| T O P I C R E V I E W |
| wasssup1990 |
Posted - Jun 01 2010 : 04:57:17 AM
Try this:
http://www.apfloat.org/apfloat_java/applet/pi.html
It is a Java applet that can calculate pi to as many decimal places as you want. I am interested in seeing what you get.
You need to have the Java Runtime Environment installed though, so go here to get it if you don't have it already.
http://www.java.com/en/
It took several tweaks to overclock my computer so that this applet didn't crash my computer. If you arn't overclocking then you probabbly shouldn't worry too much about your computer crashing with this.
builderFactory = org.apfloat.internal.IntBuilderFactory
maxMemoryBlockSize = 201326592
cacheL1Size = 8192
cacheL2Size = 262144
cacheBurst = 32
memoryTreshold = 201326592
sharedMemoryTreshold = 786432
blockSize = 134217728
numberOfProcessors = 8
Calculating pi to 20000000 radix-10 digits
Using up to 8 parallel operations for calculation
Using the Chudnovsky brothers' binary splitting algorithm
100% complete, elapsed time 110.39 seconds
Final value took 22.78 seconds
That's right, 20,000,000 decimal places.
So try it! |
| 15 L A T E S T R E P L I E S (Newest First) |
| wasssup1990 |
Posted - Jun 05 2010 : 12:07:50 PM
Hahaha! Oh man. No it wasn't.  |
| codingplanet |
Posted - Jun 05 2010 : 11:54:04 AM
Was the semi-colon intentional? :D |
| wasssup1990 |
Posted - Jun 05 2010 : 11:43:49 AM
C++ is fast but in my experience with Java and C++ I'm am more compatible with Java, I have no problem with C++, just that it is a pain cross platforming; |
| codingplanet |
Posted - Jun 05 2010 : 11:18:57 AM
True, VB does lack power. That's why I started using C++. I suppose the OS independent nature of Java is a big plus. |
| wasssup1990 |
Posted - Jun 05 2010 : 10:33:25 AM
Yeah VB using the .net framework and the old plain VB6 just doesn't cut it for me now. I'll probably never use it again. Java seems very nice and isn't attached to a specific operating system. Write code once, compile, and then you can run your program on any machine that has the JVM installed. Easy, and flexible. Hey that gave me an idea, my phone can run Java apps. Ha! My physics simulator definitely won't run on that unless the JVM creates a huge page file or whatever the phone equivalent is. I'll try it anyway but I don't think it'll work. |
| codingplanet |
Posted - Jun 05 2010 : 06:22:10 AM
I haven't done any real programming in a while. And more recently I have been using PHP for web stuff. The language that I am (or was, been about a year now) most comfortable with was VB.Net. |
| wasssup1990 |
Posted - Jun 04 2010 : 8:59:38 PM
Well for me I don't know if it will be very useful but it is very interesting, especially seeing those particles in action. Hopefully running through a number of computer simulations will crack a problem which will allow some massive breakthrough. That's the idea.
What programming language do you primarily use now? |
| codingplanet |
Posted - Jun 04 2010 : 2:06:46 PM
I played around with infinite precision maths once. It was back in the days when I was dealing with C++. I wrote a dirty pi calculator as I remember now. I think it was the Chudnovsky algorithm that I used then, but I can't be sure.
I couldn't say I'm really into partical physics (if that's waht you call it). But I'm sure it does come in very useful :) |
| wasssup1990 |
Posted - Jun 04 2010 : 12:11:51 PM
Yeah I just thought it was all very interesting.
I'm developing some programs in my spare time... well what little spare time I have. My exams are coming.
Do you remember that I was playing around with programming a physics simulator? Well I was doing that in C++ and while it worked well, especially with multi core, I wanted to start fresh with Java instead. I quite like Java and the idea of it running on a virtual machine. It's a self educating thing, plus I enjoy it. I know I can use other third party simulators and graphics engines but they just don't impress me enough with their lack of freedom to do exactly what I want, how I want. I have created a class which creates .bmp images from raw rgb24 data in a 2D array and I am now working on the physics simulator which will be developed in 2D space and graphics for now and then expanded into 3D space. I am not using standard variables like "doubles" because they seriously lack the precision I require. I am using "Apfloat" which you can find easily. It allows finite length decimal precision all the way to infinitive precision but not supporting calculations that produce irrational numbers with infinitive precision obviously.
Here's my sample of how to calculate the distance between two or more positions in 2D space. It is withing a Java utility class that I wrote.
private static final int irrationalPrecision = 100;
//Will calculate the distance, in order, between all Particles in the arg.
public static Apfloat distance(Particle... args) {
// return sqrt((b.x-a.x)^2 + (b.y-a.y)^2)
Apfloat sumOfDistances = new Apfloat(0, Apfloat.INFINITE);
for (int i = 0; i < args.length - 1; i++) {
sumOfDistances = sumOfDistances.add(ApfloatMath.sqrt((ApfloatMath.pow((args[i + 1].x.subtract(args[i].x)), 2).add(ApfloatMath.pow((args[i + 1].x.subtract(args[i].x)), 2))).precision(irrationalPrecision)));
}
return sumOfDistances;
}
As you can see to calculate sqrt((b.x-a.x)^2 + (b.y-a.y)^2) is much more complicated than simply writing Math.sqrt((b.x-a.x)*(b.x-a.x) + (b.y-a.y)*(b.y-a.y)) if I was to use standard "double" data types. |
| codingplanet |
Posted - Jun 04 2010 : 08:56:12 AM
I couldn't say really. Could be either of the two things that you mentioned. |
| wasssup1990 |
Posted - Jun 04 2010 : 08:46:08 AM
Yeah what do you think it was caused by? Perhaps the algorithm being used does not work great on multicore. Or maybe the JVM needs improvement. |
| codingplanet |
Posted - Jun 04 2010 : 05:42:12 AM
I have an i7 920 too. The 4.6GHz was on air yes. Although I had only 2 cores enabled and HT disabled because of the heat.
Scaling isn't too good on the 32M scores you posted. Less than 3.5 :/ |
| wasssup1990 |
Posted - Jun 01 2010 : 09:25:25 AM
Same test conditions but now only 1 thread and 32,000,000 digits.
builderFactory = org.apfloat.internal.IntBuilderFactory
maxMemoryBlockSize = 201326592
cacheL1Size = 8192
cacheL2Size = 262144
cacheBurst = 32
memoryTreshold = 201326592
sharedMemoryTreshold = 786432
blockSize = 134217728
numberOfProcessors = 1
Calculating pi to 32000000 radix-10 digits
Using the Chudnovsky brothers' binary splitting algorithm
100% complete, elapsed time 607.298 seconds
Final value took 69.527 seconds
Now 8 threads.
builderFactory = org.apfloat.internal.IntBuilderFactory
maxMemoryBlockSize = 201326592
cacheL1Size = 8192
cacheL2Size = 262144
cacheBurst = 32
memoryTreshold = 201326592
sharedMemoryTreshold = 786432
blockSize = 134217728
numberOfProcessors = 8
Calculating pi to 32000000 radix-10 digits
Using up to 8 parallel operations for calculation
Using the Chudnovsky brothers' binary splitting algorithm
100% complete, elapsed time 176.13 seconds
Final value took 27.98 seconds |
| wasssup1990 |
Posted - Jun 01 2010 : 09:10:32 AM
That's right I have an i7-920 overlocked to 3790MHz (Spread Spectrum since near lab equipment). QPI Bus Speed 200MHz, 19 (not 20) CPU freq. multiplier with "Turbo Mode". Voltages are obviously ramped up and speed step is turned off. Memory speed 1602MHz.
What model is you CPU? I'd be surprised if you got 4.6GHz with a i7-920 on air cooling only. I have got my CPU up to 4GHz but it is too unreliable. My CPU is supposed to run at a stock speed of 2.67GHz!
Test conditions:
Ambient Temp: 20.5C
High Air Flow computer case. |
| codingplanet |
Posted - Jun 01 2010 : 08:39:35 AM
This was run on an i7 I presume. At what speed? I haven't come across any multi-threaded pi benchmarks up to yet (except HyperPi which just runs many instances of SuperPi).
It take my i7 just under 8 minutes to calculate 32 Million digits in a single thread using SuperPi @ 4.6GHz. |
