Multithreading

Question 1

What is a thread

Answer 1

lightweight process with own instructions and data
• Thread may be a process part of a parallel program of
multiple processes, or it may be an independent program
• Each thread has all the state (instructions, data, PC, register state, and so on) necessary to allow it to execute

Question 2

What is Data Level Parallelism

Answer 2

Perform identical operations on data, and lots of data (e.g., managing vectors and matrices).

Question 3

How does the amount of computation assigned to each thread Varys?

Answer 3

Amount of computation assigned to each thread is the grain size:
– Can be from a few instructions (more suited to
uniprocessor multithreading: for 1 processor, need n threads)
– To hundreds, or thousands of instructions
(multiprocessor multithreading: for n processors, need n threads)

Question 4

What are the Types of Multithreading

Answer 4

Several different types of multithreading for a superscalar processor:
– Coarse-grained Multithreading: when a thread is stalled, perhaps for a cache miss, another thread can be executed;
– Fine-grained Multithreading: switching from one thread to another thread on each instruction, such as Sun Niagara T1 and T2;
– Simultaneous Multithreading: multiple thread are using the multiple issue slots in a single clock cycle à exploiting TLP and ILP, such as IBM Power and Intel Core i7.

Question 5

Explain the coarse-grain multithreaded

Answer 5

Long stalls (such as L2 cache misses) are hidden by switching to another thread that uses the resources of the processor.

See picture 22

Question 6

Explain the Fine-grained Multithreading

Answer 6

Basic idea: At each clock cycle we must switch to another thread, in a sort of round-robin among active threads!

See picture 22

Question 7

Explain the Simultaneous Multithreding (SMT)

Answer 7

Simultaneously schedule instructions for execution from all threads to maximize the use of resources.

See picture 22