Monta Vista Linux Supporting Floating Point operations

Monta Vista Linux Supporting Floating Point operations

[Vishwanath]

Hi All,

I am working on 440GP using Monta Vista Linux 2.1. I need floating point
operation support (for eg. function sqrt() for calculating square root of
given float variable) on Linux. I didn't find math emualtion on the Linux.
Can anyone tell me how can I get floating point support ?

Thanks in advance.

Regards,
Vishwanath


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Monta Vista Linux Supporting Floating Point operations

[Eugene Surovegin]

At 12:37 AM 10/23/2002, you wrote:
>I am working on 440GP using Monta Vista Linux 2.1. I need floating point
>operation support (for eg. function sqrt() for calculating square root of
>given float variable) on Linux. I didn't find math emualtion on the Linux.

This is strange.
Floating point emulation IS available in MVL2.1 on 440.
You probably forgot to link with libm.

I just typed trivial test and ran it on our custom 440 board:

main.cpp:
----------------------------------------------
#include <stdio.h>
#include <math.h>

int main(int argc, char* argv[]){

   printf("sqrt(10) = %f\n", sqrt(10.0));

   return 0;
}
----------------------------------------------
Compile and run:

# gcc test.cpp -lm
# ./a.out
sqrt(10) = 3.162278
#

Regards,

  Eugene Surovegin <mailto:ebs@innocent.com>


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Monta Vista Linux Supporting Floating Point operations

[Mark Hatle]

Vishwanath wrote:
> Hi All,
>
> I am working on 440GP using Monta Vista Linux 2.1. I need floating point
> operation support (for eg. function sqrt() for calculating square root of
> given float variable) on Linux. I didn't find math emualtion on the Linux.
> Can anyone tell me how can I get floating point support ?

Floating point support in MontaVista Linux is handled by the compiler and _NOT_
the kernel.  In architecting our system we have weighed the tradeoffs between
soft-floating point, or hard-floating point using kernel exceptions.
Soft-floating point was found to be superior so we use it on all platforms
without a hardware FPU.

One thing to note, because of this the ABI is slightly different then the
binaries you would download off of a random community website.  You need to use
our compiler (or a community compiler configured appropriatly) to build
soft-float.  (On a non-MontaVista compiler, you usually need to pass
-msoft-float on the command line to gcc.)


--Mark


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/