lkml.org 
[lkml]   [2003]   [Jan]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Subject[BENCHMARK] Lmbench results for 2.5.54
Date
From
Hi,
Here is a comparison of results of 2.5.52 and 2.5.53. The figures in the table below indicate median of 5 repetitions of tests. This result doesn't have many
Differences than the previous one.


2.5.54 2.5.53
==============================================================================
Processor, Processes - times in microseconds - smaller is better

1. Null call 0.44 0.46
2. stat 29 27
3. exec proc 1600 1528
------------------------------------------------------------------------------
Context switching - times in microseconds - smaller is better

1. 2p/0K ctxsw 1.560 1.430
------------------------------------------------------------------------------
*Local* Communication latencies in microseconds - smaller is better
1.AF UNIX 19 21
2.UDP 30 33
3.TCP 75 125
4.RPC TCP 106 156
------------------------------------------------------------------------------
File & VM system latencies in microseconds - smaller is better
1. PIPE 554 386
2. AF UNIX 109 136
3.
==============================================================================


*****************************************************************************
Lmbench result
kernel 2.5.54
****************************************************************************

L M B E N C H 2 . 0 S U M M A R Y
------------------------------------
(Alpha software, do not distribute)

Basic system parameters
----------------------------------------------------
Host OS Description Mhz

--------- ------------- ----------------------- ----
benchtest Linux 2.5.54 i686-pc-linux-gnu 790
benchtest Linux 2.5.54 i686-pc-linux-gnu 790
benchtest Linux 2.5.54 i686-pc-linux-gnu 790
benchtest Linux 2.5.54 i686-pc-linux-gnu 790
benchtest Linux 2.5.54 i686-pc-linux-gnu 790

Processor, Processes - times in microseconds - smaller is better
----------------------------------------------------------------
Host OS Mhz null null open selct sig sig fork exec sh
call I/O stat clos TCP inst hndl proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ----- ---- ---- ---- ---- ----
benchtest Linux 2.5.54 790 0.44 0.80 29 30 1.30 5.38 385 1541 7903
benchtest Linux 2.5.54 790 0.44 0.83 29 30 38 1.28 5.16 322 1528 7872
benchtest Linux 2.5.54 790 0.44 0.82 29 30 34 1.29 5.39 365 1534 7852
benchtest Linux 2.5.54 790 0.46 0.80 29 30 35 1.29 5.16 312 1491 7862
benchtest Linux 2.5.54 790 0.46 0.84 29 30 34 1.33 5.38 426 1511 7948

Context switching - times in microseconds - smaller is better
-------------------------------------------------------------
Host OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw
--------- ------------- ----- ------ ------ ------ ------ ------- -------
benchtest Linux 2.5.54 1.450 4.8900 14 8.9400 164 43 177
benchtest Linux 2.5.54 1.570 5.2200 14 6.1100 181 39 179
benchtest Linux 2.5.54 1.620 5.2600 26 6.2900 181 41 179
benchtest Linux 2.5.54 1.560 4.9000 14 8.5100 176 42 179
benchtest Linux 2.5.54 1.520 5.2000 15 7.0300 174 39 179

*Local* Communication latencies in microseconds - smaller is better
-------------------------------------------------------------------
Host OS 2p/0K Pipe AF UDP RPC/ TCP RPC/ TCP
ctxsw UNIX UDP TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
benchtest Linux 2.5.54 1.450 8.649 19 30 54 76 106 146
benchtest Linux 2.5.54 1.570 8.426 19 30 55 77 106 147
benchtest Linux 2.5.54 1.620 8.668 19 30 55 76 108 146
benchtest Linux 2.5.54 1.560 8.522 19 30 54 76 106 146
benchtest Linux 2.5.54 1.520 8.630 19 30 55 76 108 147

File & VM system latencies in microseconds - smaller is better
--------------------------------------------------------------
Host OS 0K File 10K File Mmap Prot Page
Create Delete Create Delete Latency Fault Fault
--------- ------------- ------ ------ ------ ------ ------- ----- -----
benchtest Linux 2.5.54 118 55 381 128 594 1.015 4.00000
benchtest Linux 2.5.54 118 55 378 126 606 1.003 4.00000
benchtest Linux 2.5.54 118 56 374 119 598 1.017 4.00000
benchtest Linux 2.5.54 118 56 374 127 599 1.005 4.00000
benchtest Linux 2.5.54 118 56 375 126 626 1.022 4.00000

*Local* Communication bandwidths in MB/s - bigger is better
-----------------------------------------------------------
Host OS Pipe AF TCP File Mmap Bcopy Bcopy Mem Mem
UNIX reread reread (libc) (hand) read write
--------- ------------- ---- ---- ---- ------ ------ ------ ------ ---- -----
benchtest Linux 2.5.54 559 110 25 296 354 124 112 353 170
benchtest Linux 2.5.54 356 98 25 296 353 124 112 352 169
benchtest Linux 2.5.54 580 105 25 294 353 124 113 352 169
benchtest Linux 2.5.54 554 109 24 295 353 124 112 354 170
benchtest Linux 2.5.54 544 114 25 287 354 124 114 354 170

Memory latencies in nanoseconds - smaller is better
(WARNING - may not be correct, check graphs)
---------------------------------------------------
Host OS Mhz L1 $ L2 $ Main mem Guesses
--------- ------------- ---- ----- ------ -------- -------
benchtest Linux 2.5.54 790 3.799 8.8820 175
benchtest Linux 2.5.54 790 3.806 60 175
benchtest Linux 2.5.54 790 3.799 8.8950 175
benchtest Linux 2.5.54 790 3.807 8.8820 175
benchtest Linux 2.5.54 790 3.807 8.8820 175

Aniruddha Marathe
WIPRO Technologies, India
aniruddha.marathe@wipro.com
+91-80-5502001 to 2008 extn 5092

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:32    [W:0.046 / U:3.964 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site