Re: Question on low execl() performance (2.0.x kernels)

cook@vs.lmco.com
Thu, 03 Apr 1997 14:49:20 -0600

on Thu, 3 Apr 1997, Stephen Costaras wrote:

> BYTE UNIX Benchmarks (Version 3.11)
> System -- Linux news 2.0.29 #5 Mon Mar 31 04:11:03 CST 1997 i686 unknown
> Start Benchmark Run: Mon Mar 31 05:16:13 CST 1997
> 2 interactive users.
> [stuff snipped]
> System Call Overhead Test 36162.8 lps (10 secs, 6 samples)
> Pipe Throughput Test 59084.2 lps (10 secs, 6 samples)
> Pipe-based Context Switching Test 24296.4 lps (10 secs, 6 samples)
> Process Creation Test 1789.7 lps (10 secs, 6 samples)
> Execl Throughput Test 235.4 lps (9 secs, 6 samples)
> [stuff snipped]
>

I can somewhat confirm your numbers. On a P5-133, I have gotten the following
results over the last four kernel configurations I have used:

Kernel
1.3.97 2.0.0 2.0.25 2.0.28

<snip>
Process Creation Test 1375.8 1267.4 993.4 982.2
lps
Execl Throughput Test 956.3 961.1 139.9 136.6 lps
<snip>
TEST INDEX INDEX INDEX INDEX

Arithmetic Test (type = double) 4.3 5.8 9.1 9.2
Dhrystone 2 without register variables 6.3 6.1 6.8 7.6
Execl Throughput Test 58.0 58.2 8.5 8.3
File Copy (30 seconds) 18.0 17.1 16.9 17.4
Pipe-based Context Switching Test 26.4 22.7 22.9 22.9
Shell scripts (8 concurrent) 6.7 6.5 10.0 9.7
====================================
SUM of 6 items 119.8 116.5 74.1 75.0
AVERAGE 20.0 19.4 12.4 12.5

Most of the other tests are either effectively the same or better in the
later kernel. I have no idea what changes may have affected the benchmark.
Does anyone else have any ideas?

Rick Cook

-- 
Work:	cook@vs.lmco.com
Home:	cookrick@ix.netcom.com