It's quite error prone. Example: If the kernel argument is e.g. float one need to pass 0.f instead of simply 0. In other words the data type has to match exactly. It would be nice to pass primitive/scalar variables by value in order to benefit from implicit conversion.
It's quite error prone. Example: If the kernel argument is e.g.
floatone need to pass0.finstead of simply0. In other words the data type has to match exactly. It would be nice to pass primitive/scalar variables by value in order to benefit from implicit conversion.