Heap Profiling Question
simonmarhaskell at gmail.com
Mon Jul 24 11:53:40 EDT 2006
Rich Fought wrote:
>> IIRC the timestamps ignore time spent in GC and time spent sampling
>> the heap, so they measure runtime of the program only.
> So if I have a server that is idle most of the time waiting for
> requests, the timestamps recorded in the heap profile will not be "real
> time" but program execution time?
That's right - the general idea is to make the profile insensitive to other
loading on the machine. I can see there might be an argument for making this
>> Still, it does look a little strange. Maybe you could send us the
>> output of +RTS -S?
> Alloc Collect Live GC GC TOT TOT Page Flts
> bytes bytes bytes user elap user elap
> 264424 262144 20592 0.00 0.00 0.03 0.01 0 0
> (Gen: 1)
> 311540 270336 63732 0.00 0.00 3.84 5.25 0 0
> (Gen: 0)
So between these first two GCs, your program spent nearly 4 seconds doing
something, but only allocated 300k. What was it doing?
More information about the Glasgow-haskell-users