Subject: Re: 25%+ improvement in in_cksum speed!
To: None <port-i386@netbsd.org, tech-perform@l8s.co.uk>
From: Michael Graff <explorer@flame.org>
List: port-i386
Date: 09/17/2002 16:53:06
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> It would be interesting to know what PIII, P4 and athlon XP get.

I increased the iterations by 100x because the test ran too fast :)

This is on a dual Athelon XP 1800 with PC2100 DDR memory:

          in_cksum.s sum f807 took   144405 usecs 0.172144 nsec/byte
           asm adc 1 sum f807 took   595019 usecs 0.709318 nsec/byte
          asm adc 1a sum f807 took   413898 usecs 0.493405 nsec/byte
          asm adc 1b sum f807 took   310851 usecs 0.370563 nsec/byte
          asm adc 1c sum f807 took   415706 usecs 0.495560 nsec/byte
          asm adc 1d sum f807 took   413917 usecs 0.493428 nsec/byte
           asm adc 2 sum f807 took   302992 usecs 0.361195 nsec/byte
           asm adc 4 sum f807 took   207616 usecs 0.247498 nsec/byte
          asm adc 4b sum f807 took   174919 usecs 0.208520 nsec/byte
          asm adc 8b sum f807 took   157807 usecs 0.188121 nsec/byte
          asm pair 2 sum f807 took   215495 usecs 0.256890 nsec/byte
          asm pair 4 sum f807 took   156356 usecs 0.186391 nsec/byte
          asm pair 8 sum f807 took   130256 usecs 0.155277 nsec/byte
         asm pair 16 sum f807 took   113312 usecs 0.135078 nsec/byte
        asm pair 16a sum f807 took    96491 usecs 0.115026 nsec/byte
         asm pair 32 sum f807 took   104955 usecs 0.125116 nsec/byte
          asm quad 8 sum f807 took   139280 usecs 0.166035 nsec/byte


Here's my P4 1.2 GHz laptop:

          in_cksum.s sum f807 took  1062577 usecs 1.266690 nsec/byte
           asm adc 1 sum f807 took  1343647 usecs 1.601752 nsec/byte
          asm adc 1a sum f807 took  1348924 usecs 1.608043 nsec/byte
          asm adc 1b sum f807 took   991374 usecs 1.181810 nsec/byte
          asm adc 1c sum f807 took  1017840 usecs 1.213360 nsec/byte
          asm adc 1d sum f807 took  1055914 usecs 1.258748 nsec/byte
           asm adc 2 sum f807 took  1212721 usecs 1.445676 nsec/byte
           asm adc 4 sum f807 took  1163888 usecs 1.387463 nsec/byte
          asm adc 4b sum f807 took  1121668 usecs 1.337132 nsec/byte
          asm adc 8b sum f807 took  1100700 usecs 1.312137 nsec/byte
          asm pair 2 sum f807 took   547090 usecs 0.652182 nsec/byte
          asm pair 4 sum f807 took   676955 usecs 0.806993 nsec/byte
          asm pair 8 sum f807 took   875602 usecs 1.043799 nsec/byte
         asm pair 16 sum f807 took   979067 usecs 1.167139 nsec/byte
        asm pair 16a sum f807 took   975152 usecs 1.162472 nsec/byte
         asm pair 32 sum f807 took  1026834 usecs 1.224082 nsec/byte
          asm quad 8 sum f807 took   671800 usecs 0.800848 nsec/byte
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (NetBSD)
Comment: See http://www.flame.org/~explorer/pgp for my keys

iD8DBQE9h8Bil6Nz7kJWYWYRAuGkAJ9Ooc5nNzXLlwhRtUemNcbHgcsdBgCdFIrI
lHrIhGuXqYk4zAgDEegEZIw=
=cZIG
-----END PGP SIGNATURE-----