[kaffe] Whoops

[Patrick Tullmann]

> >The "PS2LINUX" #define was much more readable...   :)  Probably
> >deserves at least a comment in common.h.
> 
> OK, what about I add this to common.h?
> #ifdef _R5900
> #define PS2LINUX
> #endif

Sure.

> >That's a fairly gigantic stack size for each thread.  Is there a
> >PS2-specific requirement for that?  Or did the interpreter just blow
> >up too often with a lower stack size?  (BTW, how big is a JMP_BUF on
> >the MIPS?)
> >
> 
> To be very honest, I Assumed the 32 came from register size, the ps2
> general purpose registers are 128 bits wide. However pointer are 32
> bit.

For the jthreads implementation at least, THREADSTACKSIZE determines
the default (and minimum) amount of memory to allocate for each new
thread's stack.  So, its really only a problem to make it big if
you're going to start a lot of threads.  (To small and you'll run into
out-of-stack errors.)

If you're really curious about the impact of this variable, try the
test program from this message:
	http://www.kaffe.org/pipermail/kaffe/2002-April/007946.html

> Whoops, correction I forgot the sigset_t, (I was looking at bits/setjmp.h)
> 
> It's actually 352 bytes long.

That's not too bad.  I asked because one of those (plus a bunch of
other goop) is allocated on the stack for each active interpreted
method frame.

-Pat

----- ----- ---- ---  ---  --   -    -      -         -               -
Pat Tullmann                                       tullmann at cs.utah.edu
	   "I'd kill for a Nobel Peace Prize." -- S. Wright