[ODE] stack pointers

[Shaul Kedem]

Hi,
 Changed the subject ... 
 
 I suspect, after russ validated the stack pointers I
gave, that there is a bug in the test_boxstack (or
somewhere inside ODE) - I can't finish the demo with
num = 100 , and stack at 8 mb. it fails, I thought it
fails because of the stack but if the size of the
stack in the post was smaller than 1 MB I can't see
how the stack overflew.
Reproducing the bug on my machine is very easy: simply
compile with NUM 100 and press 'B' key without
stopping, sooner or later the whole thing crash.

shul.

--- Erwin de Vries <erwin at vo.com> wrote:
> I made a huge pile of spheres using the previous
> build.  I could go VERY
> high. At least 50 spheres. But at a certain point
> the framerate went down
> very quickly, and about 2 or 3 spheres later it
> crashed. Didnt get a
> stackpointer, but i thought i'd mention.
> 
> There is a much quicker way to get a stack overflow
> i believe. Just throw
> many, many boxes at each other. Dont make a perfect
> stack. Just make sure
> every box touches 3 or 4 other boxes. Kaboom!
> 
> Erwin
> 
> ----- Original Message -----
> From: "Russ Smith" <russ at q12.org>
> To: "Shaul Kedem" <shaul_kedem at yahoo.com>
> Cc: <ODE at q12.org>
> Sent: Friday, November 09, 2001 02:38
> Subject: Re: [ODE] (no subject)
> 
> 
> > > Well, I ran it and this is what I got:
> > > grabStackPtr, simLoop :0x68bff40
> > > and
> > > grabStackPtr, dInternalStepIsland_x2 :0x68be1ec
> > >
> > > shul
> >
> > that's not a lot of stack being used. i ran the
> test myself - with
> > double precision, the boxstack demo rarely uses
> more that 10k of stack,
> > even with up to 50 boxes. but i wasn't stacking
> them carefully and
> > they were falling all over the place.
> >
> > a better test is test_chain1. with 50 spheres in
> the chain, most of
> > which are sitting on the ground, the total
> constraint dimension
> > count is M=228, and i got 506620 bytes of stack
> being used (500k).
> > so:
> >   stack space ~= 9.7457 M^2
> >
> > implying about one double precision number per M^2
> slots plus a few
> > M-sized vectors. so ODE's stack usage is
> reasonable.
> >
> > russ.
> >
> > --
> > Russell Smith
> > http://www.q12.org
> > _______________________________________________
> > ODE mailing list
> > ODE at q12.org
> > http://q12.org/mailman/listinfo/ode
> >
> >
> 
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode


__________________________________________________
Do You Yahoo!?
Find a job, post your resume.
http://careers.yahoo.com