MEGAPOST apps to compute mersenne primes, factorization and computing decimals of pi on android

[kriesel]

It would help if thorken would refrain from polluting the forum with incoherent test, atrociously formatted, 99+% bold. A little. Root requirement is the big issue. The complete pointlessness of a badly scaling application is another.

[thorken]

https://www.mersenneforum.org/showthread.php?t=23963

[Batalov]

Originally Posted by thorken

https://www.mersenneforum.org/showthread.php?t=23963

Well, that's a message that has 52 letters and has zero meaning.
What were you trying to say?

[LaurV]

Originally Posted by kriesel

It would help if thorken would refrain from polluting the forum with incoherent test, atrociously formatted, 99+% bold. A little. Root requirement is the big issue. The complete pointlessness of a badly scaling application is another.

Indeed. Right now, the crank score is incredible high. I won't waste my time to click on those links.

[preda]

Comme on guys, let's take it easy. I think I see a genuine attempt to make something useful and I applaud the effort. If it's not perfect I would like to see the effort continue and improve, rather than scare the developer away.

@thorken, did you consider publishing your project on github? (or other similar, e.g. gitlab or bitbucket). You could choose an open-source license that you consider appropriate. At once you'd dispel the mistrust visible here. Hopefully you may even get help in various forms. And maybe, others will continue and build on top of your effort.

PS: users of Spain welcome.

[thorken]

all I'm saying is that whoever wants to try the applications that try them and then say if it's malware or not is malware and the source code is in the videos I've put, I do not want to upload them anywhere for now. they are not going to solve the world, just as I had seen that in android there was no program that did this very quickly I started to do something better than what was already there, but let's not make any effort but the one that wants to try them in a cell phone that does not use and has root so you can see that nothing is happening or it is not a virus or anything strange. Greetings.

[kriesel]

Preda knows what he's talking about. He introduced new software, gpuowl. He posted his source, many versions, on github. He was open about what algorithm was being employed. He took suggestions from users and potential users, implemented some of them, and the end result is the better for it. His software is accepted and in use because he was willing to post source and make requested modifications. (Millions of programmers use Github. Many use SourceForge.)

What he did not do, is limit source access to videos, documentation to screen shots, generally making it very inconvenient, cumbersome, slow, uninviting, for the potential end user, or require root or admin privilege.

[kriesel]

Originally Posted by Batalov

Well, that's a message that has 52 letters and has zero meaning.
What were you trying to say?

Perhaps, a counterexample to using almost all bold text.

[kriesel]

From the second phone screenshot in the first post: Numeros De Quijotes Procesados 0.849 Quijotes

https://translate.google.com/#view=h...849%20Quijotes
Number of Processed Legplates 0.849 Legplates
(What's a legplate? Part of a conquistador's armor? :)

Numero De Operaciones Complejas/Seg Realizadas - 2209447.092 Ops/Seg
Number of Complex Operations / Seg Performed - 2209447.092 Ops / Sec
(That doesn't tell us much, since we have no information on how many operations it takes per iteration in your undisclosed implementation of a primality test.)

Milliseconds per LL or PRP iteration is the usual metric for performance on a specified exponent.

I think the reception would be a bit friendlier, if there was an English translation provided in the initial post of screen shots, for the various screen shots attached to it, and some more details about the algorithm. Efficiency is better. Better one person translate, than many.

Maybe it would help if the program output was in English, and put exponent, ms/iter, and run time per exponent together concisely. It's a bit hard to tell what goes with what with certainty on the screen shots. Double spacing all lines and repeating words uses screen space inefficiently. Is the one showing 2:22 for a run of 9941 and 9973, or all up to 10,000? Is T.T.C. time to complete? 2:17 for 86243 appears to be a modest 27% run time reduction from 3:07 in the first version. Another timing for 216091 might be interesting, to see if the exponential run time scaling has changed.

[Mysticial]

Originally Posted by kriesel

Preda knows what he's talking about. He introduced new software, gpuowl. He posted his source, many versions, on github. He was open about what algorithm was being employed. He took suggestions from users and potential users, implemented some of them, and the end result is the better for it. His software is accepted and in use because he was willing to post source and make requested modifications. (Millions of programmers use Github. Many use SourceForge.)

What he did not do, is limit source access to videos, documentation to screen shots, generally making it very inconvenient, cumbersome, slow, uninviting, for the potential end user, or require root or admin privilege.

Speaking of my own experience: I maintain the y-cruncher benchmark - which is closed sourced and binary only. The project is 10 years old now and the internet community seems to trust it. (at least nobody has accused me of trying to hide malware in it)

When I first launched it back in 2009, it was in the context of claiming a world record for the computation of the Euler-Mascheroni Constant. I posted a blog claiming the world record - along with how it was done (methods, algorithms, hardware, etc...). Then I included a download link to the raw binaries for people to play with.

Initially, I didn't try to advertise it. I knew right away that few people would run a naked binary from the internet. And I didn't particularly care about people using it either. I was just focused on breaking records - and the binaries for download were just a form of "proof" that I wasn't BS'ing.

But after a while, the program starting popping up by itself on various overclocking forums from other people who started using it as a benchmark. I would show up to some of those to offer support. Eventually, I started created a couple of "official" support threads on select forums.

From what I can see, people started trusting it pretty early on - despite being closed sourced and a Windows binary of all things! (at times it even required admin, and nobody questioned it) It was only until recent years when some AVs started reporting false positives did that topic finally come up.

[kriesel]

Originally Posted by Mysticial

When I first launched it back in 2009,
...

From what I can see, people started trusting it pretty early on - despite being closed sourced and a Windows binary of all things!

There is an element of "different times". I trust George's prime95 builds out of habit. The prime95 app was introduced in 1996, when, if I recall correctly, most companies were thinking about creating web sites or getting online at all, and the main vector for malware was the floppy disk that had been previously used on an infected system. With the increasing amount of information people store on their machines, the financial incentive for cybercrime went up. Increased skepticism is a rational response to increased potential damage and increased motivation for malware presenting increased risk.