Submitted by: Submitted by Fastec
Views: 22
Words: 2940
Pages: 12
Category: Science and Technology
Date Submitted: 03/03/2015 02:59 AM
5/21/2013
GPU Architectures
A CPU Perspective
Derek Hower AMD Research 5/21/2013
Goals
Data Parallelism: What is it, and how to exploit it?
◦ Workload characteristics
Execution Models / GPU Architectures
◦ MIMD (SPMD), SIMD, SIMT
GPU Programming Models
◦ Terminology translations: CPU AMD GPU Nvidia GPU ◦ Intro to OpenCL
Modern GPU Microarchitectures
◦ i.e., programmable GPU pipelines, not their fixed-function predecessors
Advanced Topics: (Time permitting)
◦ The Limits of GPUs: What they can and cannot do ◦ The Future of GPUs: Where do we go from here?
GPU ARCHITECTURES: A CPU PERSPECTIVE
2
1
5/21/2013
Data Parallel Execution on GPUs
Data Parallelism, Programming Models, SIMT
GPU ARCHITECTURES: A CPU PERSPECTIVE 3
Graphics Workloads
Streaming computation
GPU
GPU ARCHITECTURES: A CPU PERSPECTIVE
4
2
5/21/2013
Graphics Workloads
Streaming computation on pixels
GPU
GPU ARCHITECTURES: A CPU PERSPECTIVE
5
Graphics Workloads
Identical, Streaming computation on pixels
GPU
GPU ARCHITECTURES: A CPU PERSPECTIVE
6
3
5/21/2013
Graphics Workloads
Identical, Independent, Streaming computation on pixels
GPU
GPU ARCHITECTURES: A CPU PERSPECTIVE
7
Architecture Spelling Bee
P-A-R-A-L-L-E-L
Spell ‘Independent’
GPU ARCHITECTURES: A CPU PERSPECTIVE
8
4
5/21/2013
Generalize: Data Parallel Workloads
Identical, Independent computation on multiple data inputs
= ( ) = ( ) = ( ) = ( )
0,7 1,7 2,7 3,7
7,0 6,0 5,0 4,0
GPU ARCHITECTURES: A CPU PERSPECTIVE
9
Naïve Approach
Split independent work over multiple processors CPU0
0,7 = ( ) 7,0
CPU1
1,7 = ( ) 6,0
CPU2
2,7 = ( ) 5,0
CPU3
3,7 = ( )
GPU ARCHITECTURES: A CPU PERSPECTIVE
4,0
10
5
5/21/2013
Data Parallelism: A MIMD...