Messages in this thread | | | Date | Fri, 16 Apr 1999 16:19:19 -0700 | From | David Miller <> | Subject | Re: bad lmbench numbers for mmap |
| |
Date: Sat, 17 Apr 1999 04:38:41 +0530 (IST) From: V Ganesh <ganesh@vxindia.veritas.com>
I just posted the raw numbers and we are roughly twice as slow as solaris.
At least it is consistant with the fact that the solaris run had to touch roughly half as much memory during the benchmark than Linux did.
Unfortunately, the touching of memory in lat_mmap() drains away a lot of what it should be measuring (this I take part blame for as I was one of the people that suggested to Larry to make it do this).
The problem is that we had to add this to avoid cases like Solaris where the mapping state is not created at all if the process does not touch any of the pages. By touching the pages this would better guarentee that both systems in question would have to set up (and later, tear down) the same amount of state in the kernel.
But conversely, when the size of the area gets large, you begin to test the memory speed moreso than the raw mmap/munmap latency of the kernel.
Later, David S. Miller davem@redhat.com
- 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/
| |