Thread performance (was Re: dynamic context transitions)

[tridge at samba.org]

David,

 > 	That was the case with Solaris 7: it was changed to
 > 	one kernel thread per user thread in 8, vaguely similar
 > 	to the Linux one process-equivalent per thread.

ok

 > ...
 > 	The Solaris model is to assume programs are unthreaded, and 
 > 	make it the programmer's responsibility to chose the thread-safe
 > 	and high-threaded-performance interfaces, and no others. There
 > 	are no locks or warnings about using a thread-unsafe libc function,
 > 	it just fails at runtime.

lets test that theory:

bash-2.03# uname -a
SunOS sun1 5.8 Generic_108528-01 sun4u sparc SUNW,Ultra-60

bash-2.03# ./thread_perf 10 malloc

Running test 'malloc' with 10 tasks
Threads    10.18 +/- 3.87 seconds
Processes   1.51 +/- 0.18 seconds


So, on Solaris 8 malloc is 7x slower with threads than with
processes. Would you like to try a newer version of Solaris and get
back to me with the results?

See http://samba.org/ftp/unpacked/junkcode/thread_perf.c

I know there are methods on Solaris for making malloc go fast with
threads, but even with all those go-faster methods it will at best be
able to approach the speed of processes. For this sort of thing, you
can't beat hardware with software.

Cheers, Tridge