jit.3m bug..

Oct 24, 2007 at 3:05pm

jit.3m bug..

Hi,

this is a weird one: jit.3m gives funky results with 1-dimensional
matrices as input.

In example patch below the 100 x 100 noise matrix gives 0.0, 0.5, 1.0
as expected
But the 10000 x 1 noise matrix gives 0.0, 2.0, 1.0 ….

on max 4.6.2 jitter 1.6.2 macintel 2x2x2.66Ghz macos 10.4.10

max v2;
#N vpatcher 30 89 562 337;
#P window setfont “Sans Serif” 9.;
#P flonum 360 157 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
#P flonum 308 157 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
#P flonum 262 158 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
#P window linecount 1;
#P newex 291 81 125 196617 jit.noise 1 float32 10000;
#P newex 290 112 53 196617 jit.3m;
#P flonum 168 152 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
#P flonum 116 152 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
#P flonum 70 153 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
#P button 239 30 15 0;
#P newex 99 76 134 196617 jit.noise 1 float32 100 100;
#P newex 98 107 53 196617 jit.3m;
#P connect 0 0 3 0;
#P connect 1 0 0 0;
#P fasten 2 0 1 0 244 60 104 60;
#P connect 0 1 4 0;
#P connect 0 2 5 0;
#P connect 6 0 8 0;
#P connect 7 0 6 0;
#P fasten 2 0 7 0 244 63 296 63;
#P connect 6 1 9 0;
#P connect 6 2 10 0;
#P pop;

ciao,

Joost.

——————————————-

Joost Rekveld
———– http://www.lumen.nu/rekveld

——————————————-

“There is little use in devising a system of thought about
the nature of the trap if the only thing to do in order to
get out of the trap is to know the trap and find the exit.
Everything else is utterly useless: Singing hymns about
the suffering in the trap … or making poems about the
freedom outside of the trap, dreamed of within the trap …
The first thing to do is to find the exit out of the trap.
The nature of the trap has no interest whatsoever beyond
this one crucial point:
WHERE IS THE EXIT OUT OF THE TRAP?

(Wilhelm Reich)

——————————————-

#34305
Oct 24, 2007 at 3:36pm

yep. Seems the 10000 case is wrong but the 10000×1 case is correct.
thanks for the report.
wes

On 10/24/07, Joost Rekveld

wrote:
> Hi,
>
> this is a weird one: jit.3m gives funky results with 1-dimensional
> matrices as input.
>
> In example patch below the 100 x 100 noise matrix gives 0.0, 0.5, 1.0
> as expected
> But the 10000 x 1 noise matrix gives 0.0, 2.0, 1.0 ….
>
> on max 4.6.2 jitter 1.6.2 macintel 2x2x2.66Ghz macos 10.4.10
>
>
> max v2;
> #N vpatcher 30 89 562 337;
> #P window setfont “Sans Serif” 9.;
> #P flonum 360 157 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P flonum 308 157 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P flonum 262 158 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P window linecount 1;
> #P newex 291 81 125 196617 jit.noise 1 float32 10000;
> #P newex 290 112 53 196617 jit.3m;
> #P flonum 168 152 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P flonum 116 152 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P flonum 70 153 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> #P button 239 30 15 0;
> #P newex 99 76 134 196617 jit.noise 1 float32 100 100;
> #P newex 98 107 53 196617 jit.3m;
> #P connect 0 0 3 0;
> #P connect 1 0 0 0;
> #P fasten 2 0 1 0 244 60 104 60;
> #P connect 0 1 4 0;
> #P connect 0 2 5 0;
> #P connect 6 0 8 0;
> #P connect 7 0 6 0;
> #P fasten 2 0 7 0 244 63 296 63;
> #P connect 6 1 9 0;
> #P connect 6 2 10 0;
> #P pop;
>
>
> ciao,
>
>
> Joost.
>
>
>
>
>
> ——————————————-
>
> Joost Rekveld
> ———– http://www.lumen.nu/rekveld
>
> ——————————————-
>
> “There is little use in devising a system of thought about
> the nature of the trap if the only thing to do in order to
> get out of the trap is to know the trap and find the exit.
> Everything else is utterly useless: Singing hymns about
> the suffering in the trap … or making poems about the
> freedom outside of the trap, dreamed of within the trap …
> The first thing to do is to find the exit out of the trap.
> The nature of the trap has no interest whatsoever beyond
> this one crucial point:
> WHERE IS THE EXIT OUT OF THE TRAP?
>
> (Wilhelm Reich)
>
> ——————————————-
>
>
>
>
>
>
>

#115487
Oct 24, 2007 at 3:46pm

Also, if you turn parallel processing off, it will fix your problem.
There is an issue with parallel jitter and large 1D matrices.

#P window setfont “Sans Serif” 9.;
#P window linecount 1;
#P message 41 65 85 196617 ; jitter parallel 0;
#P window clipboard copytext “;jitter parallel 0″ #E;
#P window clipboard copycount 1;

wes

On 10/24/07, Wesley Smith wrote:
> yep. Seems the 10000 case is wrong but the 10000×1 case is correct.
> thanks for the report.
> wes
>
> On 10/24/07, Joost Rekveld

wrote:
> > Hi,
> >
> > this is a weird one: jit.3m gives funky results with 1-dimensional
> > matrices as input.
> >
> > In example patch below the 100 x 100 noise matrix gives 0.0, 0.5, 1.0
> > as expected
> > But the 10000 x 1 noise matrix gives 0.0, 2.0, 1.0 ….
> >
> > on max 4.6.2 jitter 1.6.2 macintel 2x2x2.66Ghz macos 10.4.10
> >
> >
> > max v2;
> > #N vpatcher 30 89 562 337;
> > #P window setfont “Sans Serif” 9.;
> > #P flonum 360 157 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> > #P flonum 308 157 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> > #P flonum 262 158 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> > #P window linecount 1;
> > #P newex 291 81 125 196617 jit.noise 1 float32 10000;
> > #P newex 290 112 53 196617 jit.3m;
> > #P flonum 168 152 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> > #P flonum 116 152 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> > #P flonum 70 153 35 9 0 0 0 3 0 0 0 221 221 221 222 222 222 0 0 0;
> > #P button 239 30 15 0;
> > #P newex 99 76 134 196617 jit.noise 1 float32 100 100;
> > #P newex 98 107 53 196617 jit.3m;
> > #P connect 0 0 3 0;
> > #P connect 1 0 0 0;
> > #P fasten 2 0 1 0 244 60 104 60;
> > #P connect 0 1 4 0;
> > #P connect 0 2 5 0;
> > #P connect 6 0 8 0;
> > #P connect 7 0 6 0;
> > #P fasten 2 0 7 0 244 63 296 63;
> > #P connect 6 1 9 0;
> > #P connect 6 2 10 0;
> > #P pop;
> >
> >
> > ciao,
> >
> >
> > Joost.
> >
> >
> >
> >
> >
> > ——————————————-
> >
> > Joost Rekveld
> > ———– http://www.lumen.nu/rekveld
> >
> > ——————————————-
> >
> > “There is little use in devising a system of thought about
> > the nature of the trap if the only thing to do in order to
> > get out of the trap is to know the trap and find the exit.
> > Everything else is utterly useless: Singing hymns about
> > the suffering in the trap … or making poems about the
> > freedom outside of the trap, dreamed of within the trap …
> > The first thing to do is to find the exit out of the trap.
> > The nature of the trap has no interest whatsoever beyond
> > this one crucial point:
> > WHERE IS THE EXIT OUT OF THE TRAP?
> >
> > (Wilhelm Reich)
> >
> > ——————————————-
> >
> >
> >
> >
> >
> >
> >
>

#115488
Oct 25, 2007 at 9:04am

Ha Wes,

thanks for tracking it down, but I wouldn’t like to sacrifice my
jitter parallellity for this..
I mainly use jit.3m for debugging, and I suppose that your
xray.jit.mean would be a good alternative in the meantime…

thanks,

Joost.

On Oct 24, 2007, at 5:46 , Wesley Smith wrote:

> Also, if you turn parallel processing off, it will fix your problem.
> There is an issue with parallel jitter and large 1D matrices.
>
> #P window setfont “Sans Serif” 9.;
> #P window linecount 1;
> #P message 41 65 85 196617 ; jitter parallel 0;
> #P window clipboard copytext “;jitter parallel 0″ #E;
> #P window clipboard copycount 1;
>
>
> wes
>

——————————————-

Joost Rekveld
———– http://www.lumen.nu/rekveld

——————————————-

“There is little use in devising a system of thought about
the nature of the trap if the only thing to do in order to
get out of the trap is to know the trap and find the exit.
Everything else is utterly useless: Singing hymns about
the suffering in the trap … or making poems about the
freedom outside of the trap, dreamed of within the trap …
The first thing to do is to find the exit out of the trap.
The nature of the trap has no interest whatsoever beyond
this one crucial point:
WHERE IS THE EXIT OUT OF THE TRAP?

(Wilhelm Reich)

——————————————-

#115489

You must be logged in to reply to this topic.