Yep, but I have plans for the weekend. I'll try to get as much work done as possible before that.

Not really, no.

So, a new update. I had some problems implementing Barrett reduction, but now I only need a division algorithm and I can finally test and hopefully send the miner out.

New thread!

Currently the newest version is beta 2. If you are a donator with an old version, just redownload the miner from the link I sent you.

I am working on an OpenCL GPU miner for Primecoin. I fully intend to release it open-source and for free when it's ready.

Progress reports: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

When will it be released?

The closed beta is now running (since 2013-09-02), and the public release will be in a week or two. If you are a donator but didn't get the miner, whine to me and I'll send it to you!

Why should I trust you?

I am the creator of the GPU miner Reaper. I was the first person to show that GPU mining Litecoins (i.e. scrypt) is possible and also a lot more efficient than on the CPU. I've made my own GPU miners for many coins, including Bitcoin, Litecoin and Yacoin, although not all of them have been released to the public. The Primecoin GPU miner will be free and open-source.

I am also a university math student, and I have made my own implementation of the Elliptic Curve Digital Signature Algorithm used by Bitcoin (and all its forks) to make transactions. From scratch. (Yes, that includes my own bignum library.) I know a lot about algebra, prime numbers and such.

Why not CUDA?

CUDA is not cross-platform. It only works on Nvidia cards, but AMD cards are the ones with the best performance. OpenCL works on all GPUs.

What operating systems will the miner support?

Windows and Linux.

How fast is the miner going to be?

Currently it's not really faster than a CPU. It'll be optimized in time, though.

Official development thread at ppcointalk: http://www.ppcointalk.org/index.php?topic=508

Time for a new thread.

Alright, I've debugged the GPU prime tester, didn't finish yet though. For some reason the main part, the modular exponentiation, isn't working right but I'll figure it out. I can't give precise speed estimates, but right now it seems fast.

Okay, for the last few days I've had a bit of a real-life problem, the details of which are private. Let's just say that if it wasn't "that bad", I would never have had to pause the development of the miner. This was a break I absolutely needed. Please understand.

But now, I'm back on track. I'll try to post a proper update later today. I'll finally start testing that primality tester on the GPU.

For the last two days, I've had some real-life related stuff take priority, I haven't gotten much work done on the miner. I should have more time in a couple of days.

Okay the new GPU sieve is done, and using a single core of an HD6990 it's around 25x faster than a single CPU core of an x6 1055T. Next up: debugging and optimizing the primality tests.

I'm working on a major optimization of the GPU sieve. Should be ready today, and the sieve should run 5-10x faster.

Yep, that's what I ended up doing and I did find and fix the problem.

I'm now working on getting the Fermat primality test done on the GPU. Everything but the bignum modulo operation is already done, but there's a problem: The OpenCL compiler crashes when I try to compile with optimizations turned on. Without optimizations the code compiles, but that's obviously not a good solution. I will try updating my graphics drivers.

I haven't forgotten about testing on a slow CPU + overkill GPU rig, but I still have some problems using multiple GPUs.

Okay, the first GPU-enabled version is ready. Right now it only does the sieving on the GPU. Primality tests are still done on the CPU. The code hasn't been optimized at all yet; the performance is equivalent (+- a few %) on these setups:
1) Phenom x6 1055T, all 6 cores
2) Phenom x6 1055T, all 6 cores, plus a HD6990 using both cores

I don't think the current version is worth releasing. It still needs a bit of work, I want it to be twice as fast. I estimate there is room for a five-fold improvement, especially when I get the primality tests done on the GPU as well.

Tomorrow I will do tests using a slow CPU with lots of GPU power (Sempron 140 with 2x6990).

Alright, I have written the GPU code for the sieve. It compiles, but I haven't gotten to test it yet. I have to code the OpenCL interface first. I can't really use my existing code since all the other coins are "GPU does all the work", but the Primecoin miner is currently "CPU and GPU work together". The ultimate plan is, of course, to do everything on the GPU, but that's not gonna be the case in the first betas.

COG.F2 is getting a low score on BTCT. Is there a plan B?

Well, now I know who's not gonna get the miner first.

This graph doesn't mean anything. I didn't make it and the beta hasn't been released yet. Someone's just being funny.

It would be a bit weird for me to send a CPU miner when I've promised a GPU miner.

If you have something against it, please say so.

Yes, because Reaper uses the exact same code.

Why not release the CPU miner to the public? The donations were for a GPU miner, which will still be in a closed beta when it's ready.

Okay, progress report. As I said earlier, the CPU miner works now. However, my initial plan of doing the sieve work on the CPU and the expensive primality tests on the GPU didn't work. I've come up with another plan but it will take a couple days to implement.

The plan is to do the sieving on the GPU. The sieving is still something I don't fully understand, but I'm sure I can do it by studying the code meticulously.