What’s the max number of bytes an array can be programmed onto the Tesla C870?
Our applicantion is real-time image processing, and our absolute worst case requires 5 arrays, each one of type complex_float32 and of length 65536 samples long (so that’s about 2.5MB) (1 array is the input coming in, 3 arrays are tables that we need to use for calculation, 1 is the output array)
Since it’s an image, so it’s 2D, which implies that there are lines. Let’s say there are N lines.
Ideally, I want to run 1 thread per line, so I would need N threads, each one works on a different like.
I would like all the threads to share the 3 table arrays.
If that’s my application, how many threads can be run concurrently, and would it fit?