1. Regarding EU benchmarks
------------10C1F1162183271A3
I tried Eu 2.4 benchmark with a program in demo/bench/sieve8k.exw
but the result is not as good as in bench.doc.
My PC is Pentium 4 1600, just as in Shootout benchmark.
Try to run hit.bat to get the result.
I changed sieve8k.exw by removing:
- include get.e (so eu will not search for that file)
- parsing of command line (now 900 is hard-coded)
- timing (I use external timing)
- printing into screen
Also I used exwc.exe to prevent Windows making a new console window.
When you run hit.bat, there will be 4 numbers:
- start time of measuring one eu start-up time (in ms)
- finish time of measuring one eu start-up time
- start time of measuring sieve, including 2 start-up times
- finish time of measuring sieve, including 2 start-up times
I got this:
7929516
7929566
7929621
7930686
so: eu startup time = 50 ms
sieve + 2 eu startup times = 1065 ms
sieve + 1 eu startup time = 1065 - 50 = 1015 ms
Second test:
7981467
7981516
7981556
7982656
so: eu startup time = 49 ms
sieve + 2 eu startup times = 1100 ms
sieve + 1 eu startup time = 1100 - 49 = 1051 ms
So it's different than that on the bench.doc (470 ms)
What makes it very different?
------------10C1F1162183271A3
Content-Type: application/x-zip-compressed; name="BENCH.zip"
2. Re: Regarding EU benchmarks
Aku writes:
> So it's different than that on the bench.doc (470 ms)
> What makes it very different?
In bench.doc I've compared all times with the start-up time
removed. I also ran the Euphoria program with 90000, rather
than 900, in order to get a more accurate measurement.
For the other languages, 900 lets the program run for a reasonable
amount of time. For Euphoria, it only takes a fraction of a second.
You can't time less than a second very reliably, especially
with other stuff going on in the background on your system.
Although Euphoria starts up as fast or faster than the others,
I wasn't interested in that. It varies too much depending
on whether the O/S has the program cached in memory cache,
or on disk etc. Most people want to know how fast their code
will run, not whether the interpreter takes 0.2 or 0.5 seconds to start.
When I run your .bat, I get numbers that are reasonably in line with
what I reported. Maybe your cache memory is different. The 8K
sieve is very sensitive to cache effects.
Regards,
Rob Craig
Rapid Deployment Software
http://www.RapidEuphoria.com
