On 01/02/16 22:51, Antti Kantee wrote:
hmm, and one of these years I
probably should fix the 600+kB of unnecessary bss wastage that mato
found; when it's close to 5% of total memory, it might start hurting.

Ok that's done in NetBSD head, will be in src-netbsd some day when I do the next pullup.

Before:
$ size scheduler.o
   text    data     bss     dec     hex filename
   3999       8  688192  692199   a8fe7 scheduler.o

After:
$ size scheduler.o
   text    data     bss     dec     hex filename
   3855       0   16448   20303    4f4f scheduler.o

(nice how dynamically allocating the cpu_info structures also reduces code size)

Reply via email to