Lesson 14.1: Data-Level Parallelism

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/33

encourage image

There's no tags or description

Looks like no tags are added yet.

34 Terms

1
New cards

Thread level parallelism

multicore processor and multithreading take advantage of _ _ _

2
New cards

Within a program

exploit thread-level parallelism is

3
New cards

Across application

exploit process-level parallelism is

4
New cards

Duplicate core, add extra logic

multicore, relatively simple extensions

5
New cards

Add storage, Associated Pipeline

multithreading, relatively simple extensions

6
New cards

ILP

executes multiple independent instructions from a single instruction stream at the same time

7
New cards

Pipeline, Superscalar design

ILP using

8
New cards

Instruction Level Parallelism

meaning of ILP

9
New cards

DLP

applies the same operation to multiple data items simultaneously

10
New cards

SIMD, GPUs

DLP using

11
New cards

Single Instruction Multiple Data

meaning of SIMD

12
New cards

Graphics Processing Unit

meaning of GPU

13
New cards

Data Level Parallelism

meaning of DLP

14
New cards

Independent Work, Process Parallel

prerequisite of data-level parallelism

15
New cards

Data level Parallelism

vector processors explicitly exploit _ _ _

16
New cards

Large Vector Register

vector processors gather data from memory into _ _ _

17
New cards

Register Register Operation

essentially performing many _ _ _ with the same opcode

18
New cards

Energy Efficient Computation

for vectorizable data, vector processors (1)

19
New cards

Hide memory latencies

for vectorizable data, vector processors (2)

20
New cards

Vector Register

holds a sequence of elements

21
New cards

Vector Instruction

applies one operation to corresponding elements

22
New cards

Vector Length

number of active elements for this instruction

23
New cards

Vector Lanes

parallel/pipelined datapaths that process elements

24
New cards

Vector Load/Store

moves contiguous or strided data to/from vector registers

25
New cards

Vector Registers, Scalar Register

Vector functional units (FUs) are fed by

26
New cards

Vector Register

provide independent data to operate on

27
New cards

Scalar Register

provide additional data or memory addresses

28
New cards

Fully Pipelined

vector FUs are _ _

29
New cards

SIMD

execution brings a (constrained) form of vector processing to general-purpose CPUs

30
New cards

32 bit register

scalar execution would put each value in its own _ _ _

31
New cards

Vector Register

SIMD packs them into a _ _

32
New cards

256 bit vector of 16 x 16 bit data

vector register

33
New cards

GPU

is originally developed as specialized hardware for graphics computation

34
New cards

Usage Patterns

GPU hardware evolution closely tied to evolution in _ _