[Trial]Contest 1 - VB (.NET & Legacy) - Prime Numbers

First of all, I'd never touch .NET unless I'd need it in my job (and my current job isn't programming related ;)). So yeah, I'm doing it with pure VB6: no API, no type libraries, no assembler etc. With a listbox it takes the steady 5 ms to show 1000 primes in the listbox (9 ms in the IDE). I formatted my code so that if the people who time the codes want to, they can leave out the listbox.

As for the algorithm, it is my own, since I didn't find any premade one. This is the first time I'm doing something with the prime numbers :) I have no idea what algorithms others might have.

Current version can count up to 32767 prime numbers (as many as you can put in the listbox without errornous behavior) at about 250 ms, listbox slowdown included (730 ms in IDE). I might submit this as it is now, as I guess I can do only minor optimization from now on that doesn't really matter.


Edit I got interested on seeing more accurate results. I took out the listbox from the timing, made the whole algorithm loop 100 times and these are my results for single run times:

- 0,83 ms to get 1000 primes compiled
- 5,22 ms to get 1000 primes under IDE
- 117 ms to get 32767 primes compiled
- 620 ms to get 32767 primes under IDE

What do we win?

I can use VB6 right? I've never heard it called legacy before...

Use a PerformanceCounter to do code tests...that has been the choice for the last 10 years as far as I know. Most routines run so fast that gettickcount or window timer never can get an accurate reading. Performance counter works in nano-seconds so it is accurate...almost.

This is a little to easy, and being there are only a few ways to achieve this, you should declare you want a unique way of finding prime numbers...Im' sure people could come up with a million mods of the basic prime number routine...I'd like to see something really unique.

Hows .0455 Seconds, compiled, Avweraged over 11 loops, Clearing and Populating the listbox every loop, sound on a 400 mHz HP?

:wave:
-Lou

Quote:

---

Originally posted by Halsafar
What do we win?

I can use VB6 right? I've never heard it called legacy before...

Use a PerformanceCounter to do code tests...that has been the choice for the last 10 years as far as I know. Most routines run so fast that gettickcount or window timer never can get an accurate reading. Performance counter works in nano-seconds so it is accurate...almost.

This is a little to easy, and being there are only a few ways to achieve this, you should declare you want a unique way of finding prime numbers...Im' sure people could come up with a million mods of the basic prime number routine...I'd like to see something really unique.

---

Hey halsafar,

firstly this is a test contest so its nothing too hard (but still makes people think) just to get everyone into the flow and routine of the contests, i think you should take a look at the FAQ located at the top of this forum it tells you everything you need to know :) there are no major prizes, the contest is just fun coding where you can learn and hopefully win :) this contests' winner gets a custom name title :) and after speed some marks are awarded for the 'sexyness' and origionality of your code :)

Pino

Quote:

---

Originally posted by NotLKH
Hows .0455 Seconds, compiled, Avweraged over 11 loops, Clearing and Populating the listbox every loop, sound on a 400 mHz HP?

:wave:
-Lou

---

That sounds fast....can you mail me the compiled version so I can test it on my machine...???? Please?

Quote:

---

Originally posted by Merri
[B]First of all, I'd never touch .NET unless I'd need it in my job (and my current job isn't programming related ;)). So yeah, I'm doing it with pure VB6: no API, no type libraries, no assembler etc. With a listbox it takes the steady 5 ms to show 1000 primes in the listbox (9 ms in the IDE). I formatted my code so that if the people who time the codes want to, they can leave out the listbox.

---

What happend to the 1ms with showing the list box...did you cheat...;)

Quote:

---

Originally posted by NoteMe
That sounds fast....can you mail me the compiled version so I can test it on my machine...???? Please?

---

Sure.

Of course, it could be only my imagination.

Hmmm, How are you going to test it compiled, if you don't have access to the code, to throw in a timer?

Or would you like the compiled one with my timing event?

err, well, the looping timer event is done with the button labeled Command 1.

INCOMING!!!
:D

INcomming? do you know my mail? Well send it to note(a)broadpark.no and yeah, add the timer too it...;) If not it won't help much...:)

Quote:

---

Originally posted by NoteMe
That sounds fast....can you mail me the compiled version so I can test it on my machine...???? Please?

---

Hmmm, can't seem to find your email button.
PM's don't allow attachements...

Ah well. I'll attach it.

:wave:

Let me know what you find!

-Lou

[EDITED: SEE BELOW]

Quote:

---

Originally posted by NoteMe
INcomming? do you know my mail? Well send it to note(a)broadpark.no and yeah, add the timer too it...;) If not it won't help much...:)

---

Ahh, well, assuming (a) means @, I've just sent it off.

Assuming you got it, then I'll go ahead and delete my attachment, above.

:wave:

Hmmmm....very weird...I get 9.9ms every time...hmmm...well at least thats not bad even if it is weird it gives me the same time every time...still it is more then twice as slow as mine tho...;)

Quote:

---

Originally posted by NoteMe
What happend to the 1ms with showing the list box...did you cheat...;)

---

I never said I got 1 ms with a listbox ;) If you reference to the EXE I posted, it shows the timing before filling up the listbox.

Of course I can try to optimize the listbox filling too... :D

Just curious,
I changed a couple of lines,
Whats the speed now?

:wave:
-Lou

Heh!

I forgot about Properly titling my entry, ie...

"Contest 1 Prime Numbers - Electroman"

All better now!

:p
-Lou

Quote:

---

Originally posted by Merri
I never said I got 1 ms with a listbox ;) If you reference to the EXE I posted, it shows the timing before filling up the listbox.

Of course I can try to optimize the listbox filling too... :D

---

Aha! So us mere mortals do stand a slim chance after all!

:lol:

ph34r w055n@|\/|e, (and his sloppy coding style) :D

...or maybe not. I figured out a new algorithm, at the moment if gives errorneus results, but if I manage to find a way to prevent the errors, it is awful fast: my current results for this new algorithm under IDE is 0,2 ms for 1000 and 6,3 ms for 32767. It will slow down though, atm I don't know how much. I just hope I can really get past the errors...


Edit Too bad, the speed is too slow with only 1000 items. I had to code sorting routine to fix the error and that ruined the speed. Now looking for alternative methods...

Quote:

---

Originally posted by NotLKH
Just curious,
I changed a couple of lines,
Whats the speed now?

:wave:
-Lou

---

Are you kidding with me...you must do. Well after first try I got 7.1ms...every time...no change at all...you must be kidding with me...well that is fast...but not fast enough....;)...but you have probably cuter code then me tho'.....:D

Cool! I just love contests! :)

Im down to ~10ms on my 1.9 GHz machine. :blush:

Damn... I forgot to do the
Contest 1 Prime Numbers - Mendhak thing.

Oy Electroman, go edit my thread and make the appropriate changes!

Quote:

---

Posted by NotLKH
Heh!

I forgot about Properly titling my entry, ie...

"Contest 1 Prime Numbers - Electroman"

All better now!

:p
-Lou

---

:lol:

Was wondering what was going on when i saw three posts by you :lol:



Mendhak: I hadn't actually noticed you'd used the title "All your prime numbers are belong to us" first of all, will edit it.... :lol:

Hey guys, is this fast enough?

I am just taking in consideration the algorithm...

:wave:

http://www.vbforums.com/attachment.p...postid=1785845

yaba daba doo ;)

r we allowed to hardcode values :confused: , like this:

VB Code:

For iCtr = 2 To 7919
    IsPrimeNo(iCtr)
Next

my app displays 0.232493515237272. how many ms is this ?

Quote:

---

Originally posted by Deepak Sakpal
my app displays 0.232493515237272. how many ms is this ?

---

Depends. How did you calculate it?

i used sTime class provided by NoteMe

Quote:

---

Originally posted by Deepak Sakpal
my app displays 0.232493515237272. how many ms is this ?

---

232ms....;)

Quote:

---

Originally posted by Tec-Nico
Hey guys, is this fast enough?

I am just taking in consideration the algorithm...

:wave:

http://www.vbforums.com/attachment.p...postid=1785845

---

I don't hope that is timed with the adding to the list box...then I am but ****ed. Even if it is not, it is still fast...;)

Quote:

---

Originally posted by Electroman

Mendhak: I hadn't actually noticed you'd used the title "All your prime numbers are belong to us" first of all, will edit it.... :lol:

---

Did I really do that? :ehh:

Quote:

---

Originally posted by NoteMe
Are you kidding with me...you must do. Well after first try I got 7.1ms...every time...no change at all...you must be kidding with me...well that is fast...but not fast enough....;)...but you have probably cuter code then me tho'.....:D

---

Thought of a couple of more lines to tweak on they way home from work yesterday. Should bump it down to more acceptable standards.

Hey, It'd be interesting if they had preliminary results on the codes, to provide incentive for us to really max out our codes.

:wave:
-Lou

Think I have stuck to the rules

tends to be 0.03075 ms

Its VB6.0 btw

think its a 1400 Duron btw :)

Its very difficult to tell at these speeds because background processing affects the numbers dramatically even when you have no apps running

0.030375
0.031

etc

had it down to 0.01 something before but it was coincidence I think :D

Hey BudwadUK... Your program only does up to 1197.. The 1000nd prime is 7919.

poop

you sure :lol:

Oops I was fiddling with my code too much :cry:

BRB

Damn! I just found a stupid thing in my program that adds 2 ms...
Is it possible to send the improved version?

Can we make assumptions e.g do we HAVE to calculate by checking all numbers or can we skip them if mathmatically they cant be prime numbers :)

I think that's ok...
I don't check even numbers (except 2) because they can't be primes.

Could you check this for us :D

Is the timing in seconds? I get 0.15-0.17 compiled.

I get around 0.008s on my app, but on the one I sent in I get around 0.01s...

I did it originally like that and thought it would be classed as cheating because the system hadnt really worked it out :D

0.2 compiler duron 1400 or so :D

0.12 if i compile with advanced debugging options set :D

Yeah... Im down to 0.0074 with the advanced options on.

Woohoo! I finally found out a superb routine! 0,03 ms compiled and 1,1 ms in IDE for 1000 primes. It gives correct results too! :D

Gotta download a better timing code, I had to use GetTickCount and loop the code 10000 times to get it accurate... worked out the whole thing on my laptop. I love modem :)


Edit 51 ms to find 41538 primes (all primes below 500000).

Get the sTime module... It's pretty accurate.

What machine are you running on?

Athlon XP-M 2400+ @ 1800 MHz

Btw, the results went down when I switched to sTime. Happiness.

Quote:

---

Originally posted by BodwadUK
Can we make assumptions e.g do we HAVE to calculate by checking all numbers or can we skip them if mathmatically they cant be prime numbers :)

---

I use 2 as a hard coded prime.

Quote:

---

a comment in my progie
'Also, since 1 is not a prime, and since the only whole positive integer below 2 is 1, then
'2 is the first prime

---

I've heard someone say they need the first 2 primes for their progie to work.

Some people, I expect, are sequencing with 6n Plus or Minus 1, which is an inherent usage of 2 and 3 as primes.

Obviously, the more primes you hard code, the faster your code will run.

Perhaps a point system should be applied, allowing you usage of n primes hard coded, but it lowers your overall for every one you use.

-Just a thought
-Lou

As for the hardcoded primes thing we decided you could hardcode as many as you wanted but be warned you will loose points for "sexiness" of the algo if you go over the top ;).

BTW Bodwad are you not entering your code? Just cos you're keep posting it publicly :confused:.

Is it still finishing on the 25th?

Dont we just post it here :confused:

:rolleyes:



I can only hope you are joking :lol:.

Quote:

---

Originally posted by BodwadUK
Is it still finishing on the 25th?

Dont we just post it here :confused:

---

you have to post your code in the contest entries forum :) otherwise people can steal your code :-(

http://vbforums.com/forumdisplay.php?s=&forumid=68

make a new thread and it will not be visible :)

How strange nobody commented about my 0.000028495 seconds...

I made it on a Pentium II. :ehh:

Which is the minimum so far? Please show your times like the messagebox does or pass it to seconds like I did.

Greetings... :)

Bodwad, use NoteMe's code (Platipus') to measure the time...

I think yours go around the 0.9 seconds and I believe you can do better than that.

Quote:

---

Originally posted by Merri
Woohoo! I finally found out a superb routine! 0,03 ms compiled and 1,1 ms in IDE for 1000 primes. It gives correct results too! :D

---

Well, mine was 0,5 ms when I posted...just figured a small improvement and it went down to 0,14 ms for debug configuration and 0,103 ms for release configuration (P4@2,4Ghz, 1GB RAM,W2K Prof,VS.NET 2002). But I think that that's the bottom line for me, I can't see how to squeeze that anymore. Out of curiosity Merri, what's the box you're running your code on ?

Cheers,
NTG

PS: OK. I got curious and wrote it for VB6. With optimizations, runtime gives me 0,076 ms on the same machine.

Quote:

---

Originally posted by Tec-Nico
How strange nobody commented about my 0.000028495 seconds...

I made it on a Pentium II. :ehh:

Which is the minimum so far? Please show your times like the messagebox does or pass it to seconds like I did.

Greetings... :)

---

How strange that you didn't notice that I commented it and even asked a question or two about it....;)

Came to tell I dropped below Tec-Nico's reported value, though I have to admit I have a more powerful computer. As for the computer specs, I have told my both computers' specs a few times already, just scroll up ;)

My current result is 0,024 ms and 0,032 ms for the second algorithm. Yea, I have two and I'm working on both. The latter one is younger (just wrote it), so there is still stuff to optimize. I'm really encouraging people, am I not? :D

Edit My newer algorithm passes the older one with big primes: the more primes to count, the faster the new algorithm is in comparison. But I need to make it faster in shorter interval too!

>>> But I think that that's the bottom line for me.
Me and my big mouth. :rolleyes: Sometimes I keep forgeting that the obvious things are always really hard to find. Saw a couple of things again and dropped to 0,06ms for release environment.

What troubles me, however, is something completely trivial that I really stumbled upon. I had a line of code:

While j < 7920 / 2

Innocent, right ? I figured that the compiler would know that this is a constant and substitute for the actual value of 3960. Built (always release env) and I get 0,085 ms. Changed the code to:

While j < 3960

Built again and I get 0,06 ms :eek: ! Geez, I don't know about other people but this came as a surprise to me. The compiler created different IL code for it. Still haven't figured it out completely but there's a conversion to float when I use 7920 / 2. OK, runtime doesn't have a way of knowing what the division will produce so it assumes float but hell, the compiler should figure out that 7920 / 2 is an integer and substitute it! When I changed the statement to While j < 7920 \ 2, I got 0,06 ms again.

Cheers,
NTG

Yeah, NoteMe, I noticed so... But nobody else besides you made a comment about it.

I am working on a better algorithm to see if I can go below what I achieved.