing? Does
>> the filtering always complete if you wait long enough?
> 
> My swap space hardly gets used with Pan.
> 
> Every time I waited long enough, the filtering completed.  Or rather,
> the times when I killed Pan were times when I didn't wait long enough,
> or else times when I could have waited forever without completion, in
> which case, I guess, I did wait long enough.

I've never done any profiling, unlike Charles who has done a lot of it
while writing pan.  Being ill-equipped for real profiling, I would try
running pan from gdb and halt it in the middle of those stalls with
Ctrl-C to see what code is actually running.  If you do that several
times you should get a feeling for what is taking up all that CPU.

And don't forget strace (aka systrace) to see if some system call is
returning errors, etc.

Also, have you tried renaming your ~/.pan2 directory to let pan make
a fresh start from scratch?  These are the blunt instruments I tend
to use first because they require little thought :o)