It's been over a month since our last update, but I now have some good news. I have made some improvements to the GPU code and am ready to start deploying the new GPU apps. I will start with the AMD OpenCL version for Linux. This will be a beta version. I have had a hell of a time with the AMD implementation of openCL, and this app still doesn't work on my Fedora system, and I believe strongly it's due to the graphics driver. But I have had the help of a volunteer named Wiktor and it runs fine for him (I believe he runs Ubuntu). Please keep in mind that AMD officially only supports RHEL and Ubuntu, so I will be interested to hear if this app works for anyone with an "unsupported" linux distro like myself. I also have openCL Windows apps that were cross compiled using mingW. I have no means of testing these, so I am not ready to deploy them just yet. But if anyone would like to take them for a spin offline, please let me know, and I can send them to you.

By numberfields@home at 2019-05-15

Subfield 5 is now on it's final batch (16x12). This is the first batch of WUs generated since the app was optimized. As a result, run times will be going up. These WUs average 2 hour run times on my AMD Ryzen 2990WX. Note there are 1.6 million of them; with the unoptimized app it would have required 16 million, so this is a huge improvement. I will be mixing in some subfield 6 WUs with this batch. These were generated a while ago for the unoptimized app, so will run very quickly by comparison. If I were to run them independently, it might put too much of a strain on the server, so I figured this is a good way to get them done (without having to re-generate them which can be tedious).

By numberfields@home at 2019-05-04

So there have been some new developments over the last week. It's both good and bad. First of all, some history. The reason I waited so long to develop a GPU app is because the calculation was heavily dependent on multi-precision libraries (gmp) and number theoretic libraries (pari/gp). Both of these use dynamically allocated memory which is a big no-no in GPUs. I found a multi-precision library online that I could use by hard coding the precision to the maximum required (about 750 bits), thereby removing the dependence on memory allocations. The next piece of the puzzle was to code up a polynomial discriminant function. After doing this, I could finally compile a kernel for the GPU. That is the history for the current GPU app. It is about 20 to 30 times faster than the current cpu version (depends on WU and cpu/gpu speeds). But then I got thinking... my GPU polynomial discriminant algorithm is different from the one in the PARI library (theirs works for any degree and mine is specialized to degree 10). So to do a true apples-to-apples comparison, I replaced the PARI algorithm with mine in the cpu version of the code. I was shocked by what I found... the cpu version was now about 10x faster than it used to be. I never thought I was capable of writing an algorithm that would be 10x faster than a well established library function. WTF? Now I'm kicking myself in the butt for not having done this sooner! This brings mixed emotions. On one side, it is great that I now have a cpu version that is 10x faster. But it also means that my GPU code is total crap. Wit...

By numberfields@home at 2019-04-05