Hi All,
Just came across this website. Feature 9 is memory management and they claim that their product "automatically releas[es] memory [that is] no longer needed."
Is this a marketing ploy, or do you think they have some trick up their sleeves? Are they just making a claim based on what they .Net runtime provides in any case (or are they gasp calling GC.Collect())?
Not really interested in the product, DotFuscator suffices for me - I am just really interested on how this could be achieved.
Interesting.
A review from 2007 contains this:
{smartassembly} uses several different techniques to bring down the memory usage.
We asked the developers of {smartassembly} for some of the specifics, and they told us that by default the CLR reserves a ton of memory for .NET assemblies - whether or not they request it. So {smartassembly} intelligently detects when the CPU is idle (or thereabout) and increases or decreases the amount of reserved memory for your assembly according to its requirements - "automated" GC in a sense, except that memory may or may not have ever been in use.
In that same vein, {smartassembly} (with the benefit of literally having access to your source code thanks to the way .NET is designed) marks any and all classes that don't have any detectable "child" classes inheriting from them as "sealed" thereby reducing the amount of memory and CPU used by the CLR during run-time to determine what functions should be made available to other classes and libraries.
The same review has a "before/after" pair of screenshots showing an application going from 8M to 420K. This suggests to me that it's actually just reducing the working set of the application, not the real memory requirements. The same "optimisation" appears to occur if you minimise an app. It doesn't suddenly take up less memory. I don't believe the desktop .NET framework can really run with only 420K.
The automatic sealing feature is interesting - I can't see that it helps other than for lookup of virtual methods. I doubt that the impact is really significant, but of course I haven't benchmarked it.
So, nothing conclusive, but I doubt that it's doing anything that I'd particularly want.