I am using UE’s compute shader to do some math computing. I am currently using DXD11 on Windows 10, with RTX GPU.
I setup two engines with same configs. I implement two compute shaders using exactly the same codes. I feed same input texture to the compute shader by reading the same file.
Oddly, the two compute shaders produce different results. The values are close, but the error can archieve 10^-4. I already use float32 every where.
Since the math computing requires very high precision, I wonders how to config the shaders so that they always produce the same output.
