Re: PATCH: get_mode_alignment fix

[Lars Brinkhoff <lars dot spam at nocrew dot org>]

Forgot the subject line, sorry about that.

Lars Brinkhoff <lars@nocrew.org> writes:
> get_mode_alignment() in stor-layout.c forces the alignment, in bits,
> to a power of two.  That works fine on 8/16/32-bit machines, but not
> on, say, a 36-bit machine.  This patch makes the alignment be a power
> of two in storage units, which makes no difference at all on machines
> where the size, in bits, of a storage unit is a power of two.
> 
> Bootstrapped on powerpc-unknown-linux-gnu and checked without regressions.

2001-04-25  Lars Brinkhoff  <lars@nocrew.org>

	* stor-layout.c (get_mode_alignment): remove the restriction
	that machine mode sizes must be powers of two.

Index: stor-layout.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/stor-layout.c,v
retrieving revision 1.96
diff -c -3 -p -r1.96 stor-layout.c
*** stor-layout.c	2001/03/23 16:59:56	1.96
--- stor-layout.c	2001/04/25 07:49:49
*************** unsigned int
*** 1844,1855 ****
  get_mode_alignment (mode)
       enum machine_mode mode;
  {
!   unsigned int alignment = GET_MODE_UNIT_SIZE (mode) * BITS_PER_UNIT;
    
    /* Extract the LSB of the size.  */
    alignment = alignment & -alignment;
  
!   alignment = MIN (BIGGEST_ALIGNMENT, MAX (1, alignment));
    return alignment;
  }
  
--- 1844,1855 ----
  get_mode_alignment (mode)
       enum machine_mode mode;
  {
!   unsigned int alignment = GET_MODE_UNIT_SIZE (mode);
    
    /* Extract the LSB of the size.  */
    alignment = alignment & -alignment;
  
!   alignment = MIN (BIGGEST_ALIGNMENT, MAX (1, alignment * BITS_PER_UNIT));
    return alignment;
  }