It seems all the examples work with just the current set of gather and scatter arrays in the stream.
It seems all the examples work with just the current set of gather and scatter arrays in the stream. I was wondering if there is any way to access for read or write, the gather arrays from the previous sets of matricies in the stream. e.g. lets say from the kernle i read into the GPU 1000 sets including one gather and one scatter array each set, and then execute a Brook+ function function on the stream. When it executes on the first set containing one gather and one scatter, can that logic also access the sets of gather and scatter arrays queued up in the stream? Also is there any way to keep a global variable in the brook+ function which persists between executions on sets in the stream or does this sort of thing always have to be managed in the kernel?