samba - May 2003 - [RESEND] crc32 optimization

Hi Samba

I noticed that the crc32 function in lib/crc32.c is somewhat
inefficient. This little patch will reduce the inner loop from 10
instructions to 8 instructions on x86 with gcc. gcc can't figure
out this simple optimization by itself on x86.

Further optimization is possible by using the impl. in the linux 2.5 kernel, but
I don't know if it is worth it. Is crc32 common in samba? What would the
typical
buffer len be?

    Jocke

Patch against 2_2_8 CVS:
--- lib/crc32.c	Sat Oct 20 23:23:35 2001
+++ lib/crc32.c.new	Tue May 13 14:45:31 2003
@@ -60,7 +60,7 @@
         uint32 crc=0xffffffff;
 	int i;
         for(i=0;i<count;i++)
-                crc = (crc>>8) ^ CRCTable[(buffer[i] ^ crc) & 0xff];
+                crc = CRCTable[(buffer[i] ^ crc) & 0xff] ^ (crc>>8);
         crc^=0xffffffff;
 	DEBUG(10,("crc32_calc_buffer: %x\n", crc));
 	dump_data(100, buffer, count);

On Tue, 2003-05-27 at 20:42, Joakim Tjernlund wrote:
> Hi Samba
> 
> I noticed that the crc32 function in lib/crc32.c is somewhat
> inefficient. This little patch will reduce the inner loop from 10
> instructions to 8 instructions on x86 with gcc. gcc can't figure
> out this simple optimization by itself on x86.
So, we are saving two instructions...  I think this e-mail consumed more
;-)
> Further optimization is possible by using the impl. in the linux 2.5
kernel, but
> I don't know if it is worth it. Is crc32 common in samba? What would
the typical
> buffer len be?
It's reasonably rare.  It's certainly not in any fast-paths that I know
of.
>     Jocke
> 
> Patch against 2_2_8 CVS:
> --- lib/crc32.c	Sat Oct 20 23:23:35 2001
> +++ lib/crc32.c.new	Tue May 13 14:45:31 2003
> @@ -60,7 +60,7 @@
>          uint32 crc=0xffffffff;
>  	int i;
>          for(i=0;i<count;i++)
> -                crc = (crc>>8) ^ CRCTable[(buffer[i] ^ crc) &
0xff];
> +                crc = CRCTable[(buffer[i] ^ crc) & 0xff] ^
(crc>>8);
>          crc^=0xffffffff;
>  	DEBUG(10,("crc32_calc_buffer: %x\n", crc));
>  	dump_data(100, buffer, count);
Looks reasonable, but I'll let somebody else chase it up.  

Andrew Bartlett

-- 
Andrew Bartlett                                 abartlet@pcug.org.au
Manager, Authentication Subsystems, Samba Team  abartlet@samba.org
Student Network Administrator, Hawker College   abartlet@hawkerc.net
http://samba.org     http://build.samba.org     http://hawkerc.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url :
http://lists.samba.org/archive/samba/attachments/20030527/376560ac/attachment.bin