home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!math.fu-berlin.de!news.netmbx.de!Germany.EU.net!mcsun!uknet!comlab.ox.ac.uk!robots.ox.ac.uk!jon
- From: jon@robots.ox.ac.uk (Jon Tombs)
- Newsgroups: comp.os.linux
- Subject: linux-0.97.pl2 speed performance
- Message-ID: <1992Aug30.143951.4417@diomedes.robots.ox.ac.uk>
- Date: 30 Aug 92 14:39:51 GMT
- Sender: jon@robots.ox.ac.uk (Jon Tombs)
- Organization: Robotics Research Group, Engineering Science Dept, Oxford, UK.
- Lines: 55
- Originator: jon@diomedes.robots
-
- There been various claims about the speed of the 0.97-pl1 and 0.97-pl2
- kernel, so here are some of the results from the byte unix benchmark, for
- the two kernels on a 486/33 16M ram 256K cache compiled with gcc2.2.2d.
- These were all run with nothing other than update and the benchmark on the
- machine.
-
- BYTE UNIX Benchmarks (Version 3.11)
-
- linux - patch level 1
- System -- Linux isabel 0.97-19 08/20/92 i386
- 2 interactive users.
- Arithmetic Test (type = arithoh) 60859.0 lps (10 secs, 6 samples)
- System Call Overhead Test 19307.1 lps (10 secs, 6 samples)
- Pipe-based Context Switching Test 4131.4 lps (10 secs, 6 samples)
- Process Creation Test 64.2 lps (10 secs, 6 samples)
- Execl Throughput Test 59.1 lps (9 secs, 6 samples)
- Shell scripts (8 concurrent) 6.0 lpm (60 secs, 3 samples)
- Recursion Test--Tower of Hanoi 516.6 lps (10 secs, 6 samples)
-
- --
- linux - patch level 2
- System -- Linux isabel 0.97.pl2-13 08/29/92 i386
- 2 interactive users.
- Arithmetic Test (type = arithoh) 60478.0 lps (10 secs, 6 samples)
- System Call Overhead Test 18995.5 lps (10 secs, 6 samples)
- Pipe-based Context Switching Test 4051.9 lps (10 secs, 6 samples)
- Process Creation Test 57.9 lps (10 secs, 6 samples)
- Execl Throughput Test 55.0 lps (9 secs, 6 samples)
- Shell scripts (8 concurrent) 5.6 lpm (60 secs, 3 samples)
- Recursion Test--Tower of Hanoi 512.5 lps (10 secs, 6 samples)
- --
-
- They all show a slight slow down. I also compared these results with the
- patch level 1 kernel and gcc2.2.2
-
- System -- Linux isabel 0.97-14 08/07/92 i386
- 2 interactive users.
- Arithmetic Test (type = arithoh) 61581.5 lps (10 secs, 6 samples)
- System Call Overhead Test 19521.2 lps (10 secs, 6 samples)
- Pipe-based Context Switching Test 4323.0 lps (10 secs, 6 samples)
- Process Creation Test 102.1 lps (10 secs, 6 samples)
- Execl Throughput Test 129.9 lps (9 secs, 6 samples)
- Shell scripts (8 concurrent) 6.0 lpm (60 secs, 3 samples)
- Recursion Test--Tower of Hanoi 458.2 lps (10 secs, 6 samples)
- --
-
-
- So it looks like Process Creation and Execl Throughput speed halved with
- the change from gcc2.2.2 to 2.2.2d - These results are probably depend
- very much on the process being invoked, but definetly for small processes
- gcc2.2.2d has added a large overhead for my set up.
-
- --
- Jon <jon@robots.ox.ac.uk>
- Dsylexics of the world, untie!
-