Week 14

Question 1

_____ overlaps a sequence of tasks

Answer 1

pipelining

Question 2

_____ is executing tasks simultaneously

Answer 2

parallelism

Question 3

symmetric vs asymmetric

Answer 3

symmetric employs replicated HW – eg “cluster” supercomputers
asymmetric uses a collection of different HW elements

Question 4

fine-grained vs coarse-grained

Answer 4

fine-grained tpyically is instruction-level parallelism, ILP
coarse-grained parallelism is at the program, perhaps even thread, level

Question 5

explicit vs implicit

Answer 5

explicit parallelism results from programmer directives in SW
implicit parallelism is handled by HW

Question 6

SIMD: Single Instruction stream / Multiple Data streams:

Answer 6

One control unit and many processing elements, eg GPU and vector processor

Question 7

MIMD: Multiple Instruction streams / Multiple Data streams

Answer 7

MIMD has multiple SISD processors (typical) or multiple SIMD (rare), each MIMD processor fetch-executes its own program, may have dedicated I/O devices, eg cluster supercomputer

Question 8

True / False: The more processors you have, the faster your program will run

Answer 8

FALSE – more processors can eventually lead to a DECREASE of speed

Question 9

What part of the CPU time equation are MIPS and MFLOPS?

Answer 9

(Clock cycles / Instructions) * (seconds / clock cycle)