cross from alpha to arm

cross from alpha to arm

[Stephen Williams]

This very simple patch allows egcs to successfully build libgcc for the
arm from a Linux/alpha host. I can't say if it fixes the root of the problem,
but it certainly gets me over the immediate hurdle. This seems to be the
only place left where 64bit host --> 32bit arm is confused.

*** egcs-971008/gcc/config/arm/arm.c	Mon Aug 11 08:57:24 1997
--- ../egcs-971008/gcc/config/arm/arm.c	Tue Oct 14 18:31:16 1997
***************
*** 407,412 ****
--- 407,416 ----
  {
    unsigned HOST_WIDE_INT mask = ~0xFF;
  
+ #if HOST_BITS_PER_WIDE_INT > 32
+   if (i & 0xffffffff00000000LL) return FALSE;
+ #endif
+ 
    /* Fast return for 0 and powers of 2 */
    if ((i & (i - 1)) == 0)
      return TRUE;

Re: cross from alpha to arm

[Richard Henderson]

On Wed, Oct 15, 1997 at 09:15:25AM +0800, Stephen Williams wrote:
> + #if HOST_BITS_PER_WIDE_INT > 32
> +   if (i & 0xffffffff00000000LL) return FALSE;
> + #endif

Make this instead

  if (i & ~(unsigned HOST_WIDE_INT) 0xffffffff) return FALSE;

The LL will mess up non-gcc 64-bit hosts, and the conditional
will be optimized away on hosts it doesn't matter for.


r~

Re: cross from alpha to arm

[Stephen Williams]

On Wed, Oct 15, 1997 at 09:15:25AM +0800, Stephen Williams wrote:
> + #if HOST_BITS_PER_WIDE_INT > 32
> +   if (i & 0xffffffff00000000LL) return FALSE;
> + #endif

rth@dot.cygnus.com said:
> Make this instead

>   if (i & ~(unsigned HOST_WIDE_INT) 0xffffffff) return FALSE; 

Even better! It tests out just fine. Thanks.


-- 
Steve Williams
steve@icarus.com
steve@picturel.com

"The woods are lovely, dark and deep.  But I have promises to keep,
And lines to code before I sleep, And lines to code before I sleep."

Re: cross from alpha to arm

[Richard Earnshaw]

> On Wed, Oct 15, 1997 at 09:15:25AM +0800, Stephen Williams wrote:
> > + #if HOST_BITS_PER_WIDE_INT > 32
> > +   if (i & 0xffffffff00000000LL) return FALSE;
> > + #endif
> 
> Make this instead
> 
>   if (i & ~(unsigned HOST_WIDE_INT) 0xffffffff) return FALSE;
> 
> The LL will mess up non-gcc 64-bit hosts, and the conditional
> will be optimized away on hosts it doesn't matter for.
> 
> 
> r~

While this is better, I think you would get better code if you failed 
numbers that weren't the equivalent of a sign-extended 32 bit number.  
Indeed some of the code in output_move_double currently assumes this 
(though this probably needs fixing).

Something like:

{
  unsigned HOST_WIDE_INT mask = ~0xFF;
  HOST_WIDE_INT j = i >> 31;

  if (j != 0 && j != (HOST_WIDE_INT) -1)
    return FALSE;

  /* Fast return for 0 and powers of 2 */
  if ((i & (i - 1)) == 0)