Professional Documents
Culture Documents
04/09/2010
Introduction
Scratch pad memory
Proposed methodology
Results
Conclusions
2
04/09/2010
INTRODUCTION
Scratch pad memory
Proposed methodology
Results
Conclusions
3
INTRODUCTION
04/09/2010
Scratch pad memory:
A high speed internal memory used for temporary
04/09/2010
Scratchpad vs. Cache:
A scratchpad doesn’t contain a copy of data that is stored
04/09/2010
The paper proposes scratchpad memory as an
alternative to cache memory as on-chip memory for
6
04/09/2010
Introduction
SCRATCH PAD MEMORY
Proposed methodology
Results
Conclusions
7
SCRATCH PAD MEMORY
04/09/2010
Memory array with the
Memory Cell decoding and the column
circuitry logic
04/09/2010
Area of scratchpad, As
As = Asde + Asda + Asco + Aspr + Asse + Asou
04/09/2010
For the memory array:
Ememcol = Cmemcol * Vdd2 * P0->1
Proposed methodology
Results
Conclusions
11
CACHE MEMORY
04/09/2010
Area model is based on
Tag Array Data Array
the transistor count in
12
04/09/2010
Introduction
Scratch pad memory
PROPOSED METHODOLOGY
Results
Conclusions
13
EXPERIMENTAL SETUP
04/09/2010
Compare same size cache with scratchpad memory
(the delay of cache is higher than scratchpad for the
04/09/2010
Performance estimation from the trace file.
An appropriate latency is added to the overall
04/09/2010
Authors assume a write through cache
Read Hit: Tag array is accessed. No write to cache and no access
to main memory
Write Miss: One cache tag read and main memory write. No
cache update.
04/09/2010
Benchmark
Cache
ARMulator Number of
trace analysis
Mapping Energy
CACTI
Algorithm Estimates
Cache/Scratch
Compiler Support Pad Size
Analytical Area
model Estimates
Scratchpad
Trace Analysis Number of 17
cycles
EXPERIMENTAL SETUP
04/09/2010
Target architecture:
AT91M40400, based on embedded ARM 7TDMI embedded processor
The area and performance estimates are made for the 0.5um 18
technology
04/09/2010
Introduction
Scratch pad memory
Proposed methodology
RESULTS
Conclusions
19
RESULTS Cache per access(2kB) 4.57 nJ
04/09/2010
Scratch pad per access(2kB) 1.53 nJ
Main memory read access, 2 bytes 24.00 nJ
The average area, time and AT Main memory read access, 4 bytes 49.30 nJ
product reductions are 34% Main memory write access, 4 bytes 41.10 nJ
04/09/2010
Spring 2010, EEL 6935, Embedded Systems
Figure: Energy consumed by the memory Figure: Comparison of cache and scratch pad
system memory area
21
04/09/2010
Introduction
Scratch pad memory
Proposed methodology
Results
CONCLUSION
22
CONCLUSION
04/09/2010
Presents an approach for selection of on-chip memory
configurations