CompuServe Thread

I contrived a test that measures process swap performance on all 3 versions of MS Windows. And, here are the results, [for a i486DX-33, 16Mb mem, 256k cache]: process swaps per second Windows 3.1: 1800 Windows/NT: 1200 Windows32s: 1200 Windows32s (dbg): 2.8 [due to DebugOutputString handling] However, performance degrades substantially when I try to do memory copies between two processes and a shared DLL. Under Win32s the performance degrades as follows, process swaps per seconds CopyMemory(DLLdest, LocalSrc, 2048): 256 CopyMemory(DLLdest, LocalSrc, 1024): 230 memcpy(DLLdest, LocalSrc, 2048): 254 byte-by-byte FOR loop of 2048 bytes: 190 At first I thought that the overhead was Win32s API overhead but the FOR loop performance is worse than the API functions. Also the size of the message doesn't make very much difference in swap rates. So, I now think the performance hit is in sparse-virtual-memory validation which must be activated on each process swap. Am I on the right track? If so, is there a way to avoid doing this on every process swap? Would CreateFileMapping help get around this? Also, is there a similar performance hit if a DLL function accesses the memory of an application rather than the other way around? There is 1 Reply.