Обсуждение: Memory Usage and libpq
I am investigating a memory usage issue and I noticed something with libpq. It appears that libpq doesn't actually free memory allocated to store results, even after PQclear() is called. In a debug program I can inspect the buffer so I know it exists. I haven't verified whether the buffer is resized with subsequent calls, but when dealing with large result sets it creates a significant memory overhead. Am I on to something here or do I need to keep looking? If so, is there any way that I can force libpq to free or in some way drastically reduce this biffer? Thanks, Keary Suska (719) 473-6431 (719) 440-9952 (cell)
On Tuesday 04 April 2006 03:14 pm, Keary Suska saith: > I am investigating a memory usage issue and I noticed something with libpq. > It appears that libpq doesn't actually free memory allocated to store > results, even after PQclear() is called. In a debug program I can inspect > the buffer so I know it exists. I haven't verified whether the buffer is > resized with subsequent calls, but when dealing with large result sets it > creates a significant memory overhead. > > Am I on to something here or do I need to keep looking? If so, is there any > way that I can force libpq to free or in some way drastically reduce this > biffer? > > Thanks, > > Keary Suska > (719) 473-6431 > (719) 440-9952 (cell) > Keary, I may be wrong about this, but just because you can access the memory after the call to PQclear, doesn't mean that the memory wasn't freed; it just means it hasn't been reused yet. The address still exists and freeing the memory doesn't "clear" it out.
on 4/4/06 11:23 AM, terry@esc1.com purportedly said: > > On Tuesday 04 April 2006 03:14 pm, Keary Suska saith: >> I am investigating a memory usage issue and I noticed something with libpq. >> It appears that libpq doesn't actually free memory allocated to store >> results, even after PQclear() is called. In a debug program I can inspect >> the buffer so I know it exists. I haven't verified whether the buffer is >> resized with subsequent calls, but when dealing with large result sets it >> creates a significant memory overhead. >> >> Am I on to something here or do I need to keep looking? If so, is there any >> way that I can force libpq to free or in some way drastically reduce this >> biffer? > > I may be wrong about this, but just because you can access the memory after > the call to PQclear, doesn't mean that the memory wasn't freed; it just means > it hasn't been reused yet. The address still exists and freeing the memory > doesn't "clear" it out. It almost goes without saying, but it does encourage me to question whether my debug program is showing deallocated memory, or further, if for some reason the system/runtime isn't giving the memory back to the kernel (something I am pursuing elsewhere). I will run a test of my own and see if my debug program is showing deallocated memory. Thanks, Keary Suska Esoteritech, Inc. "Demystifying technology for your home or business"
Keary Suska <hierophant@pcisys.net> writes:
> I am investigating a memory usage issue and I noticed something with libpq.
> It appears that libpq doesn't actually free memory allocated to store
> results, even after PQclear() is called. In a debug program I can inspect
> the buffer so I know it exists. I haven't verified whether the buffer is
> resized with subsequent calls, but when dealing with large result sets it
> creates a significant memory overhead.
This would depend entirely on how malloc/free are implemented on your
platform. Some versions can release space back to the OS, some can't.
In some it depends on what else has been allocated.
regards, tom lane