Journal of Information & Computational Science 12:9 (2015) 3373–3382 June 10, 2015
Available at http://www.joics.com
An Effective Modeling of Power Consumption and Time
Delay for SRAM Compiler
?
Jinlong Yan, Wenjuan Lu, Chunyu Peng
∗
School of Electronics and Information Engineering, Anhui University, Hefei 230601, China
Abstract
This paper presents an effective power consumption modeling, as well as access time delay modeling,
which aim to get the expenditure of power and time for various SRAM IP generated by SRAM compiler
without complicated calculation and specific circuit analysis during operation. According to a segment
analysis based on various MUX, bivariate power model on basis of two-dimensional bilinear interpolation
function is proposed. By dividing the delay time into several parts (mainly, decoder delay, word line
delay, bit line delay, and SA delay), critical path time delay network model is presented. Simulation
results by HSIM in SMIC 65 nm CMOS technology show that the average inaccuracy is about 5%.
Keywords: SRAM Compiler; Power Consumption; Time Delay; Modeling; Interpolation; Crital Path
1 Introduction
At present, Static Random Access Memory (SRAM) is being increasingly used in various ap-
plications, especially for SoC, with its high speed. Considering long design cycle of traditional
full-custom method, SRAM compiler based on semi-custom method has been required to shorter
the updating cycle of electronic products. For SRAM design, there are two major reference fac-
tors, one is the power consumption, which is crucial for low power SRAM devices, and the other
is time delay, of which major concerns is the read access time, T
cq
, which is calculated from
clock rising edge to data read out. However, owing to compiler’s flexibility and scalability [1],
calculation or simulation becomes the bottleneck to get the performance parameter of the SRAM
IP. Thus, effective modeling of power consumption and access time delay is necessary to serve
SRAM compiler [2-4]. Here, bivariate power model and critical path time delay network model
for power and delay are respectively presented.
The rest of this paper is mainly organized as follows: Section 2 reviews the relevant background
information and gives the architecture of SRAM compiler, and indicates the importance of the
modeling of power consumption and time delay for SRAM compiler. Section 3 discusses the most
important factors affecting the SRAM power dissipation and then presents the corresponding read
?
Project supported by the National Nature Science Foundation of China (No. 61474001).
∗
Corresponding author.
Email address: pengchunyu608@126.com (Chunyu Peng).
1548–7741 / Copyright © 2015 Binary Information Press
DOI: 10.12733/jics20105961