An effective way to compress Large Object Heap

Question

I think I've found an effective way to compress LOH: 
If CLR always alloc every large object at the beginning of a RAM page(usually 4KB per page)，then the large object heap(LOH) can be compressed without much
cost: CLR can compress LOH by modifying RAM page table and TLB instead of copying data. If so, small fragmentation maybe still exist (less then a memory page size per fragment), but there would be no large fragmentation, and compressing would be very fast
because of no copying. To do this, OS support may be needed, fortunately Windows OS and Visual Studio are both Microsoft's softwares, so Microsoft can implement this at least on Windows.

Answer 1

I suspect that this technique is already used in a 3rd party implementation of Java. It was briefly mentioned in a discussion about alternative GCs for CoreCLR:
https://github.com/dotnet/coreclr/issues/430
In any case, feel free to open a specific issue in the CoreCLR GitHub repository about this. Implementing an entirely new GC is a lot of work but simply improving the existing one might be another story. Though the fact that it requires kernel
support doesn't help, it may be done but it will have to wait until the kernel guys implement support for it and that will take time. There may also be legal problems if someone has a patent for such a technique.
Regarding performance: modifying page tables has a certain cost. It's possible to find that for smaller blocks it's faster to just copy the block than to modify the page tables. An implementation would probably have to employ a mix of memory copying and
page table modification and in that case you don't even need to require that objects are allocated on a page boundary. Keep in mind that it's not required to deal with one object at a time, there may be consecutive live objects that can be copied in one go.
As for Joel ramblings, well:
"I have a master in computer science and know the speed of the algorithm of compacting a large heap."
Given that you have a history of misleading/inaccurate/confusing claims such a statement is priceless. I suppose that during your master courses you were never taught basic data structures and as result you don't know that hashtable searches are O(1)
and that sorting does help with searching:
https://social.msdn.microsoft.com/Forums/en-US/09a3116a-9626-401f-8c45-c0e8b09da69c/fastest-search-algorithm?forum=csharpgeneral
In this particular case you seem to confuse the term heap as used in GC with the heap data structure. Great.
Thank you! I have opened a specific issue in the CoreCLR GitHub repository about this: https://github.com/dotnet/coreclr/issues/555