Martin Husemann <martin@duskware.de> typed:
: On Sun, Aug 10, 2003 at 05:15:43PM -0700, Bill Studenmund wrote:
:> I'm not so sure about that. Can/will the compiler notice that you're only
:> using half of a 64-bit value? Otherwise, by making this change, you're
: If gcc does miss the obvious, the 32bit code might be slightly slower
: than the old code. I can't test that right now, anyone with i386 gcc 3.3.1
: around? If the value is spread accross two 32 bit registers, and the code

Works OK.  On i386:

compiled -O2 with                       compiled -O2 with
gcc 2.95.3 (NetBSD nb4)                 gcc 3.3.1 (from NetBSD src yesterday)

00000000 <bswap64>:			00000000 <bswap64>:
   0: 55          push   %ebp		   0: 55          push   %ebp
   1: 89 e5       mov    %esp,%ebp	   1: 89 e5       mov    %esp,%ebp
   3: 53          push   %ebx	      |	   3: 8b 45 08    mov    0x8(%ebp),%e
   4: 8b 45 08    mov    0x8(%ebp),%e |	   6: 53          push   %ebx
   7: 8b 55 0c    mov    0xc(%ebp),%e	   7: 8b 55 0c    mov    0xc(%ebp),%e
   a: 89 c1       mov    %eax,%ecx	   a: 89 c1       mov    %eax,%ecx
   c: 66 c1 c9 08 ror    $0x8,%cx	   c: 66 c1 c9 08 ror    $0x8,%cx
  10: c1 c9 10    ror    $0x10,%ecx	  10: c1 c9 10    ror    $0x10,%ecx
  13: 66 c1 c9 08 ror    $0x8,%cx	  13: 66 c1 c9 08 ror    $0x8,%cx
  17: 89 d0       mov    %edx,%eax    |	  17: 89 d3       mov    %edx,%ebx
  19: 31 d2       xor    %edx,%edx    |	  19: 66 c1 cb 08 ror    $0x8,%bx
  1b: 66 c1 c8 08 ror    $0x8,%ax     |	  1d: c1 cb 10    ror    $0x10,%ebx
  1f: c1 c8 10    ror    $0x10,%eax   |	  20: 66 c1 cb 08 ror    $0x8,%bx
  22: 66 c1 c8 08 ror    $0x8,%ax     |	  24: 31 c0       xor    %eax,%eax
  26: 31 db       xor    %ebx,%ebx    |	  26: 89 ca       mov    %ecx,%edx
  28: 89 cb       mov    %ecx,%ebx    |	  28: 89 d9       mov    %ebx,%ecx
  2a: 31 c9       xor    %ecx,%ecx    |	  2a: 31 db       xor    %ebx,%ebx
  2c: 31 d2       xor    %edx,%edx    |	  2c: 09 da       or     %ebx,%edx
  2e: 09 c1       or     %eax,%ecx    |	  2e: 09 c8       or     %ecx,%eax
  30: 09 d3       or     %edx,%ebx    |	  30: 8b 1c 24    mov    (%esp,1),%eb
  32: 89 c8       mov    %ecx,%eax    |	  33: c9          leave  
  34: 89 da       mov    %ebx,%edx    |	  34: c3          ret    
  36: 5b          pop    %ebx	      <
  37: c9          leave  	      <
  38: c3          ret    	      <

Following is how gcc 3.4 20030806 (experimental) with -O2 makes it:

00000000 <bswap64>:
   0:	55                   	push   %ebp
   1:	89 e5                	mov    %esp,%ebp
   3:	8b 45 08             	mov    0x8(%ebp),%eax
   6:	53                   	push   %ebx
   7:	8b 55 0c             	mov    0xc(%ebp),%edx
   a:	89 c1                	mov    %eax,%ecx
   c:	66 c1 c9 08          	ror    $0x8,%cx
  10:	c1 c9 10             	ror    $0x10,%ecx
  13:	66 c1 c9 08          	ror    $0x8,%cx
  17:	89 d3                	mov    %edx,%ebx
  19:	66 c1 cb 08          	ror    $0x8,%bx
  1d:	c1 cb 10             	ror    $0x10,%ebx
  20:	66 c1 cb 08          	ror    $0x8,%bx
  24:	31 c0                	xor    %eax,%eax
  26:	89 ca                	mov    %ecx,%edx
  28:	89 d9                	mov    %ebx,%ecx
  2a:	31 db                	xor    %ebx,%ebx
  2c:	09 da                	or     %ebx,%edx
  2e:	09 c8                	or     %ecx,%eax
  30:	5b                   	pop    %ebx
  31:	c9                   	leave  
  32:	c3                   	ret    

Regards,
-- 
Geoff Wing : <gcw@pobox.com>
Rxvt Stuff : <gcw@rxvt.org>
Zsh Stuff  : <gcw@zsh.org>