Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Source code to measure it - results

Author: Vincent Diepeveen

Date: 17:19:34 07/15/03

Go up one level in this thread


On July 15, 2003 at 15:24:19, Gerd Isenberg wrote:

Gerd use it with a bigger hashtable. Not such a small
table.

400MB is really the minimum to measure.

So far all tests were with 133Mhz memory.

150Mhz memory a speed of nearly 210 ns is very impressive. Is that cl2 memory?

The tests performed were all with 500MB caches at PC's. Above that
win2000 cannot cache regrettably. So only under linux then you can test well.

Do not know about windows XP. I am sure still same bugs there as i didn't
get answer back from any m$ guy so far whether XP still has the bug.

i have not bought XP yet because of the license policy, i am against it.

The reason to take an as big as possible table is because this tests gets faster
when the amount of RAM drops as in the long run it all fits in L1 table when
small enough.

280 ns * (133 / 150 ) = 248 ns when projecting it.

So 248 ns sounds very fast to me then.


Here is what i get under dual K7 when using 1 processor first then 2 processors:

I took a bit quicker testtime of the RAM.

Not for the RNG measurements of course. Your machines cpu seems very dead slow
compared to this:

Welcome to RASM Latency!
RASML measures the RANDOM AVERAGE SHARED MEMORY LATENCY!

Stored in rasmexename = C:\tries\latency.exe
Trying to allocate 62500000 entries. In total 500000000 bytes
Benchmarking Pseudo Random Number Generator speed, RanRot type 'B'!
Speed depends upon CPU and compile options from RASML,
 therefore we benchmark the RNG
Please wait a few seconds.. ..took 4578 milliseconds to generate  numbers
Speed of RNG = 89471384 numbers a second
So 1 RNG call takes 11.176758 nanoseconds
Benchmarking random RNG test. Please wait..
timetaken=2828
Machine needs 56.560001 ns for RND loop
Clearing hashtable
Took 0 milliseconds to start 0 additional processes
Read latency measurement STARTS NOW using steps of 2 * 12.000 seconds :
Raw Average measured read read time at 1 processes = 433.516176 ns
Now for the final calculation it gets compensated:
  Average measured read read time at 1 processes = 376.956173 ns

but of course this is the MP760 chipset. Known for being slow latency. Dual Xeon
doesn't give much better result however.

Now with 2 cpu's:
C:\tries>latency 500000000 2
Welcome to RASM Latency!
RASML measures the RANDOM AVERAGE SHARED MEMORY LATENCY!

Stored in rasmexename = C:\tries\latency.exe
Trying to allocate 62500000 entries. In total 500000000 bytes
Benchmarking Pseudo Random Number Generator speed, RanRot type 'B'!
Speed depends upon CPU and compile options from RASML,
 therefore we benchmark the RNG
Please wait a few seconds.. ..took 4578 milliseconds to generate  numbers
Speed of RNG = 89471384 numbers a second
So 1 RNG call takes 11.176758 nanoseconds
Benchmarking random RNG test. Please wait..
timetaken=2828
Machine needs 56.560001 ns for RND loop
Clearing hashtable
Took 0 milliseconds to start 1 additional processes
Read latency measurement STARTS NOW using steps of 2 * 12.000 seconds :
Raw Average measured read read time at 2 processes = 436.924107 ns
Now for the final calculation it gets compensated:
  Average measured read read time at 2 processes = 380.364101 ns

RNG call here is considerable slower. First measurement and also second
one time in the millisecond accurate:
  11.176758 nanoseconds that calculates too.

this is 2.127Ghz K7 MP2600s here. How can it be faster than an XP2800?

Or do you mean XP1800 which is clocked at 1.53Ghz?

I compiled with visual c++ sp5 processorpack (the processor pack adds 7% in
speed for diep to vc6sp5).

Yet trivially that diff here is not so important.

>Athlon XP2.8+ ~2GHZ 512MB DDR 300MHz FSB
>
>>latency 300000000 1
>Welcome to RASM Latency!
>RASML measures the RANDOM AVERAGE SHARED MEMORY LATENCY!
>
>Stored in rasmexename = C:\Source\latency\Release\latency.exe
>Trying to allocate 37500000 entries. In total 300000000 bytes
>Benchmarking Pseudo Random Number Generator speed, RanRot type 'B'!
>Speed depends upon CPU and compile options from RASML,
> therefore we benchmark the RNG
>Please wait a few seconds.. ..took 5397 milliseconds to generate  numbers
>Speed of RNG = 75894015 numbers a second
>So 1 RNG call takes 13.176270 nanoseconds
>Benchmarking random RNG test. Please wait..
>timetaken=2935
>Machine needs 58.700000 ns for RND loop
>Trying to Allocate Buffer
>Took 0.000 seconds to allocate Hash
>Clearing hashtable
>Took 0.631 seconds to clear Hash
>Starting Other processes
>Took 0 milliseconds to start 0 additional processes
>Read latency measurement STARTS NOW using steps of 2 * 300.000 seconds :
>Raw Average measured read read time at 1 processes = 266.518624 ns
>Now for the final calculation it gets compensated:
>  Average measured read read time at 1 processes = 207.818615 ns



This page took 0.15 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.