Lecture 18
SIMD: Vector Processing
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
SIMD: Vector Processing - Lecture 18 from ECE 412, University of Illinois, Fall 2001, Study notes of Computer Architecture and Organization
A portion of a lecture from the university of illinois' ece 412 course, fall 2001, focusing on simd (single instruction, multiple data) vector processing. Topics include predication vs. Prediction, moving from general-purpose to specific application domains, and the history and issues of vector supercomputers. The lecture also covers daxpy, a common kernel in scientific codes, and vectorization challenges.
Typology: Study notes
Pre 2010
1 / 6
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download SIMD: Vector Processing - Lecture 18 from ECE 412, University of Illinois, Fall 2001 and more Study notes Computer Architecture and Organization in PDF only on Docsity!
Lecture 18
SIMD: Vector Processing
Predication vs. Prediction
- What is the penalty associated with predication?
- When should a branch be predicated, when should it be predicted?
- On the Forefront: dynamic techniques
DAXPY :
A common kernel in scientific codes
- Vector times a scalar plus a vector
- Flynn’s Bottleneck becomes a problem
Y = a X + Y
The Vector Supercomputer Realm
- Early 70s: Control Data Corp
- Mid 70s: The Cray I (1976)
- 80s through the 90s: Cray, Convex, Hitachi, Fujitsu, NEC…