Comments on: Measure memory consumption of creating object or executing functions in C#

By: Tobias Hertkorn

Tobias Hertkorn — Fri, 22 Aug 2008 10:42:58 +0000

Hey Bruno,

unfortunatelly there is little information out there about memory management in C# and no book I am aware of. Some links:

http://blog.vuscode.com/malovicn/archive/2007/12/30/net-foundations-stack-and-heap.aspx

http://www.bluebytesoftware.com/blog/

http://msdn.microsoft.com/en-us/library/f144e03t.aspx

Other than that… look at the mono runtime source code.

Cheers,
Tobi

By: Bruno

Bruno — Thu, 21 Aug 2008 10:37:47 +0000

Hello Tobias,

I think the problem is gone now. Thanks for your answer. Could you please point me a good book in the matter?

Thanks again,
Bruno

By: Tobias Hertkorn

Tobias Hertkorn — Tue, 19 Aug 2008 10:48:14 +0000

Hey Bruon,

could you email me a codesample? That would be easier for me to comment on, since I don’t fully understand the circumstances where you see your varying memory sizes.

Cheers.

By: Bruno

Bruno — Mon, 18 Aug 2008 15:00:54 +0000

Hello,

I’m facing the same problem as Andy. I’m looking for the piece of code that puts my memory in 500MB when I make a loop getting data from my database with NHibernate. I’m using GC.TotalMemory(true) but it never uses the same amount of memory after I open a session.

I hope you can help me with that.

Thank you.

By: Tobias Hertkorn

Tobias Hertkorn — Wed, 12 Dec 2007 14:49:30 +0000

Hi Andy,

please email me the complete code, since my comments seem to not like any less-than signs. I really have to fix that. sorry.

But the good thing is: I think I know what happens – but the rest of your code is quite important for me finding out what exactly is happening. So please, send me the code…
You can find my email address on the “about me” page.

Cheers,
Tobi

By: Andy

Andy — Tue, 11 Dec 2007 15:00:42 +0000

sorry… tags cut off code. Was:

byte[][] cache = new byte[1000][];
long[] memTotals = new long[1000];
for (int i = 0; i

By: Andy

Andy — Tue, 11 Dec 2007 14:59:04 +0000

Hi Tobi and Anton,
Interesting stuff… I’m having to do memory measurement for our performance unit tests. I’m using GC.GetTotalMemory(false) to query the total memory allocated, but it’s got an irritating problem – it does not accurately reflect the actual amount of memory allocated.
(I’ve had the same problem, only worse, with the .NET performance counter “# Bytes in all heaps”).
(If I do a GC.GetTotalMemory(true) I get an accurate picture, but that’s quite invasive, because it forces a GC).
If I store the results of GC.GetTotalMemory(false) in an array:
byte[][] cache = new byte[1000][]; long[] memTotals = new long[1000]; for (int i = 0; i then dump out memTotals, apparent increases only occur once every 8 allocations - i.e. GC.GetTotalMemory(false) only seems to register an increase in ~8192 byte steps... Given that this effect is (hopefully) deterministic, and tiny, I'm still going to plough ahead and use GC.GetTotalMemory in my instrumentation, but it's a pain in the arse to know that it's not 100% reliable... is full profiling the only totally accurate way to do it, or am I talking crazy?

By: Anton

Anton — Mon, 27 Aug 2007 10:51:51 +0000

Hi Tobi, Thanks first of all for this generous post!! I was anticipating a much worse approach I must say, it would have gone a little like this, To partition a section of memory and monitor the use of this section during processes. That would have been me, aproaching the bull from the wrong side not that I have the slightest idea of how to do that and would probably have lead to the creation of more memory “bubbles” (if one might say) if the use of memory exceeded the allocated bit. Implementing your approach will probably mean that I will be maintaining my sanity! Thanks again! Anton