Optimising Fixed Point Arithmetic with Meta IP Cores
| Dates: | 13 February 2013 |
|---|
| Times: | 14:00 - 15:00 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | Department of Computer Science |
|---|
| Who is it for: | Adults, Current University students, University staff |
|---|
| Speaker: | Dr Matthew Fortune |
|---|
|
|
Fixed point arithmetic provides an efficient way of operating on real data types without the software or hardware overheads that come with more general floating point arithmetic. At first glance, a fixed point arithmetic maps directly to integer arithmetic operations but corner cases such as saturation, rounding and multiplication can lead to inefficient code sequences. Algorithms that require fixed point arithmetic tend to also have a significant quantity of data to operate on and efficient access to memory is therefore key. In this seminar we provide an introduction to the Meta architecture, a description of the DSP features that can be used to optimise fixed point arithmetic and a discussion of how compiler tehnology can target such hardware features.
http://www.cs.manchester.ac.uk/research/seminars/school
Speaker
Dr Matthew Fortune
Organisation: Imagination Technologies
Travel and Contact Information
Find event
Lecture Theatre 1.4
Kilburn Building
Manchester
