[ia64-tools] Re: A glibc ia64 patch for math

[H . J . Lu]

On Thu, Sep 21, 2000 at 05:01:07PM +0200, Andreas Jaeger wrote:
> 
> Here's a patch - I'll commit it later.  Please test it and tell me if
> this fixes the issues.  I hope I didn't introduce any copy&paste
> errors :-(.

Thanks.


> @@ -2451,6 +2485,11 @@
>  static void
>  fmod_test (void)
>  {
> +  errno = 0;
> +  FUNC(fmod) (0, 0);
> +  if (errno == ENOSYS)
> +    /* Function not implemented.  */
> +    return;
>  

You cannot test fmod () like that :-).

I still got many failures on long double. I am enclosing a few here:

testing long double (without inline functions)
Failure: Test: atan2 (-0, 1) == -0
Result:
 is:          0.00000000000000000000e+00   0x0.00000000000000000000p+0
 should be:  -0.00000000000000000000e+00  -0x0.00000000000000000000p+0
 difference:  0.00000000000000000000e+00   0x0.00000000000000000000p+0
 ulp       :  0.0000
 max.ulp   :  0.0000
Failure: Test: atan2 (-inf, 1) == -pi/2
Result:
 is:          0.00000000000000000000e+00   0x0.00000000000000000000p+0
 should be:  -1.57079632679489661926e+00 -0xc.90fdaa22168c23500000000000000000000p-3
 difference:  1.57079632679489661926e+00 0xc.90fdaa22168c23500000000000000000000p-3
 ulp       :  14488038916154245685.0000
 max.ulp   :  0.0000

I put math/test-ldouble.out and math/test-ildoubl.out at

http://ftp.valinux.com/ftp/pub/support/hjl/ia64/test-ldouble.out
http://ftp.valinux.com/ftp/pub/support/hjl/ia64/test-ildouble.out

I am afraid I won't be able to look into it until the week after next
week. I hope someone can take look at those long double functions. They
are ok on ia32. But ia32 uses 80bit long double with 12bytes and ia64
uses 80bit long double with 16bytes.

Thanks.


H.J.