Amos Jeffries wrote:
We are still limited to one page,
--- 1 page or 1 segment/item? Looking at the output of ipcs they show a max seg size of 32768 (32k), but the units are in kbytes, not bytes. so the real limit looks more like 32MB Are you sure that limit was 32K and not 32k kbytes? (i.e. 32M?) I found a "ipcs -l" command that shows limits and it describes things in terms of segments, with the number of segments being fairly limited compared to amount of memory: util-linux> ipcs -l ------ Shared Memory Limits -------- max number of segments = 4096 max seg size (kbytes) = 32768 max total shared memory (kbytes) = 8388608 min seg size (bytes) = 1 ------------- Do you think it might be better resource usage to subdivide one large segment rather than trying to use many small segments so as not to use up segment descriptors? Maybe a pseudo extend-based fs like xfs has (or maybe it's feasible to strip out the extent-allocator/manager from xfs and use it to manage a memory system? I don't know if it would have any benefit over a standard alloc/malloc model, but it might minimize fragmentation over time.
