Mandelbrot example: double precision

inducer / pyopencl

OpenCL integration for Python, plus shiny features

http://mathema.tician.de/software/pyopencl

Other

1.05k stars 240 forks source link

Mandelbrot example: double precision #569

Open t0b3 opened 2 years ago

t0b3 commented 2 years ago

add double precision for

maxiter: ushort(16) -> uint(32)
output: ushort(16) -> uint(32)
z, q: complex64 -> complex128
real, nreal, imag: float -> double

works fine here

inducer commented 2 years ago

Could you explain what the presumed benefit is? An immediate downside is that devices without high-performance double support (Nvidia...) will suffer a decrease in performance. As a result, maybe this should be an option?

t0b3 commented 2 years ago

intend is to

extend the demo and explain how high precision can be implemented
enable this very example to dive into much lower zoom levels code works fine here on budget radeon

t0b3 commented 2 years ago

performance doesn't decrease compared to 32bit fyi this 64bit code performs fluent on 64bit cpus like

5$ entry level cloud vm cpu
raspi4 cpu