libvisual-devel — libvisual mailinglist for developers

[Libvisual-devel] randomness work.

[Dennis S. <sy...@yo...>]

I've put some of the randomness request in CVS, not yet finished tho.

Cheers,
Dennis

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[salsaman <sal...@xs...>]

Burkhard Plaum wrote:

>
> This one seems better:
>
> #define a 1664525L
> #define c 1013904223L
>
> return (fastrand_val= a * fastrand_val + c);
>
> a is suggested by Knuth
> c is suggested by H.W. Lewis and is a prime close to 2^32 * (sqrt(5) - 2)


Thanks for that, I've updated the code in LiVES with this.

Salsaman.

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[salsaman <sal...@xs...>]

Dennis Smit wrote:

>On Fri, 2004-10-01 at 23:10 +0300, Vitaly V. Bursov wrote:
>  
>
>>>which returns 1 with a propability of p (0.0 <= p <= 1.0)
>>>      
>>>
>>:) Nice idea.
>>
>>int lv_random_decide(float a)
>>{
>>	float x=(float)random()/RAND_MAX;
>>	return x <= a;
>>}
>>    
>>
>
>You guys probably figured by now that my math skills really suck :)
>
>Thanks for the translation to code heheheh.
>
>I think I'll be implementing this today, and some other randomness
>functions.
>
>Cheers,
>Dennis
>
>  
>
Great. I implemented a really nice feature in LiVES yesterday. Now you 
can load in an audio file, hit record, then trigger an actor. The video 
frames are grabbed and then rendered so you can have an instant video to 
accompany any audio. This feature really needs the randomness fixes, 
otherwise you just get the same visuals for every audio piece.

Hope to have another LiVES dev release out soon.

Cheers,
Salsaman.

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Dennis S. <sy...@yo...>]

On Fri, 2004-10-01 at 23:10 +0300, Vitaly V. Bursov wrote:
> > which returns 1 with a propability of p (0.0 <= p <= 1.0)
> :) Nice idea.
> 
> int lv_random_decide(float a)
> {
> 	float x=(float)random()/RAND_MAX;
> 	return x <= a;
> }

You guys probably figured by now that my math skills really suck :)

Thanks for the translation to code heheheh.

I think I'll be implementing this today, and some other randomness
functions.

Cheers,
Dennis

Re: [Libvisual-devel] Re: log verboseness

[Dennis S. <sy...@yo...>]

On Sat, 2004-10-02 at 00:43 -0300, Duilio Javier Protti wrote:
> I have implemented visual_log_set_verboseness() on
> libvisual core. We can set these:
> 
> The default level is VISUAL_LOG_VERBOSENESS_MEDIUM. (Dennis, do you
> think it shoul be LOW?).
> 
> Similar handlers can be set for warning/critical/error messages.


I'm very happy with the system you implemented, I've got 1 comment
and 1 question. First, there is no enum yet to entirely disable
visual_log output, I think that VERBOSENESS_NONE should be really
none verbose.

One more question, I really like the message handle system, I wonder
if it would be possible to also have a 'all messages' handler that
can be optionally set. This would be nice for an app that just wants to
redirect ALL output. Another thing is that the handler set function 
should accept a void *priv, if the app wants the handler function to 
handle a private. Let's say a GtkWidget in which text has to be printed.

Thanks for implementing it,
Dennis

Re: [Libvisual-devel] Re: log verboseness

[salsaman <sal...@xs...>]

Duilio Javier Protti wrote:

>I have implemented visual_log_set_verboseness() on
>libvisual core. We can set these:
>
>typedef enum {
>
>        VISUAL_LOG_VERBOSENESS_NONE,    /**< Show only VISUAL_LOG_INFO
>				and VISUAL_LOG_ERROR messages. */
>
>        VISUAL_LOG_VERBOSENESS_LOW,     /**< Show only VISUAL_LOG_INFO,
>					VISUAL_LOG_ERROR and
>					VISUAL_LOG_CRITICAL messages. */
>
>        VISUAL_LOG_VERBOSENESS_MEDIUM,  /**< Show all log messages
>					except VISUAL_LOG_DEBUG ones. */
>
>        VISUAL_LOG_VERBOSENESS_HIGH     /**< Show all log messages. */
>
>} VisLogVerboseness;
>  
>
Ermm...shouldnt' VERBOSENESS_NONE produce NO output ? Or can there be a 
mode LOG_VERBOSENESS_REALLY_REALLY_NONE :-)
Anyway haven't you got it the wrong way round, at lower verboseness 
levels, only the most critical messages should be shown.



Salsaman.

Re: [Libvisual-devel] Printing to stdout, no no no

[Dennis S. <sy...@yo...>]

On Fri, 2004-10-01 at 20:51 -0300, Duilio Javier Protti wrote:
> I think the best way to do that is to add an API function which
> let the user to set a handler for outputting messages, which 
> overrides the default behavior, like glib does. In glib you 
> have i.e g_message() with a default behavior, but you can override
> it calling g_set_message_handler() with a function that receives the
> message string and returns nothing.
> 
> This approach is very powerfull because you can process and redirect
> output whatever you like. Specially on GUI applications, you can 
> redirect messages to some GUI widget.

That sounds really good, doing this together with verboseness level API
I assume ?

Cheers,
Dennis

[Libvisual-devel] Re: log verboseness

[Duilio J. P. <dp...@fc...>]

I have implemented visual_log_set_verboseness() on
libvisual core. We can set these:

typedef enum {

        VISUAL_LOG_VERBOSENESS_NONE,    /**< Show only VISUAL_LOG_INFO
				and VISUAL_LOG_ERROR messages. */

        VISUAL_LOG_VERBOSENESS_LOW,     /**< Show only VISUAL_LOG_INFO,
					VISUAL_LOG_ERROR and
					VISUAL_LOG_CRITICAL messages. */

        VISUAL_LOG_VERBOSENESS_MEDIUM,  /**< Show all log messages
					except VISUAL_LOG_DEBUG ones. */

        VISUAL_LOG_VERBOSENESS_HIGH     /**< Show all log messages. */

} VisLogVerboseness;

The default level is VISUAL_LOG_VERBOSENESS_MEDIUM. (Dennis, do you
think it shoul be LOW?).

To avoid showing info or error messages, or to redirected them or to
process them on some special way, use:

visual_log_set_info_handler (visual_log_message_handler_func_t handler);

where handler must have signature:

void my_info_handler (const char *message, const char *funcname)
{
   /* funcname is the name of the function which invokes visual_log()
    and can be NULL on non-GNU systems */
   ...
}

Similar handlers can be set for warning/critical/error messages.


Bye,
Duilio.

Re: [Libvisual-devel] Printing to stdout, no no no

[Duilio J. P. <dp...@fc...>]

> On Fri, 2004-10-01 at 19:44 +0100, salsaman wrote:
> > This output needs to be removed or sent instead to stderr.
> > 
> > Regards,
> > Salsaman.
> 
> All has been removed from CVS, and libvisual-INFO output will be
> controlable through API.
> 
> Dennis

I think the best way to do that is to add an API function which
let the user to set a handler for outputting messages, which 
overrides the default behavior, like glib does. In glib you 
have i.e g_message() with a default behavior, but you can override
it calling g_set_message_handler() with a function that receives the
message string and returns nothing.

This approach is very powerfull because you can process and redirect
output whatever you like. Specially on GUI applications, you can 
redirect messages to some GUI widget.


Bye,
Duilio.

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Vitaly V. B. <vit...@us...>]

On Fri, 01 Oct 2004 12:25:37 +0200
Burkhard Plaum <pl...@ip...> wrote:

> Dennis Smit wrote:
> 
> > (which also includes a 
> > good pseudo random number gen, and while we add it, some nice range,
> > float random features. 
> 
> An EXTREMELY useful function would be something like
> 
> int lv_random_decide(float p);
> 
> which returns 1 with a propability of p (0.0 <= p <= 1.0)
:) Nice idea.

int lv_random_decide(float a)
{
	float x=(float)random()/RAND_MAX;
	return x <= a;
}


-- 
Vitaly
GPG Key ID: F95A23B9

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Dennis S. <sy...@yo...>]

On Fri, 2004-10-01 at 12:25 +0200, Burkhard Plaum wrote:
> Dennis Smit wrote:
> 
> > (which also includes a 
> > good pseudo random number gen, and while we add it, some nice range,
> > float random features. 
> 
> An EXTREMELY useful function would be something like
> 
> int lv_random_decide(float p);
> 
> which returns 1 with a propability of p (0.0 <= p <= 1.0)

I'm not sure if I understand this correctly, but basicly checks if p > 
0.5 or lower than ? or ?!

Dennis.

Re: [Libvisual-devel] Printing to stdout, no no no

[Dennis S. <sy...@yo...>]

On Fri, 2004-10-01 at 19:44 +0100, salsaman wrote:
> This output needs to be removed or sent instead to stderr.
> 
> Regards,
> Salsaman.

All has been removed from CVS, and libvisual-INFO output will be
controlable through API.

Dennis

Re: [Libvisual-devel] Printing to stdout, no no no

[salsaman <sal...@xs...>]

Dennis Smit wrote:

>
>>so in future if you want to see any output at all from libvisual in 
>>LiVES, it will *have to* go to stderr.
>>    
>>
>
>We're about to add an API to control message verboseness.
>
>Thanks,
>Dennis
>
>
>  
>
Perfect !

Salsaman.

Re: [Libvisual-devel] Printing to stdout, no no no

[Dennis S. <sy...@yo...>]

On Fri, 2004-10-01 at 19:48 +0100, salsaman wrote:
> salsaman wrote:
> ...
> 
> Further to my last message, I have put in libvis:
> 
> close (1);
> 
> so in future if you want to see any output at all from libvisual in 
> LiVES, it will *have to* go to stderr.

We're about to add an API to control message verboseness.

Thanks,
Dennis

Re: [Libvisual-devel] Printing to stdout, no no no

[salsaman <sal...@xs...>]

salsaman wrote:

> salsaman wrote:
> ...
>
> Further to my last message, I have put in libvis:
>
> close (1);
>
> so in future if you want to see any output at all from libvisual in 
> LiVES, it will *have to* go to stderr.
>
>
> Salsaman.


OK, can't do that...it stops output from LiVES. I'm afraid I can't 
release the libvisual plugin until all output to stdout is removed.

Salsaman.

Re: [Libvisual-devel] Printing to stdout, no no no

[salsaman <sal...@xs...>]

salsaman wrote:
...

Further to my last message, I have put in libvis:

close (1);

so in future if you want to see any output at all from libvisual in 
LiVES, it will *have to* go to stderr.


Salsaman.

[Libvisual-devel] Printing to stdout, no no no

[salsaman <sal...@xs...>]

Hi all,
in my testing of libvisual, which is almost complete now, I did:

lives-exe 2>/dev/null

this should produce no output because stdout is used in LiVES for output 
streaming. Instead I saw this, e.g. for gforce:

libvisual INFO: no progname: New Delta field: Acid_Gap
libvisual INFO: no progname: New wave shape: 3D_Cylinder
libvisual INFO: no progname: Looking up expression in dict: MAG
libvisual INFO: no progname: New Particle: Warp_Me_Away
New Delta field: Accident
libvisual INFO: no progname: New wave shape: Zoomed_Square
libvisual INFO: no progname: Looking up expression in dict: MAG
libvisual INFO: no progname: New Delta field: The_Spectrologist_s_Assistant
New Delta field: Yes
New Delta field: 6_Feet_Under

I am guessing the "New Delta Field" is temporary debug info.


Other output which needs removing:

goom2:

NEW RESIZE: 480 360 0x839dcf8


Jackdaw:

Computing table number 1
Computing table number 2
Computing table number 3
Computing table number 4


Oinksie (!)
[OINKSIE] we've got a requisition hit YEAH BABY YEAH
[OINKSIE] event handler is being called
[OINKSIE] we've got a dimension hit YEAH BABY YEAH



This output needs to be removed or sent instead to stderr.

Regards,
Salsaman.

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Burkhard P. <pl...@ip...>]

> Yes but it could be quantised (e.g. it could only be measured in 
> milliseconds then multiplied by 1000).

Oops, ok, forgot this.
BTW for my work, I also used time(NULL). It will generate
the same sequence only if it's called multiple times
within one second.


-- 
_____________________________
Dr.-Ing. Burkhard Plaum
Institut fuer Plasmaforschung
Pfaffenwaldring 31
70569 Stuttgart
Tel.: +49 711 685-2187
Fax.:            -3102

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[salsaman <sal...@xs...>]

Burkhard Plaum wrote:

>
> But assuming a libvisual program is started at random
> times within a second, even with tv_usec the propability for the
> same sequence in 2 subsequent program runs is 10^(-6).
> I guess that should be enough.
>
Yes but it could be quantised (e.g. it could only be measured in 
milliseconds then multiplied by 1000).

Salsaman.

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Burkhard P. <pl...@ip...>]

>> You can use these two functions adapted from LiVES:
>>
>> static uint32_t fastrand_val;
>>
>> inline uint32_t fastrand(void)
>> {
>>  return (fastrand_val=fastrand_val*1073741789+32749);
>> }

This one seems better:

#define a 1664525L
#define c 1013904223L

return (fastrand_val= a * fastrand_val + c);

a is suggested by Knuth
c is suggested by H.W. Lewis and is a prime close to 2^32 * (sqrt(5) - 2)

> #include <sys/time.h>
> struct timeval tv;
> 
> gettimeofday(&tv,NULL);
> fastsrand(tv.tv_usec);

Ok, maybe /dev/urandom is better, because with tv_usec, you'll never get seed
values greater than 999999.

But assuming a libvisual program is started at random
times within a second, even with tv_usec the propability for the
same sequence in 2 subsequent program runs is 10^(-6).
I guess that should be enough.

-- 
_____________________________
Dr.-Ing. Burkhard Plaum
Institut fuer Plasmaforschung
Pfaffenwaldring 31
70569 Stuttgart
Tel.: +49 711 685-2187
Fax.:            -3102

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[salsaman <sal...@xs...>]

Dennis Smit wrote:

>On Fri, 2004-10-01 at 11:48 +0200, Burkhard Plaum wrote:
>  
>
>>>#include <sys/time.h>
>>>struct timeval tv;
>>>
>>>gettimeofday(&tv,NULL);
>>>fastsrand(tv.tv_usec);
>>>
>>>      
>>>
>>Yea, that's the best method :-)
>>
>>BTW all plugins must be changed to no longer call
>>[s]rand().
>>    
>>
>
>Yep, I saw xscreensaver has something that gives a warning or an error
>on srand, not sure how they do that...
>
>Cheers,
>Dennis
>
>  
>
Well I am not exactly sure about that now. I have realised that tv_usec 
could be cyclic on some machines, so it might be better to seed with 
/dev/urandom if that device exists.

The algorithm itself seems to produce enough random noise to be useful, 
you can check this for yourself with the "noise" real time effect in LiVES.
FYI, it was originally in effecTV but used non prime numbers. My 
contribution was to switch to primes, which eliminated an apparent 
periodicity in the noise effect.


Salsaman.

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Dennis S. <sy...@yo...>]

On Fri, 2004-10-01 at 11:48 +0200, Burkhard Plaum wrote:
> > #include <sys/time.h>
> > struct timeval tv;
> > 
> > gettimeofday(&tv,NULL);
> > fastsrand(tv.tv_usec);
> > 
> Yea, that's the best method :-)
> 
> BTW all plugins must be changed to no longer call
> [s]rand().

Yep, I saw xscreensaver has something that gives a warning or an error
on srand, not sure how they do that...

Cheers,
Dennis

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Burkhard P. <pl...@ip...>]

Dennis Smit wrote:

> (which also includes a 
> good pseudo random number gen, and while we add it, some nice range,
> float random features. 

An EXTREMELY useful function would be something like

int lv_random_decide(float p);

which returns 1 with a propability of p (0.0 <= p <= 1.0)

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Burkhard P. <pl...@ip...>]

>>
>> You can use these two functions adapted from LiVES:
>>
>> static uint32_t fastrand_val;
>>
>> inline uint32_t fastrand(void)
>> {
>>  return (fastrand_val=fastrand_val*1073741789+32749);
>> }
>>
>> void fastsrand(uint32_t seed)
>> {
>>        fastrand_val = seed;
>> }
>>

This could be ok for most applications, but making a real good
random generator is not trivial.

Some detailed introduction to various random number generators
can be found in the Book "Numerical Recipes in C".
Here is the online version (scroll down to Chapter 7):

http://www.library.cornell.edu/nr/cbookcpdf.html

> And to seed it:
> 
> 
> 
> #include <sys/time.h>
> struct timeval tv;
> 
> gettimeofday(&tv,NULL);
> fastsrand(tv.tv_usec);
> 
Yea, that's the best method :-)

BTW all plugins must be changed to no longer call
[s]rand().

-- 
_____________________________
Dr.-Ing. Burkhard Plaum
Institut fuer Plasmaforschung
Pfaffenwaldring 31
70569 Stuttgart
Tel.: +49 711 685-2187
Fax.:            -3102

Re: [Libvisual-devel] Re: Release planning, quality release upcoming week.

[Dennis S. <sy...@yo...>]

On Thu, 2004-09-30 at 08:38 +0300, Vitaly V. Bursov wrote:
> Hm, I don't think it's hard. :) just few lines of code.
> 
> Fast way to do this is to define these globals:
> void visual_set_seed(uint32_t seed);
> uint32_t visual_get_seed();
> uint32_t visual_random();
> 
> With no multi-threading everything should be absolutely reproducible ;)
> 
> Next, this plugin-specific api:
> 
> void visual_plugin_set_seed(VisPluginInfo*, uint32_t seed);
> uint32_t visual_plugin_get_seed(VisPluginInfo*);
> void visual_plugin_random(VisPluginInfo*);
> 
> state of the random number generation algo will be strored in VisPluginInfo.
> 
> One thing: we need to borrow nice pseudo-random numbger gen algo.
> 
> OK?

Alright, it seems that there is real interest for this feature, in that
case. I'm not sure about the name: visual_plugin_random, because it 
kinda implies that it 'could give a random plugin' or something, we need
a good namespace here, and an lv_random module (which also includes a 
good pseudo random number gen, and while we add it, some nice range,
float random features. If you could make a patch for this, that would
be awesome. I currently really don't feel like doing development
as I can't keep my mind at it.

Thanks,
Dennis