[XPM] CUDA enabled qt client miner for primecoins. Source code inside. WIP

[K1773R]

Please check that you're using the latest SDK. I also encountered memory problems with cuda 5.0 and I'm using 5.5 now which works for me.

ACK, will do later and report back

[Entz]

Just got this compiled  (Talk about a mess, when my cuda sdk was installed the paths were completely different then they should of /nvidia-304 vs /nvidia-current etc, then some fun Qt conflicts).

Anyone have a working node for testnet they can post? Not having any luck connecting.

[bcp19]

Please check that you're using the latest SDK. I also encountered memory problems with cuda 5.0 and I'm using 5.5 now which works for me.

Just curious, have you looked at the Mfaktc source code at all?  While it is used for trial factoring Mersenne Primes, which may not be helpful, the writer did get it to sieve completely on the GPU, which may.

[Entz]

Having the same problem as K1773R  (GTX670 using CUDA 5.5 and the driver it includes). Tried it on mainnet as I still cannot connect to testnet for some reason.

Code:

Have 101 candidates after main loop
Cuda start!
{... some block messages i.e. getblocks -1 to blah, accept etc}
Have -1 candidates after main loop
Cuda+host test round finished with -1 candidates (0 host chain tests)
Cuda error: cudaMemcpy: cudaMemcpyDeviceToHost, unspecified launch failure
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow
ERROR: PrimecoinMiner() : primorial minimum overflow

from GDB

Code:

[0] start! 
sizeof(struct) = 400
mpz_print:mpz_capacity: 0 
[0] string candidate is 
[0] N is: mpz_capacity: 30 ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
[0] E is: mpz_capacity: 30 fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffe

Edit: This may just be PEBKAC / RTFM issue on my part. Just saw your note about running with 1 cpu only.

Still crashed, managed to get a few rounds. Used to crash right away

Code:

2013-07-25 19:51:56 primemeter         0 prime/h    498885 test/h         0 5-chains/h
2013-07-25 19:52:56 primemeter         0 prime/h   8404040 test/h         0 5-chains/h
2013-07-25 19:53:56 primemeter         0 prime/h   4184750 test/h         0 5-chains/h

[ReCat]

Fascinating. This CUDA miner is already vaguely functional now? Now that's some community effort. I wonder what will be the eventual result of this. Will fast CPU's and GPU's working together be the new mining rigs?

[liteuser]

I've updated to cuda-5.5 (and driver 319.21)

Running with cuda-gdb I get the following error:

Code:

Have 2400 candidates after main loop
Cuda start!
[New Thread 0x7fffacc38700 (LWP 14248)]
[Context Create of context 0x7fff700234f0 on Device 0]
[Launch of CUDA Kernel 0 (runPrimeCandidateSearch<<<(25,1,1),(192,1,1)>>>) on Device 0]

Program received signal CUDA_EXCEPTION_10, Device Illegal Address.
[Switching focus to CUDA kernel 0, grid 1, block (15,0,0), thread (0,0,0), device 0, sm 3, warp 0, lane 0]
0x00007fff7091b760 in long_multiplication(unsigned int * @generic, unsigned int * @generic, unsigned int * @generic, unsigned int, unsigned int) (
    product=0x3fff6b4, op1=0x3fff734, op2=0x3fff634, num_digits=17, 
    prod_capacity=1073741824)
    at primecoin/src/cuda/digit.h:406
406	    product[i] = 0;

[primedigger]

Please check that you're using the latest SDK. I also encountered memory problems with cuda 5.0 and I'm using 5.5 now which works for me.

Just curious, have you looked at the Mfaktc source code at all?  While it is used for trial factoring Mersenne Primes, which may not be helpful, the writer did get it to sieve completely on the GPU, which may.

I looked into it, yes. Code is not very understandable though...

[hasle2]

Is this project still active or has it been abandoned?

[ReCat]

It's abandoned. Lol. Probably everyone figured out that this is too difficult. Heck even mlmrt was having trouble.

[Kouye]

It's abandoned. Lol. Probably everyone figured out that this is too difficult. Heck even mlmrt was having trouble.

LIES! He's managed to have the same efficiency as an AMD multi-core.

With an AMD multi-core + a HD6990.

[bcp19]

It's abandoned. Lol. Probably everyone figured out that this is too difficult. Heck even mlmrt was having trouble.

LIES! He's managed to have the same efficiency as an AMD multi-core.

With an AMD multi-core + a HD6990.

WOW, we can spend ~194 watts running an AMD multi-core or ~525 watts running the AMD and a GPU and get the same results!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
I WANT I WANT I WANT I WANT!

[maco]

has anyone tested this yet? is it working?

[primedigger]

I'm still on it - with a different idea. As it turns out, doing Fermat tests on the GPU is not a no brainer and getting that fast requires too much effort for now, so I'll try to port something else to the GPU.

I'm still sure a GPU miner is possible, but right now I would say it's a lot harder than for the other coins. The other OpenCL miner project is (amusingly!) also having problems.

[refer_2_me]

I'm still on it - with a different idea. As it turns out, doing Fermat tests on the GPU is not a no brainer and getting that fast requires too much effort for now, so I'll try to port something else to the GPU.

I'm still sure a GPU miner is possible, but right now I would say it's a lot harder than for the other coins. The other OpenCL miner project is (amusingly!) also having problems.

As i'm sure you are already aware, mlmrt ported the sieve to the GPU. Is that what you are going after?

[ReCat]

It's abandoned. Lol. Probably everyone figured out that this is too difficult. Heck even mlmrt was having trouble.

LIES! He's managed to have the same efficiency as an AMD multi-core.

With an AMD multi-core + a HD6990.

Wait... that's better?

[jaakkop]

What's the progress so far?

[Lauda]

It's abandoned. Lol. Probably everyone figured out that this is too difficult. Heck even mlmrt was having trouble.

LIES! He's managed to have the same efficiency as an AMD multi-core.

With an AMD multi-core + a HD6990.

Wait... that's better?

It's not..

[ReCat]

Forget about it guys, this miner was never gonna happen. Obviously the people who started and promoted this thread had no idea what they were getting into.

The reaper guy's miner is probably the only GPU miner we will EVER be seeing.

[refer_2_me]

Forget about it guys, this miner was never gonna happen. Obviously the people who started and promoted this thread had no idea what they were getting into.

The reaper guy's miner is probably the only GPU miner we will EVER be seeing.

So it would seem, sadly. I hope once mrtlt's is open sourced we can get some real community development going for the good of the coin. Hopefully by that point, I will have finish the crunch at work and I can try to dive in as well.

[primedigger]

I was away for the past week and will look into it again this week. Yes, it's just a hobby project and it got bigger than I expected. Currently, I'm the only one working on this, so if someone wants to chip in and help (programming), send me a PM.

Status:

I will push my lastest changes soon, I have updated my code basis to hp-9 and I implemented a fast big num small prime trial division for the GPU. Depending on the settings, this can filter out 10-90% of all candidates. The CPU than computes the fermat tests on the remaining candidates. I was under the impression that the sieve would already filter out all chains versus small primes, but apparently the high performance client still filters out some candidates with trial divisions and does this before doing fermat tests.

If a fast fermat test for the GPU surfaces, than filtering+fermat tests could be chained directly on the GPU to give a better speed up.

To clarify: I didn't push my changes because I still have a silly bug somewhere, so that apparently not all prime divisors are found. But doing more prime division tests than what the high performance client does by default yields already better speed ups directly on the CPU.