NCA Performance Optimization & Debugging

Question 1

What is a common bottleneck in GPU performance?

Accepted Answer

Memory bandwidth

Answer

CPU clock speed.

Answer

Number of USB ports.

Answer

Disk space.

Question 2

What tool can be used to profile CUDA applications?

Accepted Answer

NVIDIA Nsight

Answer

Visual Studio Code.

Answer

Git.

Answer

Eclipse.

Question 3

What does kernel occupancy indicate?

Accepted Answer

Ratio of active warps to max warps

Answer

Percentage of CPU used.

Answer

Amount of memory allocated.

Answer

Number of threads created.

Question 4

What is the purpose of CUDA memory coalescing?

Accepted Answer

Group memory accesses for efficiency

Answer

Reduce compute power.

Answer

Increase cache misses.

Answer

Disable threads.

Question 5

Which technique helps identify race conditions in GPU code?

Accepted Answer

Using synchronization and detection tools

Answer

Ignoring synchronization.

Answer

Disabling threads.

Answer

Increasing thread count.

Question 6

What is warp divergence?

Accepted Answer

Threads take different paths

Answer

Threads execute in lock-step.

Answer

Threads share registers.

Answer

Threads communicate.

Question 7

How can shared memory be optimized for performance?

Accepted Answer

Avoiding bank conflicts

Answer

Ignoring access patterns.

Answer

Using global memory instead.

Answer

Increasing thread divergence.

Question 8

What is the main cause of GPU kernel launch overhead?

Accepted Answer

CPU-GPU synchronization

Answer

Data transfer speed.

Answer

Memory coalescing.

Answer

Thread creation.

Question 9

Which debugging tool is designed specifically for CUDA applications?

Accepted Answer

CUDA-GDB

Answer

GDB.

Answer

Visual Studio.

Answer

Valgrind.