[Bluez-devel] Elimination of pow() in SBC code

[Bluez-devel] Elimination of pow() in SBC code

[Marcel Holtmann]

Hi Henryk,

the pow() function is the only reason for linking with -lm and it is
only used in one place with a limit range of parameters.

	scalefactor[ch][sb] = pow(2.0, (frame->scale_factor[ch][sb] + 1));

So what is the range of scale_factor[ch][sb]. From a quick testing a saw
that it ranges from 0 - 15. Is this always true? If yes, then I would
use a pre-calculated array for 2^x.

Regards

Marcel




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Elimination of pow() in SBC code

[Henryk Plötz]

[-- Attachment #1: Type: text/plain, Size: 885 bytes --]

Moin,

Am Sun, 28 Nov 2004 00:51:46 +0100 schrieb Marcel Holtmann:

> the pow() function is the only reason for linking with -lm and it is
> only used in one place with a limit range of parameters.

Oh yes, that occurence of pow(2.0, x) is in older code I forgot about.
(And before -lm was necessary for cos(), too)

> So what is the range of scale_factor[ch][sb]. From a quick testing a
> saw that it ranges from 0 - 15. Is this always true? 

Yes, scale_factor is a 4 bit integer (Table 12.13 of A2DP) and there
should be no way of getting it bigger. If there was this would be a bug.

> If yes, then I would use a pre-calculated array for 2^x.

Isn't 1<<x good enough for that?

-- 
Henryk Plötz
Grüße aus Berlin
~~~~~~~ Un-CDs, nein danke! http://www.heise.de/ct/cd-register/ ~~~~~~~
~ Help Microsoft fight software piracy: Give Linux to a friend today! ~

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [Bluez-devel] Elimination of pow() in SBC code

[Brad Midgley]

guys

that's a good idea to break the need for libm and for another 
int->float->int conversion

> 	scalefactor[ch][sb] = pow(2.0, (frame->scale_factor[ch][sb] + 1));
> 
> So what is the range of scale_factor[ch][sb]. From a quick testing a saw
> that it ranges from 0 - 15. Is this always true? If yes, then I would
> use a pre-calculated array for 2^x.

or like

  scalefactor[ch][sb] = 4 << frame->scale_factor[ch][sb];

the compiler doesn't know about the limited range of scale_factor so it 
may actually generate a loop here.

brad


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Elimination of pow() in SBC code

[Marcel Holtmann]

Hi Henryk,

> > the pow() function is the only reason for linking with -lm and it is
> > only used in one place with a limit range of parameters.
> 
> Oh yes, that occurence of pow(2.0, x) is in older code I forgot about.
> (And before -lm was necessary for cos(), too)
> 
> > So what is the range of scale_factor[ch][sb]. From a quick testing a
> > saw that it ranges from 0 - 15. Is this always true? 
> 
> Yes, scale_factor is a 4 bit integer (Table 12.13 of A2DP) and there
> should be no way of getting it bigger. If there was this would be a bug.
> 
> > If yes, then I would use a pre-calculated array for 2^x.
> 
> Isn't 1<<x good enough for that?

stupid me. Of course power of two is done by bitshifting. I committed
the following to the CVS:

	2 << frame->scale_factor[ch][sb]

Regards

Marcel




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Elimination of pow() in SBC code

[Marcel Holtmann]

Hi Brad,

> that's a good idea to break the need for libm and for another 
> int->float->int conversion

we don't need -lm anymore, but there is still the math.h include and
this is because of the fabs() function. Any ideas?

> > 	scalefactor[ch][sb] = pow(2.0, (frame->scale_factor[ch][sb] + 1));
> > 
> > So what is the range of scale_factor[ch][sb]. From a quick testing a saw
> > that it ranges from 0 - 15. Is this always true? If yes, then I would
> > use a pre-calculated array for 2^x.
> 
> or like
> 
>   scalefactor[ch][sb] = 4 << frame->scale_factor[ch][sb];
> 
> the compiler doesn't know about the limited range of scale_factor so it 
> may actually generate a loop here.

It must be

	scalefactor[ch][sb] = 2 << frame->scale_factor[ch][sb];

However bitshifting is the right way to solve this and I don't see any
problems for the compiler. We shift the integer value 2.

Regards

Marcel




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Re: [Bluez-devel] Elimination of pow() in SBC code

[Brad Midgley]

Marcel

> we don't need -lm anymore, but there is still the math.h include and
> this is because of the fabs() function. Any ideas?

the abs functions are usually done with a macro

#define fabs(x) ((x)<0?(-x):(x))

but we could define our own similar macro if you want to stop using math.h

brad


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel