Are the hardware threads spawned automatically determined by the processor itself? Or it is determined by the program just like software threads?
blah329
To summarize this slide: In contrast to the high latency hiding example from the previous slide, this has four execution contexts, meaning that the processor will execute within one of the four contexts on any given clock cycle. This exhibits low latency hiding ability, because if thread 1 has a long stall, while threads 2, 3, and 4 run and terminate quickly, it is possible for thread 1's stall to still not be over, and as such, this would bring the utilization of the processor down, versus if there were more threads.
Are the hardware threads spawned automatically determined by the processor itself? Or it is determined by the program just like software threads?
To summarize this slide: In contrast to the high latency hiding example from the previous slide, this has four execution contexts, meaning that the processor will execute within one of the four contexts on any given clock cycle. This exhibits low latency hiding ability, because if thread 1 has a long stall, while threads 2, 3, and 4 run and terminate quickly, it is possible for thread 1's stall to still not be over, and as such, this would bring the utilization of the processor down, versus if there were more threads.