Profiling Options for CUDA and HIP¶
Quiz Questions and Answers¶
Question 1¶
What API is used in CUDA to calculate elapsed execution time?
- A)
cudaEventElapsedTime
- B)
cudaEventSynchronize
- C)
cudaEventRecord
- D)
cudaDeviceSynchronize
Answer:
A) cudaEventElapsedTime
Question 2¶
True or False: HIP and CUDA use the same APIs for event-based time measurement.
Answer:
False
Explanation: HIP provides analogous APIs (hipEvent_t
, hipEventCreate
, etc.) but not identical to CUDA.
Question 3¶
What tool would you use for system-wide profiling in ROCm?
- A) Nsight Systems
- B) Nsight Compute
- C) rocprof
- D) rocminfo
Answer:
C) rocprof
Question 4¶
True or False: The CUDA Occupancy Calculator determines the percentage of GPU memory used by the application.
Answer:
False
Explanation: The CUDA Occupancy Calculator determines the ratio of active warps to the maximum number of warps supported per multiprocessor.
Question 5¶
Which of the following is NOT a step in using CUDA events for timing?
- A) Record start and stop events using
cudaEventRecord
- B) Calculate elapsed time using
cudaEventElapsedTime
- C) Free memory using
cudaFreeEvent
- D) Synchronize events using
cudaEventSynchronize
Answer:
C) Free memory using cudaFreeEvent
Explanation: cudaEventDestroy
is used to destroy events, not cudaFreeEvent
.