hetero lecture slides 002 lecture 1 lecture 1 1 overview
TRANSCRIPT
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
1/14
Course Overview
Lecture 1 1
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
2/14
Cour
Learn how to program heterogparallel computing systems a
high performance and ener
efficiency functionality and maintai
scalability across futuregenerations
Portability across vendor
Technical subjects
Parallel programming API, techniques
principles and patterns oalgorithms
processor architecture feconstraints
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
3/14
Instructor:
Wen-mei Hwu w-hwu@illinuse [Coursera HPP] to e-mail subject line
Teaching Assistants:Abdul Dakkak, Izzat El HaJablin, Andy Schuh, and c
ContributorsDavid Kirk, John StrattoGelado, John Stone, JavMichael Garland, TAs, an
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
4/14
Web Res
Website: Handouts, quizzes, labs,
slides/recordings Weekly view vs. classic v
Sample textbook chapters,documentation, software r
Electronic announcements
Forum discussions Forum for Q&A - the commuread and answer the postiyour classmates often hav
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
5/14
Quizzes: 50%
Weekly, repeatab
Labs (Machine Probl
Weekly, with optlater assignment
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
6/14
Academic Hone
You are allowed and enco
discuss assignments with
students in the class.
Getting verbal advice/hepeople who have already course is also fine.
Any sharing of code is
unacceptable
Includes posting your owthe Forum, reading someocode and then going off your own.
Giving/receiving help on
is unacceptable
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
7/14
Recommended Textbo
1. D. Kirk and W. Hwu, P
Massively Parallel Pro
Hands-on Approach, 2nMorgan Kaufman Publish
ISBN 0123814723
2. Lab assignments will h
accompanying instructi
notes3. NVIDIA, NVidia CUDA C
Guide, version 5.0, NV
(reference book)
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
8/14
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
9/14
ECE498AL ECE408/CS483 - Co
2/07 3/07
NVIDIA released
CUDA, UIUC lecture
and lab materialwent on-line
NAMD and other
apps group post
ECE498AL classprojects
VSC
China
relscho
6/08
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
10/14
ECE498AL ECE408/CS483 - Co
8/11 11/12
ECE498AL became
an official UIUC
courseECE408/CS483
First Coursera
HPP Offering
(25,000 students,10,000 finished)
S
Cour
COf
1/14
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
11/14
Tentative
Week One:
Introduction to HeterogenComputing, Overview of CUKernel-Based Parallel Pro
Lab tour and programming of vector addition in CUD
Week Two:
Memory Model for Localityfor Conserving Memory BanHandling Boundary ConditiPerformance Consideration
Programming assignment ofmatrix-matrix multiplicatC
Week Three:
Parallel Convolution Patt
Programming assignment ofmatrix-matrix multiplicatC
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
12/14
Tentative
Week Four:
Parallel Scan Pattern
Programming assignment o
convolution in CUDA C.
Week Five:
Parallel Histogram Patte
Atomic Operations
Programming assignment o
scan in CUDA C. Week Six:
Data Transfer and Task P
Programming assignment o
histogram in CUDA C.
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
13/14
Tentative
Week Seven:
Introduction to OpenCL, IntroC++AMP, Introduction to OpenA
Programming assignment of vecusing streams in CUDA C.
Week Eight:
Course Summary, Other Related
Models Thrust, Bolt, and CUD
Programming assignment of sim
matrix multiplication in choi
OpenCL, C++AMP, or OpenACC. Week Nine:
Complete any remaining lab as
with optional, bonus programmassignments in choice of Open
or OpenACC.
-
8/13/2019 Hetero Lecture Slides 002 Lecture 1 Lecture 1 1 Overview
14/14Welcome Ab
Recorded with Perso