Pages:
Author

Topic: [XPM] Working on a GPU miner for Primecoin, new thread :) - page 15. (Read 166591 times)

newbie
Activity: 23
Merit: 0
I've made the primality tester 4-5x faster, but I haven't managed to fix the "fractional assert" problem. Should I just release the faster version (many people won't be able to use it) or try to fix the bug first (will take time)?
If this is of any help, looking at the console as all the data flies by, my setup (i7 chip, win8, 64bit) reels out all the data, starting with 0 fermats/sec and 786 gandalfs/s, then as time passes, the gandalfs/s progressively drops and if the app doesnt crash earlier each time the console reports the gandalfs/s it shows a smaller and smaller number, reducing to 60 gandalfs/s then it usually gives up and passes onto windows error reporting. It always only shows 0 fermats/s - ? what is this meant to be?

I had this same issue with the gandalfs dropping and I managed to keep it at a regular average number by playing around with the "worksize" parameter in the primecoin.conf , got the best result with 64

I can confirm worksize of 64 was optimal for me too.
newbie
Activity: 20
Merit: 0
I've made the primality tester 4-5x faster, but I haven't managed to fix the "fractional assert" problem. Should I just release the faster version (many people won't be able to use it) or try to fix the bug first (will take time)?
If this is of any help, looking at the console as all the data flies by, my setup (i7 chip, win8, 64bit) reels out all the data, starting with 0 fermats/sec and 786 gandalfs/s, then as time passes, the gandalfs/s progressively drops and if the app doesnt crash earlier each time the console reports the gandalfs/s it shows a smaller and smaller number, reducing to 60 gandalfs/s then it usually gives up and passes onto windows error reporting. It always only shows 0 fermats/s - ? what is this meant to be?

I had this same issue with the gandalfs dropping and I managed to keep it at a regular average number by playing around with the "worksize" parameter in the primecoin.conf , got the best result with 64
member
Activity: 104
Merit: 10
What GPU?
sr. member
Activity: 771
Merit: 258
Trident Protocol | Simple «buy-hold-earn» system!
I've made the primality tester 4-5x faster, but I haven't managed to fix the "fractional assert" problem. Should I just release the faster version (many people won't be able to use it) or try to fix the bug first (will take time)?
If this is of any help, looking at the console as all the data flies by, my setup (i7 chip, win8, 64bit) reels out all the data, starting with 0 fermats/sec and 786 gandalfs/s, then as time passes, the gandalfs/s progressively drops and if the app doesnt crash earlier each time the console reports the gandalfs/s it shows a smaller and smaller number, reducing to 60 gandalfs/s then it usually gives up and passes onto windows error reporting. It always only shows 0 fermats/s - ? what is this meant to be?
sr. member
Activity: 771
Merit: 258
Trident Protocol | Simple «buy-hold-earn» system!
I've made the primality tester 4-5x faster, but I haven't managed to fix the "fractional assert" problem. Should I just release the faster version (many people won't be able to use it) or try to fix the bug first (will take time)?
Great work! Might be fairer to fix the bug, I am one of the ones who hasnt been able to use the miner yet though.
member
Activity: 104
Merit: 10
I've made the primality tester 4-5x faster, but I haven't managed to fix the "fractional assert" problem. Should I just release the faster version (many people won't be able to use it) or try to fix the bug first (will take time)?
newbie
Activity: 23
Merit: 0
The TOTAL counts aren't comparable between use_gpu_fermat_test true/false, should look at 2-chains. I'll remove the TOTAL count in the next version.
Good to know, thanks! Looking at 2/3 chains then the gpu_fermat_test is faster by ~ 1.4x (I do have a weak CPU in there).  Is the candidate difference a red herring (1/9 as many) or just the way it is reported?
I'm just not seeing stats anything like this.

With the fermat test done on the GPU I get...
0 fermats/s, 6.73269 gandalfs/s.
70.2502k TOTAL     1.79828k 2-chains     39.093 3-chains     39.093 4-chains

With the fermat done on the CPU I get ...
60338.8 fermats/s, 3523.21 gandalfs/s.
350.383M TOTAL     1.01852M 2-chains     64.7127k 3-chains     3.58685k 4-chains     298.904 5-chains

Ignoring the total, the fermat test on the GPU gives 1000x fewer 2-chains,  and 100x fewer 4-chains.  This is with 2 x 7950, a AMD FX(tm)-4130 Quad-Core Processor, and ubunutu 13.

Am I missing something somewhere?

I got similar results enabling/disabling the fermat test on the GPU.
sr. member
Activity: 364
Merit: 253
Any successful blocks?
hero member
Activity: 639
Merit: 500
Please help me compile a file libblkmaker, so that I could use address different than mu1DxMiRUgzjNrVyVeHeRhqDm3XVK12SK2 in my config. I'm not a programmer.
edit base58.c in the libblkmaker dir and add these middle lines:

      case   0:  // Bitcoin pubkey hash
      case  23:  // Primecoin pubkey hash
      case 111:  // Testnet pubkey hash

and

      case   5:  // Bitcoin script hash
      case  83:  // Primecoin script hash
      case 196:  // Testnet script hash

So, look up those bitcoin/testnet lines and add the primecoin one. Then compile Smiley (don't forget autogen.sh)
You can also just redownload the library sources (from https://dl.dropboxusercontent.com/u/55025350/bitcoin-libblkmaker.zip). I have updated the source.
Thanks! Now next problem is GPU is not used. How can I fix it?

Quote
Share thread started
GeneratePrimeTable() : prime table [1, 262144] generated with 23000 primes
Available CPU mining algorithms: hp7
Using default: hp7
Creating 2 CPU threads.
1...Warning: can't set thread priority
2...done
List of platforms:
        0       AMD Accelerated Parallel Processing
Using platform number 0
2013-09-05 19:08:29 Error: Error getting OpenCL device IDs
member
Activity: 104
Merit: 10
Please help me compile a file libblkmaker, so that I could use address different than mu1DxMiRUgzjNrVyVeHeRhqDm3XVK12SK2 in my config. I'm not a programmer.
edit base58.c in the libblkmaker dir and add these middle lines:

      case   0:  // Bitcoin pubkey hash
      case  23:  // Primecoin pubkey hash
      case 111:  // Testnet pubkey hash

and

      case   5:  // Bitcoin script hash
      case  83:  // Primecoin script hash
      case 196:  // Testnet script hash

So, look up those bitcoin/testnet lines and add the primecoin one. Then compile Smiley (don't forget autogen.sh)
You can also just redownload the library sources (from https://dl.dropboxusercontent.com/u/55025350/bitcoin-libblkmaker.zip). I have updated the source.
sr. member
Activity: 415
Merit: 250
Please help me compile a file libblkmaker, so that I could use address different than mu1DxMiRUgzjNrVyVeHeRhqDm3XVK12SK2 in my config. I'm not a programmer.
edit base58.c in the libblkmaker dir and add these middle lines:

      case   0:  // Bitcoin pubkey hash
      case  23:  // Primecoin pubkey hash
      case 111:  // Testnet pubkey hash

and

      case   5:  // Bitcoin script hash
      case  83:  // Primecoin script hash
      case 196:  // Testnet script hash

So, look up those bitcoin/testnet lines and add the primecoin one. Then compile Smiley (don't forget autogen.sh)
hero member
Activity: 639
Merit: 500
Please help me compile a file libblkmaker, so that I could use address different than mu1DxMiRUgzjNrVyVeHeRhqDm3XVK12SK2 in my config. I'm not a programmer.
sr. member
Activity: 266
Merit: 250
I just finished to read all the thread, I'm so happy that I didn't "donate".

I'm better without a GPU miner since CPU it's what I'm most using right now, however XPM needs the GPU miner to gain popularity otherwise it will remain as the botnet coin.

cheers!
newbie
Activity: 23
Merit: 0
Can this be used to mine in ypool?

I don't think so.  Ypool uses XPT exclusively, IIRC.  I think primecoin-reaper uses getwork, so someone would need to port XPT to it.
hero member
Activity: 812
Merit: 1000
It seems like memory corruption, the most annoying kind of bug. Might take some time to fix, I haven't been able to reproduce it on my computer.
This is a serious problem that may not have a fix.  I have a GPU that I still have running on SHA256 because it starts throwing massive errors if I try to use the same intensity on scrypt coins.

I don't think this is a GPU error, but rather an error in the CPU code somewhere.

Agreed.happens at stock clocks too and it's too much of a coincidence that all our cards are faulty.
member
Activity: 104
Merit: 10
It seems like memory corruption, the most annoying kind of bug. Might take some time to fix, I haven't been able to reproduce it on my computer.
This is a serious problem that may not have a fix.  I have a GPU that I still have running on SHA256 because it starts throwing massive errors if I try to use the same intensity on scrypt coins.

I don't think this is a GPU error, but rather an error in the CPU code somewhere.
member
Activity: 67
Merit: 10
I have a hd7970 and hd7950.

1) How old is your GPU? - 3 months old
2) Have you ever OC'd or declocked it? - yes to recommended scrypt overclocks.
3) What intensity settings have you run it at on other coins and for how long? - i13 (cgminer) for most scrypt, i20 for copperlark (reaper miner). I dont mine full time, probably 3-4 days in one hit has been my longest non stop mine.
4) Have you ever over/under volted it? No.

Crashes with Prime Reaper on any sievesize setting after so long, sometime instant and sometime after a few mins depending on the setting. If I remove the sievesize setting and sievesize percentage entirely from the config file reaper sometimes runs, but always produces 0 fermats, and a value for gandalfs that declines over time. The longest it has run without crashing in this mode is around 2 hrs, eventually it crashed too. I have pretty much given up attempting to get it to work now.
donator
Activity: 1064
Merit: 1000
It seems like memory corruption, the most annoying kind of bug. Might take some time to fix, I haven't been able to reproduce it on my computer.
This is a serious problem that may not have a fix.  I have a GPU that I still have running on SHA256 because it starts throwing massive errors if I try to use the same intensity on scrypt coins.  300MH vs 174KH running 16 vs 11 intensity.  Many of these GPUs have been running on hyperdrive for months on end, but until now they have not had to deal with the intensive memory load that your miner is throwing at them.  I'd bet your program hits GPU the same way P95 hits CPU... it stresses them to the point of failure not seen in normal/somewhat heavy operation.  With all the underclocking/overclocking/undervolting/overvolting I'd bet there are a lot of cards that appear to work fine yet are hiding a major problem. 

Memtest is a fairly intensive memory testing program, but it's not as intensive as P95.  I've seen many posts about people saying the P95 program was faulty cause it crashed their computer yet memtest said no problems.  Not once have those claims been proven out though, it's always ended up that P95 stressed something harder than memtest and found a problem that needed fixing.

I vaguely remember someone on the Mersenne forums having written a crude GPU program to simulate what memtest does... maybe you need to have people start looking into that.

Until then, I'd suggest you start having people answer a few basic questions to start the troubleshooting...

1) How old is your GPU?
2) Have you ever OC'd or declocked it?
3) What intensity settings have you run it at on other coins and for how long?
4) Have you ever over/under volted it?

I'd be willing to bet the answers to these 4 questions will start a pattern for those who have it working and those that don't.
1, About one to two months old.
2, yes slightly both ways, but it runs so hot that I most often let it run on default.
3, I usually use 10-15, I haven't been running any mining application 24/7, just a couple of hours to try the cards out.
4, Yes but not through any manual process, I have only used the built in +-20% voltage.
//DeaDTerra
hero member
Activity: 532
Merit: 500
It seems like memory corruption, the most annoying kind of bug. Might take some time to fix, I haven't been able to reproduce it on my computer.
This is a serious problem that may not have a fix.  I have a GPU that I still have running on SHA256 because it starts throwing massive errors if I try to use the same intensity on scrypt coins.  300MH vs 174KH running 16 vs 11 intensity.  Many of these GPUs have been running on hyperdrive for months on end, but until now they have not had to deal with the intensive memory load that your miner is throwing at them.  I'd bet your program hits GPU the same way P95 hits CPU... it stresses them to the point of failure not seen in normal/somewhat heavy operation.  With all the underclocking/overclocking/undervolting/overvolting I'd bet there are a lot of cards that appear to work fine yet are hiding a major problem. 

Memtest is a fairly intensive memory testing program, but it's not as intensive as P95.  I've seen many posts about people saying the P95 program was faulty cause it crashed their computer yet memtest said no problems.  Not once have those claims been proven out though, it's always ended up that P95 stressed something harder than memtest and found a problem that needed fixing.

I vaguely remember someone on the Mersenne forums having written a crude GPU program to simulate what memtest does... maybe you need to have people start looking into that.

Until then, I'd suggest you start having people answer a few basic questions to start the troubleshooting...

1) How old is your GPU?
2) Have you ever OC'd or declocked it?
3) What intensity settings have you run it at on other coins and for how long?
4) Have you ever over/under volted it?

I'd be willing to bet the answers to these 4 questions will start a pattern for those who have it working and those that don't.
member
Activity: 100
Merit: 10
It seems like memory corruption, the most annoying kind of bug. Might take some time to fix, I haven't been able to reproduce it on my computer.
I get the same bug as DeaDTerra. Let me know if I can help test, or if you need a pastebin dump to figure it out. I am trying all sorts of settings but still get the same issue, running on Win8, 64bit. I get the same issue on three other PCs. The most it runs for is a minute or so.

I also have problems with reaper crashing,  win 7 x64,  played with the settings but no joy
Pages:
Jump to: