Is there a way to get 32 bit (float) values out of Math.sin() or a similar class, maybe smth like Math.sin(float a)

Question

I am experimenting with using my GPU in java, I am using Aparapi but I have an issue where if I use (float)Math.sin(x) it still falls back to the CPU because my GPU doesn't support F64. is there a Math library for floats? should I be using the gpu to calculate sins at all?

Try using Java 32-bit. Just be wary that operations on ```long``` and ```double``` are not thread-safe. — Radu Mirea, Apr 04 '20 at 11:28
@MireaRadu Java 32-bit has nothing to do with the size of double. All types in Java are the same regardless of architectures — phuclv, Apr 04 '20 at 14:49

score 2 · Accepted Answer · answered Apr 05 '20 at 23:21

I would expect a project like this to come with math routines of its own so that you can call functions implemented on the GPU.

And it does: Kernel.sin(float) in Aparapi should map to a call to the OpenCL sin function.

Whether it makes sense to do trig on the GPU, that's for you to evaluate.

Is there a way to get 32 bit (float) values out of Math.sin() or a similar class, maybe smth like Math.sin(float a)

1 Answers1