Re: kernel performance degradation & first benchmarks

Matjaz Breskvar (phoenix@fiz.uni-lj.si)
Mon, 04 May 1998 09:27:19 +0200


Hello!

Somehow the attachment didn't get in last mail, so here it is now.
I hope it is usefull!

Sorry for the mistake.
regards
phoenix

Benchmark iX-SSBA 1.21E

Date: 03.05.1998

System: i586
CPU: PENTIUM 166MMX
RAM: 32 Mb
HD: 2.1 Gb IDE
OS: Linux 2.1.99 (compiled using pgcc, with no
changes to
default optimizations)
Compiler: pgcc-2.90.27 980315 (egcs-1.0.2 release)
Options (for ssba): -O6 -mpentium

--1--> MP-Dhrystone [MP-Dhry/s]:
mp_dhry2nr [without reg,without optimization,1000000 iter]
mp_dhry2r [with reg,without optimization,1000000 iter]
mp_dhry2nro [without reg,with optimization,1000000 iter]
mp_dhry2ro [without reg,with optimization,1000000 iter]

2.0.33 2.1.99
----------------------- -----------------------
mp_dhry2nr 154083 156985 (101 %) 161030 157356 (97 %)
mp_dhry2r 170357 171379 (100 %) 169779 174216 (102 %)
mp_dhry2nro 259740 260416 (100 %) 258397 260078 (100 %)
mp_dhry2ro 259740 260416 (100 %) 260416 262123 (100 %)

--2--> Whetstone [KWhet/s]:
whets(single precision,with optimization,50M inst)
whetd(double precision,with optimization,50M inst)

2.0.33 [KWhetstones/sec] 2.1.99 [KWhetstones/sec]
------------------------ ------------------------
whets 87719 87719
whetd 81967 83333

--3--> Linpack [KFlops]
linpackrs (single precision,with optimization,rolled)
linpackus (single precision,with optimization,unroll)
linpackrd (double precision,with optimization,rolled)
linpackud (double precision,with optimization,unroll)

2.0.33 2.1.99
------------ ------------
linpackrs 21458 Kflops 21458 Kflops
linpackus 22889 Kflops 24524 Kflops
linpackrd 18559 Kflops 19074 Kflops
linpackud 20196 Kflops 19619 Kflops

--5--> Utah
2.0.33 time for utah in seconds: real: 0.46 user: 0.23 syst: 0.05
2.1.99 time for utah in seconds: real: 0.44 user: 0.24 syst: 0.04

--6--> Tools
2.0.30 total time for tools : real: 4.54 user: 1.34 syst: 1.6
2.1.99 total time for tools : real: 3.97 user: 1.38 syst: 1.31

--7--> Multi Users

2.0.33 2.1.99
-------------------- --------------------
8 0.78 0.33 0.44 0.72 0.42 0.30
16 1.53 0.68 0.85 1.42 0.64 0.78
32 3.45 1.38 1.72 2.90 1.56 1.27

--8--> Machine
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec
%CPU
2.0.33 70 2630 49.7 2611 14.3 1259 24.7 2734 62.3 2892 34.3 35.9
3.8
2.1.99 70 2480 29.1 2654 10.7 1071 8.1 2858 27.7 1973 6.3 61.9
1.8

--9--> Test C, make -f testc.mk testc.o
2.0.33 16.14 10.73 0.71
2.1.99 15.24 10.11 0.58

--10-> Doducd
2.0.33 precision=50.00274969 itera=5484 (correct) time=22.81
R=2128
2.1.99 precision=50.00274969 itera=5484 (correct) time=22.62
R=2146

--11-> BSD
2.0.33

---------------------------------------
total time for memory in sec: real: 5.23 user: 5.02 syst:
0.17
total time for calls in sec: real: 1.23 user: 0.11 syst:
0.56
total time for pipes in sec: real: 1.06 user: 0.04 syst:
0.54
total time for fork/exec in sec: real: 78.58 user: 34.33 syst:
42.75

2.1.99

---------------------------------------
total time for memory in sec: real: 5.27 user: 4.97 syst:
0.22
total time for calls in sec: real: 0.64 user: 0.07 syst: 0.3
total time for pipes in sec: real: 1.65 user: 0.07 syst:
0.33
total time for fork/exec in sec: real: 72.45 user: 34.68 syst:
37.19

--12-> Misc SSBA info
2.0.33 2.1.99
------------- -------------
executable average size: 81486.4 bytes 80358.5 bytes
total time for the SSBA complete run: 1015 seconds 942 seconds

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu