2002-12-13 04:51:37

by Aniruddha M Marathe

Subject: [BENCHMARK] short summary of LM bench performance of mm2 patch

Hi,
Josh, please comment on the format.
Here is a short summary of performance of 2.5.51 mm2 as compared to 2.5.51 mm1 on LM bench. Figures below are 'median' of 5 results.

-------------------------------------------------------------------------------
For detailed results, please mail me so that I will send results individually.
-------------------------------------------------------------------------------

2.5.51 mm1 2.5.51-mm2
==================================================================================
Processor, Processes - times in microseconds - smaller is better

1. Null call time decreased a bit. 0.46 0.44
2. increase in time for fork proc 362 403
----------------------------------------------------------------------------------
Context switching - times in microseconds - smaller is better

1. 2p/16K ctxsw 5.01 4.84
----------------------------------------------------------------------------------
*Local* Communication latencies in microseconds - smaller is better
1. UDP 33 35
2. TCP 122 125
----------------------------------------------------------------------------------
File & VM system latencies in microseconds - smaller is better
1. 10K file delete 57 59
2. mmap latency 615 651
==================================================================================

Rest of the results are not much different.


Regards,
Aniruddha Marathe
WIPRO technologies, India
[email protected]
+91-80-5502001 extn 5092

2002-12-13 04:54:23

by Larry McVoy

Subject: Re: [BENCHMARK] short summary of LM bench performance of mm2 patch

On Fri, Dec 13, 2002 at 10:29:11AM +0530, Aniruddha M Marathe wrote:
> Hi,
> Josh, please comment on the format.

Well, I'm not josh, but what is wrong with the default LMbench output
format? It's much easier to see what's going on than what you posted.
--
---
Larry McVoy lm at bitmover.com http://www.bitmover.com/lm