Note that fragments running on different ALUs on the same core share the same instruction stream!
Michael-hsiu
Are there heuristics to determine which ALU to send an instruction to? Or does the stream send to whichever ALU opens up first? I'm wondering if shader instructions can have "priorities", and one that comes later can be bumped up to an earlier spot in the stream to be done first.
fywu85
I think as soon as possible is one-way to do it. The priority thing sounds really CPU-esque and is what perhaps GPU lacks since each core is a lite version of a CPU. For more information, check out this stackoverflow post: https://computergraphics.stackexchange.com/questions/3627/how-does-a-gpu-process-a-task-by-using-multiple-alus
Note that fragments running on different ALUs on the same core share the same instruction stream!
Are there heuristics to determine which ALU to send an instruction to? Or does the stream send to whichever ALU opens up first? I'm wondering if shader instructions can have "priorities", and one that comes later can be bumped up to an earlier spot in the stream to be done first.
I think as soon as possible is one-way to do it. The priority thing sounds really CPU-esque and is what perhaps GPU lacks since each core is a lite version of a CPU. For more information, check out this stackoverflow post: https://computergraphics.stackexchange.com/questions/3627/how-does-a-gpu-process-a-task-by-using-multiple-alus