Is it true that within GPU:

a$.$ thread block is a set of threads running on one multiprocessor,

b$.$ all threads of a single thread block can communicate with each other through shared

memory,

c$.$ threads of a single thread block can be synchronized,

d$.$ blocks of threads are divided into warps,

e$.$ warps are groups of threads that are scheduled together for execution,

f$.$ small number of threads per block cause load latency in device memory reads,

g$.$ assigning one block per multiprocessor makes the multiprocessor idle during thread

synchronization.