mysql scaling questions

[Bruce Evans]

On Wed, 2 Jan 2008, Gary Stanley wrote:

> At 06:00 AM 1/2/2008, Kris Kennaway wrote:
>
>> Per later discussion you will also need to either comment out the syscalls 
>> that are (might be) being cached by glibc to artificially inflate its 
>> reported rate, or verify that it is not doing so.
>
> IIRC linux uses vsyscalls on x86_64, and calling things like gettimeofday() 
> isn't as expensive as it normally should be.. vsyscalls are similar to 
> darwins' commpage.

How does it implement gettimeofday()'s 1us precision using direct access?
I see one way: put the time and other volatiles on an unmapped page, and
update the time in the trap handler.  With a lot of work, this should
be only slightly slower than the gettimeofday() syscall.

Bruce