Your question is GPU Algorithm Optimization. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you optimize a specific algorithm for a GPU architecture? Explain how you would choose an algorithm, expose parallelism, map work to a Qualcomm Adreno GPU, and reduce memory and synchronization costs. Use one concrete example, such as matrix multiplication, image convolution, or reduction, and discuss how you would validate the optimization with Qualcomm Snapdragon Profiler or Adreno GPU Profiler. Scope: focus on GPU compute concepts and OpenCL or Vulkan compute, not application-specific business requirements.