i want to implement a BigFloat (arbitrary precision float) application on the GPU and i thought using SyCL would be a good solution for this.
The basic C++ implementation works on a vector of bytes and functions good.
My idea was to create a seperate class and copy of all methods that don’t allocate data and use an accessor to a buffer that is created from the vector of the original class.
This buffer is correctly allocated and the accessor is stored as an attribute in the SyCL BigFloat class. The original instance is alive during the complete process.