How to test your cell phone warranty

[Damian]

I have been verifying some mersenne primes on my cell phones using LLMidlet j2me program.

I tested it on my two phones: an old Siemens C66, and a new Blackberry 8700.

Surprisingly it turned out the siemens was 1,84 times faster than the blackberry (almost twice!)

Here are some benchmarks on the 5 Robinson's mersennes

Mersenne | Blackberry 8700 / Siemens C66
M13 | 1 sec / 1 sec = 1
M14 | 18 sec / 13 sec = 1,37
M15 | 4 min / 2 min 13 sec = 1,47
M16 | 19 min 44 sec / 10 min 35 sec = 1,86
M17 | 21 min 51 sec / 11 min 52 sec = 1,84

The midlet jad can be downloaded here LLMidlet.jad
And the midlet jar can be downloaded here LLMidlet.jar

Anyone interested on adding some benchmarks to this list?

[Damian]

Also, I forgot to say the blackberry almost eat all of its battery during this test, and the siemens battery was almost not affected.

The record I hold is M22 on the blackberry and it lasted 1 day 5 hours 12 min 15 sec.
(this was before I knew the siemens was faster, and the phone was connected charging battery all that time)

[henryzz]

i have just started testing M23 to beat ur record
might take a while

[Damian]

Originally Posted by henryzz

i have just started testing M23 to beat ur record
might take a while

At what percentage it is now?
What phone is it?

Someone changed the title of this post XD
It was "How to test mersenne primes on your cell phone", not how to test the phone warranty. Of course I can not guarantee it wont burn out your phone.

But, who want to waste those idle cpu phone cycles?

[henryzz]

after 24 hours it still didnt have a progress marker so i stopped it as i was going to need the phone before it finished at that rate
i think i may have pushed something that had stopped it
i will do the benchmarks tonight

[henryzz]

a couple of days ago i found this program on my mobile and i tested M8191 with it in ~90918 seconds which is ~25.25 hours
it was a Sony Ericsson W810i
Mersenne | Sony Ericsson W810i
M13 | 10 sec
M14 | 15 sec
M15 | 2 min 10 sec
M16 | 35 min 2 sec
M17 | 101 min 48 sec

[10metreh]

Originally Posted by henryzz

a couple of days ago i found this program on my mobile and i tested M8191 with it in ~90918 seconds which is ~25.25 hours
it was a Sony Ericsson W810i
Mersenne | Sony Ericsson W810i
M13 | 10 sec
M14 | 15 sec
M15 | 2 min 10 sec
M16 | 35 min 2 sec
M17 | 101 min 48 sec

I normally consider it standard practice not to use M for both of its meanings (Mersenne number and Mersenne prime) in the same post, so M13 should have been M521, M14 should have been M607 etc. No offence intended.

[henryzz]

Originally Posted by 10metreh

I normally consider it standard practice not to use M for both of its meanings (Mersenne number and Mersenne prime) in the same post, so M13 should have been M521, M14 should have been M607 etc. No offence intended.

yes
it is something that could be rather confusing except that most of the time it is obvious that M14 means the 2^607-1 because 14 is a low number
to my knowledge no one has ever gotten confused on this forum with people using both interchangeably
i agree though that both forms probably shouldn't be used in the same post

[retina]

I imagine anything <M61 could be used without generating much confusion. I think most people here accept that small numbers are the prime count, and larger numbers are the exponent.

[starrynte]

this is inspiring me to try to write a midlet that also does ll tests...or maybe trial factors...
UPDATE: LL tests are too complicated (unless you don't use FFT's) thus I will attempt to write a trial factoring midlet

[henryzz]

Originally Posted by starrynte

this is inspiring me to try to write a midlet that also does ll tests...or maybe trial factors...
UPDATE: LL tests are too complicated (unless you don't use FFT's) thus I will attempt to write a trial factoring midlet

i think that there is a multiprecision integer for java
i dont know weather that works for mobiles
something that does llr tests or prp tests might be good