An alpha IEEE bug

An alpha IEEE bug

[RHS Linux User]

Hi,

There is a bug in the IEEE math on linux/alpha:

# gcc -mieee ieee.c
# a.out
7f800000
1
# gcc -mieee ieee.c -O
# a.out
zsh: 744 abort      ./a.out
# gdb a.out
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.16 (alpha-redhat-linux), Copyright 1996 Free Software Foundation, Inc...
(gdb) b isinff
Breakpoint 1 at 0x120000780: file ieee.c, line 16.
(gdb) r
Starting program: /home/work/alpha/bugs/math/a.out

Breakpoint 1, isinff (x=Erroneous arithmetic operation.) at ieee.c:16
16        if (ix == 0)
(gdb)

Can someone please run this on other alpha platforms?


Thanks.


-- 
H.J. Lu (hjl@gnu.org)
----
#include <stdio.h>

typedef int int32_t __attribute__ ((__mode__ (  __SI__ ))) ;
typedef union
{
  float value;
  int32_t word;
} ieee_float_shape_type;

int isinff(float x)
{
  int32_t ix,t;
  ieee_float_shape_type gf_u;
  gf_u.value = x;
  ix = gf_u.word;
  if (ix == 0)
    abort ();
  else
    printf ("%x\n", ix);
  t = ix & 0x7fffffff;
  t ^= 0x7f800000;
  t |= -t;
  return ~(t >> 31) & (1 - ((ix & 0x80000000) >> 30));
}

main ()
{
  float x = 1.0 / 0.0;

  printf ("%d\n", isinff (x));
  return 0;
}

Re: An alpha IEEE bug

[Richard Henderson]

On Sat, Feb 07, 1998 at 12:28:16PM -0800, RHS Linux User wrote:
> There is a bug in the IEEE math on linux/alpha:

It is a bug in combine.  Look at -O without -mieee -- it doesn't
abort but it gives wrong results.

I'll see what I can do.


r~

Re: An alpha IEEE bug

[Klaus Kaempf]

> 
> Hi,
> 
> There is a bug in the IEEE math on linux/alpha:
> 
> # gcc -mieee ieee.c
> # a.out
> 7f800000
> 1
> # gcc -mieee ieee.c -O
> # a.out
> zsh: 744 abort      ./a.out

On openVMS/Alpha I get a different output (without optimization):

3f800000
0

It also calls abort() when compiled with optimization.

Klaus
-- 
proGIS Software                 E-Mail: kkaempf@progis.de
Dipl.-Inform. Klaus K"ampf      Fax:    0241-47067-29
Jakobstr. 117                   Voice:  0241-47067-11
D-52064 Aachen                  WWW:	http://www.progis.de

Re: An alpha IEEE bug

[Richard Henderson]

On Mon, Feb 09, 1998 at 11:27:24AM +0100, Klaus Kaempf wrote:
> On openVMS/Alpha I get a different output (without optimization):
> 
> 3f800000
> 0

OpenVMS uses F and G format floats, no?

> It also calls abort() when compiled with optimization.

Probably the same combine problem though.


r~

Re: An alpha IEEE bug

[Klaus Kaempf]

> 
> On Mon, Feb 09, 1998 at 11:27:24AM +0100, Klaus Kaempf wrote:
> > On openVMS/Alpha I get a different output (without optimization):
> > 
> > 3f800000
> > 0
> 
> OpenVMS uses F and G format floats, no?
> 
No ! GNU C on openVMS/Alpha uses IEEE floats, I never tried to get it running
with the old VAX floating formats.

-- 
proGIS Software                 E-Mail: kkaempf@progis.de
Dipl.-Inform. Klaus K"ampf      Fax:    0241-47067-29
Jakobstr. 117                   Voice:  0241-47067-11
D-52064 Aachen                  WWW:	http://www.progis.de