Since no-one else is byte-ing, I will attempt an answer. Noe, although I have been programming under VM/CMS and its cousin VP/CSS since the mid-70s, I am not an efficiency expert.
As I understand things, when a program (or CMS itself) releases memory back to CMS free storage, if RELPAGE is ON, then the CMS memory manager will release the pages back to CP freeing them from its working set and its page table. This additional overhead can become a BIG performance problem if your program/CMS turns around and gets those pages again almost immediately. SET RELPAGE OFF causes CMS memory manager to NOT release the pages back to CP. This can cause the working set to be larger than it needs to be for the length of time that it takes to need those pages again. SET RELPAGE ON can reduce total system overhead when the use of memory in a virtual machine tends to vary greatly (2M in use, then 8M, then back to 2) but tends to remain static for periods of time. I don't have a clue why this would be recommended for a LINUX virtual machine. Does Linux get memory from CMS? It may be that IBM (or someone) has implemented something for similiar reasons under Linux. Lloyd Fuller On Thu, 7 Mar 2002 21:04:57 +0100, Rob van der Heij wrote: >>Loren Charnley posted his sample profile.exec and it included the line >> >>set relpage off /*Avoids a Linux problem * > >Since nobody bites... What fascinated me is that this is a CMS >setting. It is used by CMS when a program ends to tell CP to drop >the pages that were holding the program (otherwise CP would still >keep those pages in storage and on page volumes). >My understanding is that when your PROFILE EXEC issues an >IPL command to start Linux, CP will clear storage anyway. > >Maybe in the dark ages people had a different loader program that >would use CMS to read the kernel image from disk and stuff it into >memory and then branch into it. In that situation you would not want >CMS to clear memory in between. > >Rob >
