x
Menu

Advanced algorithmic techniques for GPUs

University of Illinois, , Prof. Wen-mei Hwu

Updated On 02 Feb, 19

Overview

Includes

Lecture 3: Lectures 3 Avoidance of conflicts in resources

4.1 ( 11 )


Lecture Details

The GPU architecture is characterized by memory access bandwidth that, although fast, is often limiting in comparison to compute throughput. Thus, achieving performance critically depends on finding ways to reduce and regularize global memory access. Three important algorithmic strategies for conserving bandwidth are “registermemory tiling”, “layout transformation” and “thread coarsening”. These come at a cost of increased on-chip memory usage, which is also a limited resource. We will discuss a variety of examples from PDE solvers, linear algebra, and convolution.

Ratings

0


0 Ratings
55%
30%
10%
3%
2%
Comments
comment person image

Sam

Excellent course helped me understand topic that i couldn't while attendinfg my college.

Reply
comment person image

Dembe

Great course. Thank you very much.

Reply
Send