Re: 2.1.xxx makes Electric Fence 22x slower
David Gadbois (gadbois@cyc.com)
Sat, 5 Sep 1998 22:44:03 -0500
My main application is VM-intensive. The current O(n) search in
do_munmap() looks like it will be a killer: The application maps a
largish (100-300MB) VM image in 64KB chunks, and the allocator and
garbage collector are constantly mapping and unmapping random 64KB
chunks of the address space. With the address space divvied up this
way, there are several hundred VMA entries to search through for every
mmap() or munmap() call versus a handful with some O(log n) search
structure. The locality on the VM modifications is FIFO (the
allocator maps a chunk, and the garbage collector unmaps it much
later), so a cache won't help at all.
I shall report some numbers as soon as I get a 2.1.119 system up and
running.
--David Gadbois
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/faq.html