[MOAB-dev] Measuring memory

[Jason Kraftcheck]

On 07/28/2010 03:25 PM, Steve Jackson wrote:
> 
> On Jul 28, 2010, at 14:59 , Jason Kraftcheck wrote:
>>
>> Checking memory usage is basically a non-portable mess.  Just getting the
>> code to compile on all platforms and return useful information on at least
>> those platforms that developers use is about the best that can be
>> expected.
> 
> I wonder if we should just replace the calls to non-portable system
> functions with calls to MOAB's own estimated_memory_use() function.  Is
> anyone relying on having system-level memory use measurements in mbskin
> and mbperf? 

MOAB's calls are a) not quite the same and b) not as accurate.  They are not
quite the same in that they will never measure peak transient memory
required by some algorithm (e.g. the temporary memory used by the skinner.)
 They only measure the memory used for persistent data in the database.
They are not as accurate because they can't know things like the overhead
for chunks allocated via malloc, the fragmentation within the memory managed
by malloc, etc.

- jason