Pick Your Poisson: P-1 Bounds

[storm5510]

Originally Posted by kriesel

I have not incorporated any of the tools folder in the Windows builds I've posted. Until recently it was all python or shell script, no c. I haven't identified a python compilation method yet that both (a) succeeds and (b) produces code small enough to fit in a forum attachment. (But I have eliminated several candidate approaches.)

A person has to understand the language(s) first, then the math. I recognize a few of the languages when I see them. The math is a near total loss on me. I learned enough Strawberry Perl to attempt to get it to deal with some large numbers. Anything beyond 2^1400 was a no-go for it. Of course, it is very possible I was not using it properly.

I found a way to compile a Perl program about a year ago. It was about 35 lines of code in a 2K script file. It became an 830K binary. Talk about turning the hare into the tortoise. It was beyond useless like that.

[kriesel]

Originally Posted by storm5510

I found a way to compile a Perl program about a year ago. It was about 35 lines of code in a 2K script file. It became an 830K binary. Talk about turning the hare into the tortoise. It was beyond useless like that.

What you report for Perl is a several-fold lower bloat factor than what I've seen for Python compilation. I've a Windows license for Perl2Exe from IndigoStar Software. ~2500 lines (131K) becomes ~6MiB executable, compressible to ~3.5MiB .7z. MUCH lower bloat factor than the Python executable generators I've tried on Preda's primenet.py. And it turns out Indigostar Python2Exe is not one I've tried yet. Maybe it's a language characteristic though. http://www.indigostar.com/python2exe/

Perhaps for your short script file, or certainly for "Hello world" in Python, a lot gets pulled in, a lot of which may be unnecessary. In Preda's make file for gpuowl there is a final step called strip. As I understand it, it removes the unreferenced library content that gets added automatically by the ton during the compile & link process. As I recall it cuts the final executable size by more than a factor of two. I have not found the equivalent for Python compilers.

[Xyzzy]

http://timelessname.com/elfbin/

[preda]

Try this online python notebook, let me know how it works:
https://mybinder.org/v2/gh/preda/gpu...lculator.ipynb

For Windows, one may try to install conda/anaconda to get python locally.

Google Collab does not work, as it does not find the import from github (but surely somebody dedicated can get it to work)
https://colab.research.google.com/gi...lculator.ipynb

[storm5510]

Originally Posted by preda

Try this online python notebook, let me know how it works:
https://mybinder.org/v2/gh/preda/gpu...lculator.ipynb

This works quite well. I bookmarked it. I gave it a wavefront exponent, M99785099 I was assigned. Prime95 would use 750,000 and 11,250,000. The script's minimum seems too small, 260,000 and 7,000,000. Mid range is 600,000 and 20,000,000. Big is 1,000,000 and 34,000,000.

[kriesel]

Originally Posted by Xyzzy

http://timelessname.com/elfbin/

Interesting, and at 142 bytes, such bloat. https://www.gnostice.com/nl_article.asp?id=225&t=The_Smallest_Hello_World_Program_At_20_Bytes

If allowing interpreted code, even smaller: https://www.quora.com/What-programmi...rogram?share=1
bash shell script or win/dos bat or cmd files can be short.
Win hi.bat or hi.cmd: (No CR LF) 17 bytes.

echo Hello World!

Powershell: (No CR LF) 12 bytes, the output string only. Invoke by .\hi.

Hello World!

These are smaller than the Python source file's 21 bytes (No CR LF).

print('Hello World!')

[storm5510]

Originally Posted by preda

Try this online python notebook, let me know how it works:
https://mybinder.org/v2/gh/preda/gpu...lculator.ipynb

For Windows, one may try to install conda/anaconda to get python locally.

For some reason, this page will only load from here. If I try the link I have, error 404. I bookmarked it again. Same result. I will take a look at your recommendations.

[James Heinrich]

Originally Posted by preda

There is a new P-1 calculator in the pm1/ folder in GpuOwl.

My probability calculator has been using Mihai's new P-1 probability calculations for a few weeks, but I have just updated it to more clearly show the min/mid/max values that are calculated.

[chalsall]

Originally Posted by preda

For Windows, one may try to install conda/anaconda to get python locally.

And for Perl, I recommend Strawberry Perl. This is apparently the version Larry Wall himself uses when he's not on Unix.

Just to say, I also have found that trying to compile Perl into a Windows exec is not scalable. Not only are the executables huge, but they take a very long time to launch because of all the unpacking of the environment involved.

Not a good User Experience.

Those who get the scripts from this codebase are going to be technically savvy. Presume they already have the interperters installed, or will do so if they wish to use the scripts.

[masser]

Shouldn't a calculator answer the question posed?

Question posed, view the answer here.

[storm5510]

Originally Posted by masser

Shouldn't a calculator answer the question posed?

Question posed, view the answer here.

I have been using James's calculator for quite a while now. He indicates he has updated it. That is great. I will continue to use it.