I was curious about this:
http://www.mysqlperformanceblog.com/2006/09/29/what-to-tune-in-mysql-server-after-installation/
http://www.mysql.com/news-and-events/newsletter/2003-11/a0000000269.html

so I did a couple ad hoc tests.

Celeron 1.7Ghz 512MB ram
MySQL 5.0

(Note: I got errors when I tried to change innodb_log_file_size with MySQL version 4.1 on Debian)

# You can set innodb_buffer_pool_size up to 50 - 80 %
# of RAM but beware of setting memory usage too high
# Set innodb_log_file_size to 25 % of innodb_buffer_pool_size
# innodb_buffer_pool_size = 128M

Before each test:

DROP DATABASE sa_bayes;
CREATE DATABASE sa_bayes;
USE sa_bayes;
SOURCE /root/bayes_mysql.sql;
ALTER TABLE bayes_expire TYPE=InnoDB;
ALTER TABLE bayes_global_vars TYPE=InnoDB;
ALTER TABLE bayes_seen TYPE=InnoDB;
ALTER TABLE bayes_token TYPE=InnoDB;
ALTER TABLE bayes_vars TYPE=InnoDB;
quit

Create initial user:

su amavis -c 'sa-learn --spam sample-spam.txt'

Restoring a 9MB Bayes database from a backup:
time su amavis -c 'sa-learn --restore backup.txt'

The Cpu % jumps around quite a bit, so these are only representational.
The queries per second avg tends to improve as time goes on.

##############################################################################

innodb_buffer_pool_size         | 8388608    # default 8M
innodb_log_file_size            | 5242880    # default 5M 

top - 19:29:15 up  1:36,  2 users,  load average: 1.17, 1.07, 0.75
Tasks:  64 total,   1 running,  63 sleeping,   0 stopped,   0 zombie
Cpu(s):  8.3%us,  3.8%sy,  0.0%ni,  0.0%id, 87.1%wa,  0.8%hi,  0.0%si,  0.0%st
Mem:    516780k total,   470532k used,    46248k free,    37288k buffers
Swap:   120448k total,        0k used,   120448k free,   213904k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
11861 amavis    16   0 32220  25m 3084 S  1.5  5.1   0:43.97 sa-learn
 3471 mysql     15   0  124m  19m 5508 S  9.0  3.9   1:14.24 mysqld

Queries per second avg: 76.882

This took:
real    196m40.059s # 3.26 hours
user    5m24.612s
sys     0m52.603s


##############################################################################

If I had more available ram, this may have done better...

In the [mysqld] section of my.cnf:
innodb_buffer_pool_size = 128M
innodb_log_file_size = 32M

innodb_buffer_pool_size         | 134217728
innodb_log_file_size            | 33554432

top - 20:28:49 up  2:35,  2 users,  load average: 2.00, 1.92, 1.55
Tasks:  64 total,   1 running,  63 sleeping,   0 stopped,   0 zombie
Cpu(s): 59.5%us, 15.9%sy,  0.0%ni,  0.0%id, 24.3%wa,  0.3%hi,  0.0%si,  0.0%st
Mem:    516780k total,   490356k used,    26424k free,    32284k buffers
Swap:   120448k total,       44k used,   120404k free,   234028k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
12960 amavis    15   0 32220  25m 3084 S 30.9  5.1   3:57.83 sa-learn
12887 mysql     15   0  215m  23m 4564 S 44.5  4.7   5:20.57 mysqld

Queries per second avg: 630.082

This took:
real    25m17.483s
user    5m25.160s
sys     0m58.264s

##############################################################################

If I had more available ram, this may have done better...

innodb_buffer_pool_size         | 268435456  #256MB
innodb_log_file_size            | 67108864   #64MB

top - 05:17:08 up 11:24,  2 users,  load average: 2.11, 1.96, 1.40
Tasks:  64 total,   1 running,  63 sleeping,   0 stopped,   0 zombie
Cpu(s): 79.4%us, 19.9%sy,  0.0%ni,  0.0%id,  0.0%wa,  0.3%hi,  0.3%si,  0.0%st
Mem:    516780k total,   510488k used,     6292k free,    29828k buffers
Swap:   120448k total,       52k used,   120396k free,   271004k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
20585 amavis    15   0 32220  25m 3084 S 42.9  5.1   4:24.14 sa-learn
20485 mysql     15   0  348m  29m 4588 S 56.8  5.8   5:54.25 mysqld

Queries per second avg: 431.094

This took:
real    25m33.937s
user    5m28.089s
sys     0m56.528s

##############################################################################

Used:
status;
show variables;

Net result: 8X faster in this test (I'm guessing mostly inserts).
It might have been even faster if I had more available free ram.