I would like to make use of the #GPU in my #Apple #Silicon system, but unfortunately it seems to only work with #Metal and in #SinglePrecision. If anyone has managed to do #DoublePrecision calculations (and can point me to a working example), that would be great! #Astrodon
#gpu #apple #silicon #metal #singleprecision #doubleprecision #Astrodon
What I found out:
- there’s a (simple) #metal #nbody example on GitHub that I could make work. It seems to be really fast.
- not sure yet how to link with #C++ but I found headers that should make this work
- metal doesn’t seem to support #DoublePrecision. Boo! #apple, please implement this!
#Metal #nbody #c #doubleprecision #apple
Hearsay has it that 64bit #DoublePrecision #FloatingPoint on #GPU is #slow, but #SoftFloat using two 32bit integers is even much #slower (I tested with #OpenCL).
#doubleprecision #floatingpoint #gpu #slow #softfloat #slower #opencl