> I tried build.sh on my SS1+ with around 20040418 kernel,
> but system CPU time during compilation seems much higher:
> 
> I think old sparc kernels (~1.6W) had the similar problem
> and then it was fixed:
> http://mail-index.netbsd.org/port-sparc/2003/07/31/0000.html
> http://mail-index.netbsd.org/port-sparc/2003/08/03/0000.html
> but now it returns again?

Most likely this is caused by an increase in fragmentation of kernel
virtual address space causing `pmeg thrashing' again. I guess the
buffer cache changes are responsible for this.

I'll see if I can work around it.

-pk