[BENCHMARK] Lmbench result for 5.52 mm2 patch.

Aniruddha M Marathe (aniruddha.marathe@wipro.com)
Thu, 19 Dec 2002 13:38:24 +0530


Hi,
Here are the results of Lmbench on 2.5.52-mm2. I have made a table of prominent changes. The table entries are median values of 5 results.
* mmap latency that was on rise for some 5.52 and mm1 patches, is now decreased.
* also fork proc is giving better performance now.

2.5.52-mm2 2.5.52-mm1
==============================================================================
Processor, Processes - times in microseconds - smaller is better

1.fork proc 441 400
------------------------------------------------------------------------------
Context switching - times in microseconds - smaller is better

1. 2p/0K ctxsw 1.390 1.290
2. 2p/16k ctxsw 5.08 4.62
------------------------------------------------------------------------------
*Local* Communication latencies in microseconds - smaller is better
1. 2p/0K 1.390 1.290
------------------------------------------------------------------------------
File & VM system latencies in microseconds - smaller is better
1. mmap latency 597 640
------------------------------------------------------------------------------
*Local* Communication bandwidths in MB/s - bigger is better
1. pipe 404 335
==============================================================================

Here are the complete results for mm2 patch.

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 2.5.52 i686-pc-linux-gnu 790
benchtest Linux 2.5.52 i686-pc-linux-gnu 790
benchtest Linux 2.5.52 i686-pc-linux-gnu 790
benchtest Linux 2.5.52 i686-pc-linux-gnu 790
benchtest Linux 2.5.52 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 2.5.52 790 0.44 0.82 28 29 1.28 5.20 458 1546 7942
benchtest Linux 2.5.52 790 0.44 0.80 28 29 34 1.28 5.21 445 1564 8118
benchtest Linux 2.5.52 790 0.44 0.80 28 29 32 1.28 5.42 351 1550 8021
benchtest Linux 2.5.52 790 0.46 0.82 28 30 32 1.30 5.43 441 1547 8096
benchtest Linux 2.5.52 790 0.46 0.83 28 30 34 1.30 5.42 381 1572 8048

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 2.5.52 1.390 5.0800 15 8.2900 178 43 180
benchtest Linux 2.5.52 1.360 5.2000 14 8.8000 178 42 180
benchtest Linux 2.5.52 1.430 4.7800 17 6.9300 181 46 181
benchtest Linux 2.5.52 1.370 5.1800 14 6.0600 179 45 180
benchtest Linux 2.5.52 1.450 5.2200 14 9.0300 179 42 180

*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 2.5.52 1.390 8.173 21 34 59 125 156 171
benchtest Linux 2.5.52 1.360 7.991 21 35 60 127 156 171
benchtest Linux 2.5.52 1.430 8.242 21 35 59 126 156 175
benchtest Linux 2.5.52 1.370 8.415 21 35 59 125 158 174
benchtest Linux 2.5.52 1.450 8.335 21 35 60 125 156 173

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 2.5.52 118 58 373 118 586 0.912 4.00000
benchtest Linux 2.5.52 118 58 381 129 599 0.899 4.00000
benchtest Linux 2.5.52 119 58 381 127 587 0.891 4.00000
benchtest Linux 2.5.52 119 58 379 129 597 0.903 4.00000
benchtest Linux 2.5.52 119 58 377 126 601 0.980 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 2.5.52 335 137 23 299 355 124 113 355 171
benchtest Linux 2.5.52 432 130 22 295 352 123 112 353 169
benchtest Linux 2.5.52 322 129 22 295 352 123 112 352 169
benchtest Linux 2.5.52 407 134 22 293 352 123 112 352 168
benchtest Linux 2.5.52 292 131 22 290 350 124 113 350 168

Memory latencies in nanoseconds - smaller is better
(WARNING - may not be correct, check graphs)
---------------------------------------------------
Host OS Mhz L1 $ L2 $ Main mem Guesses
--------- ------------- ---- ----- ------ -------- -------
benchtest Linux 2.5.52 790 3.799 8.8840 174
benchtest Linux 2.5.52 790 3.807 8.8830 176
benchtest Linux 2.5.52 790 3.798 8.8830 176
benchtest Linux 2.5.52 790 3.808 8.8740 176
benchtest Linux 2.5.52 790 3.808 56 177

Regards,
Aniruddha Marathe
WIPRO Ttechnologies, India
Aniruddha.marathe@wipro.com
+91-80-5502001 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/