What is the difference between Limiter Counters and Utilization Counters about GPU??

As the documentation says, Limiter counters tell you which subsystems of the GPU are active by providing a percentage of the total number of processor cycles during which this subsystem was active. Besides, Instrument also provides some Utilization Counters and the value is different from the Limiter. What do Utilization Counters mean?

Accepted Reply

Limiter = work + stall, utilization = work. Typically you will expect to see one limiter at 100% and that will be the bottlenecked block, but the utilization may not be 100% depending on the amount of stalling the block is doing. Stalling is any waiting on dependencies. For example, a cache may stall waiting for a request to come back from main memory, or the ALU may stall when it issues a low-throughput instruction that takes several cycles to complete. Basically, you would use the limiter to locate the bottlenecked block and then look at the utilization to decide how efficiently that block is being used.

Replies

Limiter = work + stall, utilization = work. Typically you will expect to see one limiter at 100% and that will be the bottlenecked block, but the utilization may not be 100% depending on the amount of stalling the block is doing. Stalling is any waiting on dependencies. For example, a cache may stall waiting for a request to come back from main memory, or the ALU may stall when it issues a low-throughput instruction that takes several cycles to complete. Basically, you would use the limiter to locate the bottlenecked block and then look at the utilization to decide how efficiently that block is being used.