Throughput Degradation with 4GB Kernel

[Partha Narayanan]

Here are some VolanoMark results that show throughput
differences for 1GB and 4GB kernel build options.

This test runs in loopback mode, with no disk activity,
so the 4GB degradation should not be related to bounce
buffers.

A time based profile did not provide any insight into
why the 4GB kernel on a 4-way degrades.  We will get a
a kernprof acg for both the 1GB and 4GB kernels.

This degradation was unexpected.
Any ideas on what may be causing this degradation ?

VolanoMark 2.1.2 10/100 loopback test,
8-way 700MHZ Pentium III,
IBM JVM 1.3. (build cx 130 -20010626)
Throughput in msg/sec

Kernel            UP          4-way       8-way
=======           =====       =====       ======
2.4.14 1GB        11024       17565       13734
2.4.14 4GB        11060       13378       13030
-----------------------------------------------

2.4.16 1GB        11064       16292       11840
2.4.16 4GB        11105       12605        9434
-----------------------------------------------

2.4.17 1GB        11005       15894       11595
2.4.17 4GB        11057       12754        9363
-----------------------------------------------

Partha Narayanan
[e-mail]