unresolved symbol litodp,dptoli,dpmul - floating point operations in kernel

unresolved symbol litodp,dptoli,dpmul - floating point operations in kernel

[durai]

[-- Attachment #1: Type: text/plain, Size: 599 bytes --]

hello,
I am using a mips cross compiler (mips-linux-gcc, version 2.95.3) to build my driver
I am using some floating point operations in a wireless lan driver for a mips platform in ucLinux, When i load the driver I am getting unresolved symbols

> 
> insmod: unresolved symbol dptoli
> insmod: unresolved symbol dpmul
> insmod: unresolved symbol litodp

And somebody told me that we cannot use floating point operations in kernel code, but i desperately need to use floating point operations. 
please tell me how to use floating point operations in kernel code.

thanks & regards
durai

[-- Attachment #2: Type: text/html, Size: 1349 bytes --]

Re: unresolved symbol litodp,dptoli,dpmul - floating point operations in kernel

[Geert Uytterhoeven]

On Fri, 10 Oct 2003, durai wrote:
> I am using a mips cross compiler (mips-linux-gcc, version 2.95.3) to build my driver
> I am using some floating point operations in a wireless lan driver for a mips platform in ucLinux, When i load the driver I am getting unresolved symbols
> 
> > 
> > insmod: unresolved symbol dptoli
> > insmod: unresolved symbol dpmul
> > insmod: unresolved symbol litodp
> 
> And somebody told me that we cannot use floating point operations in kernel code, but i desperately need to use floating point operations. 
> please tell me how to use floating point operations in kernel code.

Do not use floating point operations in kernel code.
Re-implement using fixed point or something like that.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: unresolved symbol litodp,dptoli,dpmul - floating point operations in kernel

[Jun Sun]

On Fri, Oct 10, 2003 at 04:28:27PM +0200, Geert Uytterhoeven wrote:
> On Fri, 10 Oct 2003, durai wrote:
> > I am using a mips cross compiler (mips-linux-gcc, version 2.95.3) to build my driver
> > I am using some floating point operations in a wireless lan driver for a mips platform in ucLinux, When i load the driver I am getting unresolved symbols
> > 
> > > 
> > > insmod: unresolved symbol dptoli
> > > insmod: unresolved symbol dpmul
> > > insmod: unresolved symbol litodp
> > 
> > And somebody told me that we cannot use floating point operations in kernel code, but i desperately need to use floating point operations. 
> > please tell me how to use floating point operations in kernel code.
> 
> Do not use floating point operations in kernel code.
> Re-implement using fixed point or something like that.
>

If you are really really desparate, something like the following
might work.

void use_fpu(void)
{
	if (is_fpu_owner()) {
		save_fp(current);
		loose_fpu();
		enable_fpu();
	}
	local_irq_save(flags);
	
	/* now use fpu and store the results */

	local_irq_restore(flags);
}

I like to emphsize this is just a hack and I am not even sure if it will work
at all.  If compiler complains you might have to change the
CC flag for that file or use fpu with inline assembly.

Jun

Re: unresolved symbol litodp,dptoli,dpmul - floating point operations in kernel

[Ralf Baechle]

On Fri, Oct 10, 2003 at 09:58:24AM -0700, Jun Sun wrote:

> > > > insmod: unresolved symbol dptoli
> > > > insmod: unresolved symbol dpmul
> > > > insmod: unresolved symbol litodp

> If you are really really desparate, something like the following
> might work.
> 
> void use_fpu(void)
> {
> 	if (is_fpu_owner()) {
> 		save_fp(current);
> 		loose_fpu();
> 		enable_fpu();
> 	}
> 	local_irq_save(flags);
> 	
> 	/* now use fpu and store the results */
> 
> 	local_irq_restore(flags);
> }
> 
> I like to emphsize this is just a hack and I am not even sure if it will work
> at all.  If compiler complains you might have to change the
> CC flag for that file or use fpu with inline assembly.

The symbols he was missing are used for software floating point.  Software
floating point in kernel space is perfectly ok since it doesn't use the FPU
so your hack isn't even needed.

  Ralf