[XPM]unofficial jhPrimeminer thread - page 43.

mumus

sr. member

Activity: 291

Merit: 250

Quote from: bitwarrior on July 30, 2013, 03:27:12 AM

pls do make one thanks.. i cant receive shares on the original client running on win 32 bit..or my cpu is too slow.. thanks in advance.

Done. https://dl.dropboxusercontent.com/u/1189851/jhPrimeminer-GMP-32bit.zip
But, again, on my quad core i7 I have about 30k PPS with the 64bit version and only 3.5K PPS with the 32bit version. This means it's about 8x slower.
If anybody else has suggestions are welcome to comment. I guess the problem is with the 32bit version of the mpir library.

bitwarrior

legendary

Activity: 1764

Merit: 1000

pls do make one thanks.. i cant receive shares on the original client running on win 32 bit..or my cpu is too slow.. thanks in advance.

mumus

sr. member

Activity: 291

Merit: 250

Quote

Anyone was able to make one for 32 bit? thanks

I can build a 32bit version if you want but the one I compiled last time is like 5 times more slower than the 64bit version, so I'm not sure it's faster than the official 32bit client.

bitwarrior

legendary

Activity: 1764

Merit: 1000

Quote from: mumus on July 29, 2013, 08:17:16 AM

Quote from: mumus on July 25, 2013, 11:34:43 PM

Quote from: Foamy on July 25, 2013, 07:53:19 PM

Quote from: mumus on July 25, 2013, 05:35:36 PM

If somebody want to give a try I made some modifications to the jhPrimeminer and I think at least 2 times faster than the current release.
Source https://github.com/hg5fm/jhPrimeminer
Compiled 64bit version with all the dependencies https://github.com/hg5fm/jhPrimeminer/blob/master/src/jhPrimeminer-GMP.zip

dude chrome hates it, it wouldnt even let me download it and when i tried using firefox my antivirus went off.

I just moved the binaries to dropbox. https://dl.dropboxusercontent.com/u/1189851/jhPrimeminer-GMP.zip
What do you mean on "antivirus went off"? I've just submitted the .exe file from the zip to some online antivirus scanners and found no problems. http://r.virscan.org/report/1dd9f52c4f970778a2a156de36655e03.html

Added even more optimizations and Share Value information based on the newest change in the pool.
https://dl.dropboxusercontent.com/u/1189851/jhPrimeminer-GMP.zip (only 64bit, the 32bit is too slow)
More details and the source on https://github.com/hg5fm/jhPrimeminer

Anyone was able to make one for 32 bit? thanks

gatra

hero member

Activity: 583

Merit: 505

CTO @ Flixxo, Riecoin dev

Quote from: jh00 on July 29, 2013, 08:35:45 PM

Quote from: gatra on July 29, 2013, 08:01:51 PM

I have an optimization that I wanted to apply, so that's why I wanted sources.
I'll just give it to you so you can include it in next versions: in prime2.cpp
this:

Code:

nSolvedMultiplier = ((fixedInverseU64) * (pU64 + 1ULL)) % pU64;

is the same as this:

Code:

nSolvedMultiplier = fixedInverseU64;

because pU64 is obviously 0 modulo pU64, and fixedInverseU64 is already less than pU64

Ahh true. Thanks, I didn't think this way. It will be included in the next release.

It works in the other case too: if nBiTwinSeq is odd, you can do this:

Code:

nSolvedMultiplier = pU64 - fixedInverseU64;

because what we want is ( (-1) * fixedInverseU64 ) modulo pU64

jh00

newbie

Activity: 39

Merit: 0

Quote from: gatra on July 29, 2013, 08:01:51 PM

I have an optimization that I wanted to apply, so that's why I wanted sources.
I'll just give it to you so you can include it in next versions: in prime2.cpp
this:

Code:

nSolvedMultiplier = ((fixedInverseU64) * (pU64 + 1ULL)) % pU64;

is the same as this:

Code:

nSolvedMultiplier = fixedInverseU64;

because pU64 is obviously 0 modulo pU64, and fixedInverseU64 is already less than pU64

Ahh true. Thanks, I didn't think this way. It will be included in the next release.

spoid

hero member

Activity: 532

Merit: 500

bearded, drunk, fat, naked

this is confusing. so do I use what mumus posted or the latest version from the ypool website? i'm trying to find out by benchmarking, but they changed the complete share system while I was at it... D: the mumus version finds way more shares and shows more info, so I'm going with that one for now.

Also, Kouye, when compiling yourself, do you have to change lots of parameters with specific knowledge about your CPU, or do programs simply run better on the machine/architecture they have been compiled on and you just follow some standard windows compiling tutorial?

bcp19

hero member

Activity: 532

Merit: 500

Quote from: Kouye on July 29, 2013, 08:05:34 PM

No, I'm running the hp8 on a very low end loaned virtual machine.

7e64d60c73d284a0d9095765e14a8b528f5866fc9ab6b05ba98964d9fd3f30eb
5c29b461e31980239373c3e27d0189ae032163e593d454a34fa5ad3e5ef2ea84

Those are the txid for the 2 blocks I mined last week.

Just stating I feel much more confortable running source code I compiled myself and has proven quite efficient, given my cpu power, than running precomîled binaries working for a pool, with a x10 speed increase. Cheesy

You can get the source code for the pool miner as well.

Kouye

sr. member

Activity: 336

Merit: 250

Cuddling, censored, unicorn-shaped troll.

No, I'm running the hp8 on a very low end loaned virtual machine.

7e64d60c73d284a0d9095765e14a8b528f5866fc9ab6b05ba98964d9fd3f30eb
5c29b461e31980239373c3e27d0189ae032163e593d454a34fa5ad3e5ef2ea84

Those are the txid for the 2 blocks I mined last week.

Just stating I feel much more confortable running source code I compiled myself and has proven quite efficient, given my cpu power, than running precomîled binaries working for a pool, with a x10 speed increase. Cheesy

gatra

hero member

Activity: 583

Merit: 505

CTO @ Flixxo, Riecoin dev

Quote from: jh00 on July 29, 2013, 07:48:08 PM

Quote from: gatra on July 29, 2013, 07:03:55 PM

and again... new binaries, no updated sources...

Can't you just wait even a few minutes? I usually release the binaries first, upload them on ypool.net, maybe try some other stuff out and then push the source to the github.

hehe, ok, that was fast, I'm sorry
I have an optimization that I wanted to apply, so that's why I wanted sources.
I'll just give it to you so you can include it in next versions: in prime2.cpp
this:

Code:

nSolvedMultiplier = ((fixedInverseU64) * (pU64 + 1ULL)) % pU64;

is the same as this:

Code:

nSolvedMultiplier = fixedInverseU64;

because pU64 is obviously 0 modulo pU64, and fixedInverseU64 is already less than pU64

bitwarrior

legendary

Activity: 1764

Merit: 1000

Quote from: forsetifox on July 29, 2013, 07:16:16 PM

Quote from: Kouye on July 29, 2013, 07:06:52 PM

In the meantime, I'm solo-mining with 300pps and I got 2 blocks last week.

Have fun!

Isn't this restricted to using the pool? I've tried several versions and they crash if you try to solo mine.

How do you solo mine using this jhprimeminer?

jh00

newbie

Activity: 39

Merit: 0

Quote from: gatra on July 29, 2013, 07:03:55 PM

and again... new binaries, no updated sources...

Can't you just wait even a few minutes? I usually release the binaries first, upload them on ypool.net, maybe try some other stuff out and then push the source to the github.

Quote from: forsetifox on July 29, 2013, 07:16:16 PM

Isn't this restricted to using the pool? I've tried several versions and they crash if you try to solo mine.

Currently yes, only pool mining. Solo mining support is planned.

forsetifox

sr. member

Activity: 266

Merit: 250

Quote from: Kouye on July 29, 2013, 07:06:52 PM

In the meantime, I'm solo-mining with 300pps and I got 2 blocks last week.

Have fun!

Isn't this restricted to using the pool? I've tried several versions and they crash if you try to solo mine.

Kouye

sr. member

Activity: 336

Merit: 250

Cuddling, censored, unicorn-shaped troll.

In the meantime, I'm solo-mining with 300pps and I got 2 blocks last week.

Have fun!

gatra

hero member

Activity: 583

Merit: 505

CTO @ Flixxo, Riecoin dev

Quote from: bcp19 on July 29, 2013, 06:54:46 PM

ypool reposts the release of version .35

and again... new binaries, no updated sources...

bcp19

hero member

Activity: 532

Merit: 500

ypool reposts the release of version .35

mumus

sr. member

Activity: 291

Merit: 250

The most recent version (v3) includes the following:

Add -d (SievePercentage) & -primes (SievePrimeLimit) command line parameters that can be used for more tweaking.
SievePercentage ranges from 1 - 100, recommended is around 10 (default is Cool

. It limits how many base primes are used to filter out candidate multipliers in the sieve

The idea of SievePrimeLimit was taken from the latest official jhPrimeMiner source. The default value is equal to MaxSieveSize which I still recommend.

Changed the sampling time for the multiplier auto adjustment to 10 sec.

Download link for the 64bit binaries: https://www.dropbox.com/s/tauj21tvm06xk7y/jhPrimeminer-GMP-v3.zip
Source: https://github.com/hg5fm/jhPrimeminer

SievePercentage is similar to cabin's '-d' (SieveDivisor) parameter but it's using percentage instead of division of the sieve size. (ex: '-d 100' is equivalent to '-d 1' in cabins build, 50 = 2, 25=4 and so on)

mumus

sr. member

Activity: 291

Merit: 250

Quote from: cabin on July 29, 2013, 12:47:44 PM

I've tried some different settings that worked well for me:

jhPrimeMiner.exe -d 5 -h 7

duo core 2.9GHz
Valid share found![ 55 / 55 ] Mon Jul 29 12:45:48 2013
Val/h: 3.17 Shares/h: 64.2 - 5ch/h: 210.1 - PPS: 1415 - Len: 5.80/7.68 5.13

But I'll be away and won't be doing any more builds for a while so if the -h an -d params prove useful maybe Mumus can add them into his builds. They are both used in the sieve weave at the bottom here:

https://www.dropbox.com/s/f39cxsus0u6xary/prime2.cpp
nSieveDepth and nSieveDivisor

also there is a line:
//printf("U");

that would print out when the sieve needs to be torn down and rebuilt.. I suspect this sucks tons of time and I've been trying to minimize it with various combinations of settings.

I'll review the code and I'll include what I think it's important. I'm planning to do also an auto adjustment on the SievePercentage (you call it nSieveDivisor).

cabin

sr. member

Activity: 604

Merit: 250

Last Update July 30th: https://www.dropbox.com/s/szayx15skhevv7x/jhPrimeminer.zip

I've tried some different settings that worked well for me:

jhPrimeMiner.exe -h 8 -s 1500000 -o http://ypool.net:10034 -u cabin.xpm_1 -p x

---- New Block: 89831 - Diff: 9.25495 / 5.800000 PrimorialMultiplier: 41
---- Total/Valid shares: [ 236 / 240 ] 6CH_count: 68 - 7CH_count: 7 - Max: 8.42722
═══════════════════════════════════════════════════════════════
Val/h 7.69 - SPH: 74.62 - 5ch/h: 275 - PPS: 1406 Best: 5.09 Avg: 5.24

But I'll be away and won't be doing any more builds for a while so if the -h proves useful maybe Mumus can add it into his builds. nSieveDepth is used in the sieve weave at the bottom of prime.cpp:
https://www.dropbox.com/s/32tvjs9xays96ir/primecoinMiner_diffMumu.zip

also there is a line:
//printf("U");

that would print out when the sieve needs to be torn down and rebuilt.. I suspect this sucks tons of time and I've been trying to minimize it with various combinations of settings.

pyromaniac

hero member

Activity: 639

Merit: 500

Quote from: kuusj98 on July 29, 2013, 11:54:19 AM

EDIT 3: SHE's ALIVE! ~32.000 PPS, is that alright for my CPU?

Absolutely! I have around 55 k on my i7 2600k. So it's definitely normal result.

Topic: [XPM]unofficial jhPrimeminer thread - page 43. (Read 180201 times)