Win32 VM updates & benchmarks

Dwight Hughes dwighth at ipa.net
Fri Sep 10 20:32:10 UTC 1999


On a PII/300, NT 4.0, sp5, on the 2.5 build 6 VM, I am seeing:

 '0.944 microseconds per call to Time>>millisecondClockValue'
 '0.061 microseconds overhead per primitive call'
and on average about
 13,850,000 bytecodes/sec; 825,000 sends/sec

Note: I am seeing rather inconsistent results in bytecodes/sec on the
"standard" Squeak benchmark - especially on faster processors. Your
bytecodes/sec results on a PII/450 should be up in the 18M-19M range. My
results above are from running:

'From Squeak 2.5 of August 6, 1999 on 10 September 1999 at 1:21:41 pm'!

!Integer methodsFor: 'benchmarks' stamp: 'dwh 9/10/1999 13:21'!
tinyBenchmarks
	"Report the results of running the two tiny Squeak benchmarks."
	"0 tinyBenchmarks"
	"On a 292 MHz G3 Mac: 22727272 bytecodes/sec; 984169 sends/sec"

	| n t1 t2 r |
	n _ 200.
	t1 _ Time millisecondsToRun: [n benchmark].
	t2 _ Time millisecondsToRun: [r _ 32 benchFib].
	^ ((n * 500000 * 1000) // t1) printString, ' bytecodes/sec; ',
	  ((r * 1000) // t2) printString, ' sends/sec'
! !

Where I changed n _ 1 to n _ 200 and 28 benchFib to 32 benchFib. This
makes these benchmarks take several seconds each and so far gives much
more stable and reasonable results.

-- Dwight

Dean_Swan at Mitel.COM wrote:
> 
> Andreas,
> 
> Here are benchmark results for my PC at work.  I'll try this on my P133 at home
> also.
> 
>      PII/450, NT 4.0, SP 4
> 
>      '0.665 microseconds per call to Time>>millisecondClockValue'
>      '0.043 microseconds overhead per primitive call'
> 
>       '16666666 bytecodes/sec; 1222897 sends/sec'
> 
> Have you rebuilt the updated VM for WinCE?
> 
>                          -Dean Swan





More information about the Squeak-dev mailing list