Hi. does anyone have an idea what would be faster, reading a big texture, where each thread reads a sample from it or reading from a buffer, say 64 floats, so they all read from a much smaller buffer, but each thread reads 64 floats. The thread is running on a compute shader. Thanks.