Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

A Survey on Minimizing Energy Consumption of VLSI Processors Using Multiple Supply Voltages by ijcsiseditor


The International Journal of Computer Science and Information Security (IJCSIS) is a well-established publication venue on novel research in computer science and information security. The year 2010 has been very eventful and encouraging for all IJCSIS authors/researchers and IJCSIS technical committee, as we see more and more interest in IJCSIS research publications. IJCSIS is now empowered by over thousands of academics, researchers, authors/reviewers/students and research organizations. Reaching this milestone would not have been possible without the support, feedback, and continuous engagement of our authors and reviewers. Field coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. ( See monthly Call for Papers) We are grateful to our reviewers for providing valuable comments. IJCSIS December 2010 issue (Vol. 8, No. 9) has paper acceptance rate of nearly 35%. We wish everyone a successful scientific research year on 2011. Available at IJCSIS Vol. 8, No. 9, December 2010 Edition ISSN 1947-5500 � IJCSIS, USA.

More Info
									                                                                (IJCSIS) International Journal of Computer Science and Information Security,
                                                                Vol. 8, No. 9, December 2010

      A Survey on Minimizing Energy Consumption of
      VLSI Processors Using Multiple Supply Voltages
                                                        B. Sathiyabama and Dr. S. Malarkkan

Abstract—Due to the continuous increase in earth's population,                        The interesting thing to be noted is that the inventor of this
adequate supply of resources is going to be a major issue. One basic               device states that this unique power controller design analyses
essential resource in rising demand is energy and in particular                    power consumption using an artificial intelligence algorithm
electrical energy. The contributions of the scientific community                   implemented on a high-end micro-controller. The key point
toward the goal of sustainability with regard to energy consumption
                                                                                   here is the “high-end micro-controller", which is a good
of embedded systems are previously discussed in many research
                                                                                   example for an embedded system and how they can be present
works. Low power has become one of the major design issues due to
the increased demand in personal computing devices and portable                    without being really noticed, often in surprising quantities.
communication system. In this paper a survey on minimizing energy                  This leads to the main topic, the energy consumption of
consumption of VLSI Processors using multiple supply voltages is                   embedded systems or VLSI Processors and the various
presented. This survey discusses on search method for a scheduling                 strategies available to reduce it. Besides helping to save the
and module selection problem using multiple supply voltages so as to               environment, reducing energy consumption of embedded
minimize dynamic energy consumption under time and area                            systems can lead to immediate monetary rewards for their
constraints. The algorithm based on a genetic algorithm is surveyed                producers, for example increased sales of the mobile phone
to find near-optimal solutions in a short time for large-size problems.
                                                                                   with the longest standby time, which is suspected to be the
The literature related to the multiple supply voltages with genetic
                                                                                   main reason behind the efforts to minimize power
approach and energy consumption minimization in various VLSI
systems is presented.                                                              consumption. In this survey the Genetic Approach based
                                                                                   Minimizing energy consumption of VLSI Processors Using
 Keywords— Energy minimization, Functional                    pipelining,          Multiple Supply Voltages is presented.
Multiple supply voltages, dynamic power, scheduling.
                                                                                                       II.    LITERATURE SURVEY
                           I.     INTRODUCTION                                        This section presents the literature survey on the
                                                                                   minimization of power consumption in VLSI processors. The
A     T the time of writing, 6.7 billion people live on earth. It
      is estimated that the world population will reach 9 billion
by 2040. This increase alone will result in rapidly rising
                                                                                   power consumption can be reduced only if the cause for the
                                                                                   power dissipation is found.
energy consumption. This problem is amplified by rising                               Massoud Pedram [17] presented the cause for the power
living standards in developing countries. The present world                        dissipation in VLSI processors. Power dissipation in CMOS
electricity production is about 20 trillion kilowatt-hours. By                     circuits is caused by the three sources: 1) the leakage current
2030, it will reach 30 trillion kilowatt-hours, mostly through                     which is primarily determined by the fabrication technology,
coal and natural gas. Coal is a very dirty form of energy,                         consists of reverse bias current in the parasitic diodes formed
especially when its emissions are not properly filtered, which                     between source and drain diffusions and the bulk region in a
is mostly the case in developing countries. This will result in                    MOS transistor as well as the subthreshold current that arises
serious repercussions for the environment. It stands to reason                     from the inversion charge that exists at the gate voltages
to try to limit those consequences. One approach is sustainable                    below the threshold voltage, 2) the short-circuit (rush-through)
development, which means to use resources to meet ones                             current which is due to the DC path between the supply rails
needs in such a way that future generations have the ability to                    during output transitions and 3) the charging and discharging
meet their needs in the available environment. Producing                           of capacitive loads during logic changes.
energy in ways that destroy the environment obviously                                 The diode leakage takes place when a transistor is turned
contradicts this goal. In theory, there is more than enough                        off and another active transistor charges up or down the drain
solar energy available to supply the entire world. In practice,                    with respect to the first transistor’s bulk potential. The ensuing
harvesting this energy in a cheap and efficient way is not easy.                   current is proportional to the area of the drain diffusion and
Therefore it is important to put the created energy to good use                    the leakage current density. The diode leakage is typically 1
instead of wasting it.                                                             picoA for a 1 micro-meter minimum feature size! The
                                                                                   subthreshold leakage current for long channel devices
                                                                                   increases linearly with the ratio of the channel width over
                                                                                   channel length and decreases exponentially with VGS-Vt where
     B. Sathiyabama, Research Scholar Sathyabama University, Chennai,
                                                                                   VGS is the gate bias and Vt is the threshold voltage. Several
India.                                                                             hundred millivolts of “off bias” (say, 300-400 mV) typically
   Dr. S. Malarkkan, Principal, Manakulavinayagar Institute of Technology,         reduces the subthreshold current to negligible values. With
Pondicherry, India.

                                                                                                               ISSN 1947-5500
                                                          (IJCSIS) International Journal of Computer Science and Information Security,
                                                          Vol. 8, No. 9, December 2010

reduced power supply and device threshold voltages, the sub                 circuit delay remains less than the clock period provides the
threshold current will however become more pronounced. In                   opportunity to reduce power consumption of VLSI circuits.
addition, at short channel lengths, the subthreshold current                The objective is to change the voltage of the body bias to
also becomes exponentially dependent on drain voltage                       reduce leakage, allowing the circuit to consume less power
instead of being independent of VDS (see [18] for a analysis).              whenever the clock edge can be met as detected beforehand.
The subthreshold current will remain 102 - 105 times smaller                   Hariyama et al, [1] proposed a novel approach on
than the “on current” even at submicron device sizes.                       minimizing energy consumption of VLSI processors based on
   The short-circuit (crowbar current) power consumption for                dual-supply-voltage       assignment     and    interconnection
an inverter gate is proportional to the gain of the inverter, the           simplification. His work presents a design technique to
cubic power of supply voltage minus device threshold, the                   minimize energy of both functional units (FUs) and an
input rise/fall time, and the operating frequency [19]. The                 interconnection network between FUs. To reduce complexity
highest short circuit current flows when there is no load; this             of the interconnection network, data transfers among FUs are
current decreases with the load. If gate sizes are selected so              classified according to FU types of operations in a data flow
that input and output rise/fall times are about equal, the short-           graph. The basic idea behind reducing the complexity of
circuit power consumption will be less than 15% of the                      interconnection network is that the interconnection resource
dynamic power consumption. If, however, design for high                     can be shared among data transfers with the same FU type of a
performance is taken to the extreme where large gates are used              source node and the same FU type of a destination node.
to drive relatively small loads, then there will be a stiff penalty         Furthermore, an efficient method based on a genetic algorithm
in terms of short-circuit power consumption.                                is presented for large-size problems.
   The short-circuit and the leakage currents in CMOS circuits                 Xiaoying et al, [25] studied the problem of leakage power
can be made small with proper circuit and device design                     reduction by means of input vector control, and develop a
techniques. The dominant source of power dissipation is thus                platform for CMOS combinational circuit leakage power
the charging and discharging of the node capacitances (also                 reduction. Genetic algorithm is worn for searching minimum
referred to as the dynamic power dissipation) and is given by:              leakage vector with circuit status difference as fitness
                                                                            function. Experimental results indicate that the proposed
                          0.5                                               method can achieve satisfied leakage power reduction, and the
where C is the physical capacitance of the circuit, Vdd is the              run time is reasonable. This method has no necessity for Spice
supply voltage, E(sw) (referred as the switching activity) is the           simulation and independent from target technology.
average number of transitions in the circuit per 1/fclk time, and              Hariyama et al, [2] proposed a Genetic approach to
fclk is the clock frequency.                                                minimizing energy consumption of VLSI processors using
    Ishikawa et al., [20] discussed on the power dissipation                multiple supply voltages. The author presents an efficient
control. A bit-serial multiple-valued reconfigurable VLSI                   search method for a scheduling and module selection problem
using current-mode logic circuits has been introduced by the                using multiple supply voltages so as to minimize dynamic
author. A Differential-Pair Circuit (DPC) is used as a basic                energy consumption under time and area constraints. The
component of a cell, so that the static power is dissipated even            proposed algorithm is based on the genetic algorithm so that it
in the nonactive cells. To solve the problem, autonomous                    can find near-optimal solutions in a short time for large-size
ON/OFF control of the current sources is presented based on                 problems, n efficient search can be achieved by crossover that
superposition of bit-serial data and current-source control                 prevents generating nonvalid individuals and a local search is
signals. In the proposed switched current control technique,                also utilized in the algorithm. Experimental results for large-
the static power dissipation can be greatly reduced because                 size problems with 1,000 operations exhibits that the proposed
current sources in nonactive circuit blocks are turned off. The             method can achieve significant energy reduction up to 50
superposition of data and control signals in a single                       percent and can find a near-optimal solution in 10 minutes.
interconnection is effectively utilized to reduce complexity of             Conversely, the ILP-based method cannot find any feasible
switches and interconnections, and to eliminate skew between                solution in one hour for the large-size problem, even if a state-
data and control signals. It is evaluated that the reduction of             of-art mathematical programming solver is used.
the power dissipation is remarkable, if the operating ratio is                 W. Hung et al, [22] discussed briefly on the techniques in
less than 75%.                                                              reduction of power consumption. In this paper, the author
    Xin He Al-Kadry, [21] proposed a novel concept to                 presents an algorithm for the minimization of total power
control power dissipation in VLSI processors. This paper                    consumption via multiple VDD assignment, multiple VTH
emphasizes on adaptive leakage control using body bias                      assignment, device sizing and stack forcing, while maintaining
technique to reduce the power dissipation of the 65 nm MOS                  performance requirements. These four power reduction
devices. Through adding forward body biasing, the leakage is                techniques are correctly encoded in genetic algorithm and
reduced in sub-100 nm CMOS devices (unlike above-100 nm                     evaluated simultaneously. The overhead imposed by insertion
devices) while slightly increasing the signal propagation                   of level converters is also taken into account. The
delay. For the conditions where the circuit does not use up the             effectiveness of each one of power reduction mechanism is
entire clock cycle, this slack can be used to reduce the power              verified, as are the combinations of different approaches.
dissipation without any loss in performance. The fact that the              Experimental results are provided by the author for a number

                                                                                                        ISSN 1947-5500
                                                       (IJCSIS) International Journal of Computer Science and Information Security,
                                                       Vol. 8, No. 9, December 2010

of 65 nm benchmark circuits that span typical circuit                    diminution in power consumption comes at no additional
topologies, including inverter chains, SRAM decoders,                    impact to area or performance and does not require any
multiplier and a 32bit carry adders. From the experimental               alteration to the design flow. It is exposed that the number of
results, the author shows that the combination of four low               possible input orders increases exponentially in the number of
power techniques is the effective way to achieve low power               inputs to the checker. As a result, the computational cost of
budget.                                                                  finding the optimum input order can be very expensive as the
   Power dissipation in CMOS digital circuits consists of                number of inputs to the checker increases. The author presents
dynamic power, short circuit power and static power. Short               a very effective technique to build a reduced cost function to
circuit power consumption can be kept within bounds by                   solve the optimization problem to find a near optimal input
careful design and tuning the switching characteristics of               order. It scales well with growing number of inputs to the
complementary logic (slope engineering); it is usually                   checker, and the computational costs are independent of the
negligible compared to dynamic power and leakage power.                  complexity of the checker. Experimental results illustrates that
Dynamic power was once the dominant power consumption                    a reduction in power consumption of 16% on the average for
term. However, as the result of technology scaling and VTH               several types of checkers can be obtained using the proposed
(threshold voltage) decreasing, leakage power will soon                  technique.
account for a large portion of total power consumption.                     Mohanty et al, [4] puts forth an Energy efficient scheduling
   Although there are many techniques to reduce power                    for datapath synthesis. In his paper, two new algorithms are
dissipation, most existing works focus on one technique in               described for datapath scheduling which aim at energy
isolation instead of concurrently applying a number of power             reduction while maintaining performance. The given
minimization techniques. In [22], a power optimization                   algorithms, time constrained and resource constrained, utilize
framework based on the genetic algorithm is presented. The               the concepts of multiple supply voltage and dynamic clocking
optimization strategy combines four power reduction                      for energy minimization. In dynamic clocking, the functional
techniques: multiple VDD assignment, multiple VTH                        units can be worked at different frequencies depending on the
assignment, gate sizing, and stack forcing. It simultaneously            computations occurring within the datapath during a given
applies and evaluates the effects of these techniques to achieve         clock cycle. The plan is to schedule high energy units, for
maximum power saving under a hard timing constraint. The                 instance the multipliers at lower frequencies such that they can
framework can be easily extended to include other power                  be operated at lower voltages to reduce energy consumption
reduction techniques. To the best of our knowledge, this is the          and the low energy units, such as adders at higher frequencies,
first power optimization framework that simultaneously uses              to compensate for speed. The algorithms have been applied to
all of these four power reduction techniques.                            a variety of high level synthesis benchmark circuits under
   Compared to the ILP approach that were used by [23] [24],             different time and resource constraints. The experimental
one advantage of GA approach is that the parallel nature of              results demonstrate that for the time constrained algorithm,
genetic algorithms suggests parallel processing as the natural           energy savings in the range of 33-75% are obtained. Similarly,
route to explore. The author implements a parallel version of            for the resource controlled algorithm, under various resource
their algorithm, by dividing the population processing among             constraints using two supply voltage levels (5.0 V, 3.3 V)
multiple processors. The authors notice that in average more             energy savings in the range of 24 - 53% can be obtained.
than 3X run-time speed-up on a 4-processor workstation                      Muthumala et al., [5] presents a technique to minimize the
against the single-processor version of the algorithm (The               total energy consumption under time and area constraints,
reason that the author cannot achieve a 4X speedup is the                considering interconnection and functional unit energy.
interaction overhead among parallel processes). Another                  Multiple supply and threshold voltage method is used to
advantage is that for ILP approach, the running time for a               minimize the static and dynamic energy in the functional
large circuit may be prohibitively long; while for the GA-               units. A genetic algorithm based search technique is proposed
based strategy, it can be set a proper termination criterion to          for the energy consumption minimization problem, so that
tradeoff the runtime and power saving.                                   near-optimal solution can be found in a reasonable time for
   Mohanram et al, [3] discussed on the topic of energy                  large-size problems. Interconnection simplification is attained
consumption in which Lowering power consumption in                       by increasing the sharing of interconnections among
concurrent checkers via input ordering is presented. The                 functional units. Experimental results show that up to 30% of
author presents an efficient and scalable technique for                  energy savings can be achieved by this proposed method.
lowering power consumption in checkers used for concurrent                  Kamble et al, [6] presented a comparative study on Energy-
error detection. The basic idea is to make use of the functional         efficiency of VLSI caches. The author investigates the use of
symmetry of concurrent checkers with respect to their inputs,            organizational alternatives that lead to more energy-efficient
and to order the inputs such that switching activity (and hence          caches for contemporary microprocessors. Dissipative
power consumption) in the checker is minimized. The inputs               transitions are likely to be very correlated and skewed in
of the checker are typically driven by the outputs of the                caches, precluding the use of simplistic hit/miss ratio based
function logic and check symbol generator logic-spatial                  power dissipation models for accurate power estimations. The
correlations between these outputs are analyzed to compute an            authors use a detailed register-level simulator for a typical
input order that minimizes power consumption. The                        pipelined CPU and its multi-level caches, and simulate the

                                                                                                    ISSN 1947-5500
                                                       (IJCSIS) International Journal of Computer Science and Information Security,
                                                       Vol. 8, No. 9, December 2010

execution of the SPECint92 benchmarks to glean accurate                  Dynamic Voltage Scaling (DVS), Dynamic Frequency Scaling
transition counts. A detailed dissipation model for CMOS                 (DFS) can be used to find optimized solution. Various
caches is brought up for estimating the energy dissipation               concepts such as pipelining, parallel processing, retiming,
based on electrical parameters of a typical circuit                      unfolding, systolic array etc. are used in design of modern
implementation and the transition counts collected by                    VLSI based low power.
simulation. A block buffering method is presented to allow                  Implementation of VLSI design algorithms includes high
cache energy requirements to be reduced without increasing               level architectural transformations. Pipelining ,parallel
access latencies. The authors report results for a system with           processing, retiming ,unfolding, folding and systolic array
an off-chip L2 cache. The authors conclude that block                    design methodologies plays an important role for optimized
buffering, with sub-banking to be very effective in reducing             high performance design. Similarly, high level algorithm
energy dissipation in the caches, and in the off-chip I/O pad            transformations such as strength reduction look ahead and
drivers.                                                                 relaxed look ahead are also utilized for design implementation.
   Santosh Chede et al, [7] proposed on the Significance of              Strength reduction transformations are applied to minimize the
VLSI Techniques for Low Power Real Time Systems. In                      number of multiplications in convolution, parallel infinite
microelectronics design, power consumption and the speed of              impulse response (FIR) digital filters, discrete cosine
operation, are crucial constraints. Propagation delay of the             transforms (DCTs) and parallel rank –order filters. Look ahead
circuit component has an impact on such factors. Pipelining              and relaxed look ahead transformations are pertained to design
and parallel processing strategies are used for desirable                pipelined direct form and lattice recursive digital filters and
propagation delays and hence for clock and throughput                    adaptive filters and parallel recursive digital filters. And these
variation respectively. To some extent variation in propagation          strategies are used to develop and design architectures for
delay is accountable for power consumption reduction. In his             multiplication, addition, digital filters, pipelining styles, low
work, pipelining and parallel processing concepts are analyzed           power computations and architectures for high performance
with reference to task scheduling in real time system. Power             programmable or ultra low power embedded digital signal
consumption and speed of operation problems of such systems              processors, applicable to various biomedical, industrial,
are analyzed.                                                            defense, consumer applications etc [8].
   Main goal of most of the system level or circuit design are              Jui-Ming Chang et al, [9] proposed a novel technique for
high performance and power optimization. For high                        Energy Minimization Using Multiple Supply Voltages. A
performance system design, propagation delay minimization                dynamic programming technique is presented for solving the
plays an important role. Basically size, cost, performance and           multiple supply voltage scheduling problems in both
power consumption are the crucial issues in low power                    nonpipelined and functionally pipelined data-paths. The
portable battery operated system design. Excessive power                 scheduling problem refers the assignment of a supply voltage
dissipation which overheats thereby degrading the                        level (selected from a fixed and known number of voltage
performance and lifetime is not at all affordable. Energy                levels) to each operation in a data flow graph so as to
consumption being an important constraint for battery life               minimize the average energy consumption for given
estimation, VLSI based low power design of dedicated                     computation time or throughput constraints or both. The
multimode signal conditioning integrated circuit is desirable.           energy model is accurate and accounts for input pattern
Modern systems consist of digital realization of analog                  dependencies, re-convergent fanout induced dependencies,
processes and this helps to design system with high precision,           and the energy cost of level shifters. Experimental results
high signal to noise ratio (SNR), repeatability and flexibility.         illustrate that using three supply voltage levels on a number of
DSP systems can be realized with custom designed hardware                standard benchmarks, an average energy saving of 40.19%
circuits or ultra low power high performance programmable                (with a computation time constraint of 1.5 times the critical
processors fabricated using VLSI circuit technology.                     path delay) can be attained compared to using a single supply
   Essentially the role of digital system is to maximize the             voltage level.
performance with minimum cost and less time to market.                      One driving factor behind the push for low power design is
Performance measures are throughput, clock rate, circuit                 the growing class of personal computing devices as well as
complexity and power dissipation or total energy consumed to             wireless communications and imaging systems that demand
execute a real/non real time task. In order to design complex            high-speed computations and complex functionalities with low
digital system using VLSI technology, modeling with node                 power consumption. Another driving factor is that excessive
identification is essential. Generally to carry out design, DSP          power consumption has become a limiting factor in integrating
algorithms are realized and transformed to hardware. To                  more transistors on a single chip. Unless power consumption
investigate and analyze data flow and data paths i.e.                    is considerably reduced, the resulting heat will limit the
parallelism and pipelining among tasks and subtasks, system              feasible packing and performance of VLSI circuits and
modeling methods like block diagrams, Signal flow graph                  systems. The most effective way to lessen power consumption
(SFG), Data flow Graph (DFG), Dependence graph etc. is                   is to lower the supply voltage level for a circuit. Reducing the
very much required. In such design there is trade off between            supply voltage however increases the circuit delay.
sampling frequency, operating frequency and power                        Chandraskan et al. [10] compensate for the increased delay by
consumption, in order to design high performance system.                 shortening critical paths in the data-path using behavioral

                                                                                                     ISSN 1947-5500
                                                        (IJCSIS) International Journal of Computer Science and Information Security,
                                                        Vol. 8, No. 9, December 2010

transformations such as parallelization or pipelining. The                and to aid in the comparison of results. Results show that
resulting circuit consumes lower average power while meeting              GALOPS achieves significant power reductions in the
the global throughput constraint at the cost of increased circuit         presented benchmark designs. In addition, GALOPS generates
area. More recently, the use of multiple supply voltages on the           a family of unique solutions for each design, all of which
chip is attracting attention. This has the advantage of allowing          satisfy the multiple design objectives, providing flexibility to
modules on critical paths to use the highest voltage level (thus          the VLSI designer.
meeting the required timing constraints) while allowing                      The power consumption of the very large scale integration
modules on noncritical paths to use lower voltages (thus                  (VLSI) devices has become an important parameter in recent
reducing the energy consumption). This method tends to result             years, largely due to the explosion in the use of portable
in smaller area overhead compared to parallel architectures.              communication and computing systems. Of particular interest
   There are, however, a number of practical problems that                in such systems are digital signal processing (DSP) devices.
must be overcome before use of multiple supply voltage                    These devices are specialized processors used widely in
becomes prevalent. These problems include routing of                      complex functions such as telecommunications, data
multiple supply voltage lines, area/delay overhead of required            compression, and speech processing. The increasing
level shifters, and lack of design tools and methodologies for            requirements for portable systems to incorporate these
multiple supply voltages. The first issue is an important                 functions have led to increased demand for low-power DSP
concern which should be considered by any designer who                    devices. However, the design of DSP devices that offer
wants to use multiple supply voltages. That is, there is a                complex functions with low-power consumption requires the
tradeoff between lower energy dissipation and higher routing              use of advanced automated synthesis tools.
cost. The remaining problems (that is, level shifter cost and                Traditional automated synthesis tools optimized the VLSI
lack of tools) are addressed in the authors work. That is,                device for speed and area. This presented a complex solution
shown that the area/delay overhead of level shifters is                   space, especially when considered at the high level of
relatively small and will present an effective algorithm for              abstraction (the design capture stage). The addition of power
using multiple supply voltages during behavioral synthesis.               as a design objective compounds the complexity of the high-
   In this context, an important problem is to assign a supply            level synthesis task. Therefore, high-level low-power
voltage level (selected from a finite and known number of                 synthesis tools require a more robust search and optimization
supply voltage levels) to each operation in a data flow graph             mechanism to produce designs with optimal tradeoffs between
(DFG) and schedule various operations so as to minimize the               the objective parameters. Genetic algorithms (GAs) [12] have
energy consumption under given timing constraints. To this                proven to be a successful technique for tackling the complex
problem is referred as the multiple-voltage scheduling                    problems inherent in the design and optimization of VLSI
problem or the MVS problem for short. In his work, the                    devices. Examples are VLSI synthesis tools such as that
problem is tackled in its general form. Chandraskan et al in              developed by Arslan et al. [13], which uses a GA for the
[10] shows that the MVS problem is NP-hard even when only                 structural synthesis of logic circuits. A tool for reducing VLSI
two points exist on the energy-delay curve for each module                implementation area by using a GA to reduce the size of
(these curves may be different from one module to another),               functional operators was developed by Martin and Knight
and then propose a dynamic programming approach for                       [14]. The authors have previously demonstrated the
solving the problem. This algorithm which has pseudo-                     application of GAs in low-power synthesis [15], [16] using a
polynomial complexity (cf., Section IV-C) produces optimal                restricted GA with limited genetic operators.
results for trees, but is suboptimal for general directed acyclic            In his work, genetic algorithm for low-power synthesis
graphs. The dynamic programming technique is then                         (GALOPS) is presented, a GA for the high-level synthesis of
generalized to handle functionally pipelined designs. This is             low power CMOS-based DSP designs. GALOP uses problem-
the first time that the use of multiple supply voltages in a              specific techniques for low-power synthesis. As illustrated
functionally pipelined design is considered. A novel revolving            here, the incorporation of these techniques into the GA
schedule for handling these designs is presented by this                  framework requires modification of the standard genetic
author.                                                                   operations. The primary objective of GALOPS is to construct
   Bright et al, [11] presents a new tool for the synthesis of            a minimum power design from an initial high-level
low-power VLSI designs, specifically, those designs targeting             specification, while tracking other performance constraints
digital signal processing applications. The synthesis tool                such as area and speed. The author presents the application of
genetic algorithm for low-power synthesis represented as                  GA for minimizing power consumption in four contributions.
GALOPS uses a genetic algorithm to apply power-reducing                   (1) The formulation of a GA that can be able in handling the
transformations to high-level signal-processing designs,                  VLSI low-power synthesis problem, considering the specified
producing designs that satisfy power requirements as well as              performance constraints of CMOS devices (2) The
timing and area constraints. GALOPS use the problem-                      development of problem-specific genetic operators that
specific genetic operators that are specifically tailored to              manipulate a library of low-power transformations. Standard
incorporate VLSI-based digital signal processing design                   genetic operators are modified to incorporate these
knowledge. A number of signal-processing benchmarks are                   transformations, in particular, a crossover operator is
utilized to facilitate the analysis of low-power design tools,            developed which recognizes and applies power-saving

                                                                                                     ISSN 1947-5500
                                                                  (IJCSIS) International Journal of Computer Science and Information Security,
                                                                  Vol. 8, No. 9, December 2010

transformations to designs. The developed crossover operator                         [13] T. Arslan, D. H. Horrocks, and E. Ozdemir, “Structural cell-based VLSI
                                                                                          circuit design using a genetic algorithm,” in Proc. IEEE Int. Symp.
preserves the notion of inheritance in a genetic algorithm (3)                            Circuits Syst., vol. 4, 1996, pp. 308–311.
Results that show the significant power reductions obtained                          [14] R. S. Martin and J. P. Knight, “Genetic algorithms for optimization of
using the GA synthesis technique (GALOPS). The effects of                                 integrated circuits synthesis,” in Proc. 5th Int. Conf. Genetic Algorithms,
                                                                                          1993, pp. 432–438.
relaxing design constraints, such as area, have also been                            [15] M. S. Bright and T. Arslan, “A genetic framework for the high-level
analyzed and compared. (4) Analysis of the capability of a                                optimization of low power VLSI DSP systems,” Electron. Lett., vol. 32,
GA-based synthesis tool to present multiple solutions to a                                pp. 1150–1151, June 1996.
problem, exploiting the multiple solution nature of the GA                           [16] Bright, M.S. and Arslan, T., “Transformational-based synthesis of VLSI
                                                                                          based DSP systems for low power using a genetic algorithm,” in IEEE
search technique.                                                                         Int. Symp. Circuits and Systems, CA, May 1998, pp. 45–48.
                      III.    CONCLUSION                                             [17] Massoud Pedram, "Design Technologies for Low Power VLSI",
                                                                                          Encyclopedia of Computer Science and Technology, 1995
   The reduction in the power consumption is an important                            [18] T. A. Fjeldly and M. Shur. " Threshold voltage modeling and the
issue in modern portable personal computing and wireless                                  subthreshold regime of operation of short-channel MOSFET’s. " IEEE
                                                                                          Transactions on Electron Devices, 40(1):137–145, Jan. 1993.
communication systems. Minimizing the energy consumption                             [19] H. J. M. Veendrick. " Short-circuit dissipation of static CMOS circuitry
of VLSI Processors using multiple supply voltages with                                    and its impact on the design of buffer circuits. " IEEE Journal of Solid
Genetic Approach is a challenge task. Therefore, any method                               State Circuits, 19(8):468–473, August 1984.
                                                                                     [20] Ishikawa, Akitaka Okada, Nobuaki Kameyama, Michitaka," Low-Power
that provides a way to reduce this consumption must be                                    Multiple-Valued Reconfigurable VLSI Based on Superposition of Bit-
studied, evaluated and applied to the system in development.                              Serial Data and Current-Source Control Signals", Multiple-Valued Logic
The techniques reviewed from the literature presented in this                             (ISMVL), 2010 40th IEEE International Symposium on 26-28 May
                                                                                          2010 , pp: 179 – 184
survey promise an interesting way to achieve this issue. Also                        [21] Xin He Al-Kadry, S. Abdollahi, A.,"Adaptive leakage control on body
this literature study can be helpful for the further development                          biasing for reducing power consumption in CMOS VLSI
in the minimization of power consumption in future.                                       circuit",Quality of Electronic Design, 2009. ISQED 2009, pp: 465 – 470
                                                                                     [22] W. Hung, Y. Xie, N. Vijaykrishnan, M. Kandemir, M. J. Irwin and Y.
                                                                                          Tsai, "Total Power Optimization through Simultaneously Multiple-VDD
                              REFERENCES                                                  Multiple-VTH Assignment and Device Sizing with Stack Forcing",
[1]  Hariyama, M. Yamadera, S. Kameyama, M., "Minimizing energy                           Proceedings of the 2004 International Symposium on Low Power
     consumption of VLSI processors based on dual-supply-voltage                          Electronics and Design (ISLPED’04), pp: 144-149
     assignment and interconnection simplification", Circuits and Systems,           [23] Srivastava and D. Sylvester, "Minimizing total power by simultaneous
     2005. 48th Midwest Symposium on 7-10 Aug. 2005, pp: 1867 - 1870                      Vdd/Vth assignment," IEEE/ACM Asia-South Pacific Design
     Vol.2                                                                                Automation Conference, pp. 400-403, 2003.
[2] Hariyama, M. Aoyama, T. Kameyama, M., "Genetic approach to                       [24] D. Nguyen, A. Davare, M. Orshansky, D. Chinnery, B. Thompson, K.
     minimizing energy consumption of VLSI processors using multiple                      Keutzer, “Minimization of Dynamic and Static Power Through Joint
     supply voltages", Computers, IEEE Transactions on Issue Date : Jun                   Assignment of Threshold Voltages and Sizing Optimization,” ISLPED
     2005, Volume : 54 , Issue:6, pp: 642 – 650                                           2003.
[3] Mohanram, K. Touba, N.A., "Lowering power consumption in                         [25] Xiaoying Zhao, Jiangfang Yi, Dong Tong and Xu Cheng,"Leakage
     concurrent checkers via input ordering", Very Large Scale Integration                Power Reduction for CMOS Combinational Circuits",Micro-Processor
     (VLSI) Systems, IEEE Transactions on Nov. 2004, Volume : 12 ,                        Research & Development Center, Peking University, Beijing
     Issue:11, pp: 1234 – 1243
[4] Mohanty, S.P. Ranganathan, N., "Energy efficient scheduling for
     datapath synthesis", VLSI Design, 2003. Proceedings. 16th International
     Conference on 4-8 Jan. 2003, pp: 446 – 451
[5] Waidyasooriya Hasitha Muthumala, Masanori Hariyama and Michitaka
     Kameyama, "GA-Based Assignment of Supply and Threshold Voltages
     and Interconnection Simplification for Low Power VLSI Design", IEEE
[6] Kamble, M.B. and Ghose, K., "Energy-efficiency of VLSI caches: a
     comparative study", VLSI Design, 1997. Proceedings., Tenth
     International Conference on 4-7 Jan 1997, pp: 261 – 267
[7] Santosh Chede, Kishore Kulat and Rajesh Thakare, "A Significance of
     VLSI Techniques for Low Power Real Time Systems", 2010
     International Journal of Computer Applications (0975 - 8887), Volume 1
     – No. 22, pp: 86-91
[8] Flavius Gruian, “Hard real time scheduling for low energy using
     stochastic data and DVS processors,” in proc. of Int. symposium on Low
     Power Electronics and Design California, USA, 2001, 46-51.
[9] Jui-Ming Chang and Massoud Pedram. "Energy Minimization Using
     Multiple Supply Voltages", IEEE Transactions On Very Large Scale
     Integration (VLSI) Systems, VOL. 5, NO. 4, DECEMBER 1997, pp:
[10] Chandrakasan, M. Potkonjak, J. Rabaey, and R. W. Brodersen,
     “HYPER-LP: A system for power minimization using architectural
     transformations,” in Proc. IEEE Int. Conf. Computer-Aided Design,
[11] Marc S. Bright and Tughrul Arslan, "Synthesis of Low-Power DSP
     Systems Using a Genetic Algorithm", IEEE Transactions on
     evolutionary computation, vol. 5, no. 1, February 2001, pp: 27-41
[12] D. E. Goldberg, Genetic Algorithms in Search, Optimization and
     Machine Learning. Reading, MA: Addison-Wesley, 1989.

                                                                                                                        ISSN 1947-5500

To top