Question \#1 (a) What is memory coalescing? Which thread access pattern allows or not allow to perform memory coalescing? Your friend is a CPU fan boy and thinks that maximum memory bandwidth is always achieved when a single thread reads consecutive, increasing memory addresses. Using the concept of memory coalescing, explain why this is not the case on the GPU. (b) What is synchronization? Why we need synchronization primitive in GPU? Explain how syncthreads 0 should be used when bringing data from global to shared memory for use by all threads in a block