Savaş KOÇ

Hi, thanks for opening this issue to here. I think there is something wrong in OpenCl. I am sending kernel and host code in C language. It STILL gives wrong...

It's default implementation in macOS. I'm using 10.12 (16A320) MacBook Pro (Retina, 13-inch, Late 2013). It has i5 4258U CPU and Iris 5100 GPU

Anyone check the c code i posted?

I think this is an OpenCl issue because GPU produces incorrect results independently from platform and/or language (C, Java, Python etc.).

I'm aware of that. I tried another types than long but still gives incorrect results. Maybe some GPU's can't process 64 bit data types?

@grfrost If I am correct, why `(long) tc + 212600881053l;` line works but `(long) tc * 100;` not?

I doubt. even original OpenCl does not produce correct results. Check my kernel above.

@CC007 You're saying that you get same results both GPU and CPU mode right? Can you try my kernel in C?

It generated within build phase. Did you add kernel.cl to compile sources?

I built successfully with the modification @tawalaya did. Instead of giving another version of protobuf, I just used x86-64 artifact. So, in core/scheduler/build.gradle protobuf { protoc { artifact = 'com.google.protobuf:protoc:3.4.0:osx-x86_64'...