A Comparative Study of VLSI 3D Placement for power management and wirelength reduction

A Comparative Study of VLSI 3D Placement for power management and wirelength reduction

1T. Kapilachander,2 B. Vijayashree, 3 E. Rajthilak

1,3Assistant Professor, Panimalar Engineering College, Chennai,Tamil Nadu, India

2Assistant Professor, Arulmighu Meenakshi Amman College of Engineering, Kancheepuram,Tamil Nadu, India

Abstract In recent days every application must need power management and area management such physical problems and aspects of VLSI design. Minimization of power and area are highly complex integrated systems in microelectronics have led to the 3 Dimension developments a technological approach. 3 Dimension offers numerous advantages: Size, power consumption, hybrid integration etc., various techniques to handle the power management in IC. Power dissipation in a IC is base on power used by the IC and also by heat dissipation. To reduce energy use or to minimize heat dissipation some of the techniques are available. Three-dimensional (3D) integration is a viable approach that allows designers to add functionality to the devices while maintaining the same die area without the need for new design process. Stack dies in 3D integrated circuits (ICs) also reduces the die area of designs. In addition, die area reduction helps decrease wire length, thus improving the performance of the designs.

1. INTRODUCTION

   The task of the very large scale integration (VLSI) placement is to assign exact location to various circuit components within the chip area. It involves number of objective such as wire length, area of the die, timing and power. Placement treats the shapes of all blocks as fixed; i.e., it only determines the location of each block on the chip. The variables are the xy locations of the blocks; most blocks are standard cells the y-locations of cells are restricted to standard- cell rows. Placement instance sizes range in to the tens of millions and will continue to increase. Placement is usually divided in to two steps: global placement and detailed placement .Global placement assigns blocks to certain sub regions of the chip without determining the exact location of each component within its sub region. As a result, the blocks may still overlap. Detailed placement starts from the result of global placement, removes all overlap between blocks, and further optimizes the design. Placement objectives include the estimated total wire length needed to connect blocks in nets, the maximum expected wiring congestion in subsequent routing, and the timing performances of the circuit. To solve such large scale mixed size VLSI placement problem. Many algorithms are used to solve the placement problem in VLSI among these algorithm

   The placement algorithms are classified into constructive and iterative improvement methods. The constructive algorithm starts with the empty set and builds up the partition by adding one element at a time. These algorithms are faster but the quality of result is not good. An iterative algorithm starts with initial placements and repeatedly modifies it. These algorithms give good result but it takes long time.

   The placement problems are, the wire length and area of the die , routability, power minimization and delay . Out of these mention problem the area minimization and the wire

   length minimization are the most critical part. For the area and wire length optimization a modern placer needs to handle the large-scale design with millions of object, heterogeneous object with different sizes and various constrained placement such as preplaced blocks and chip density. The traditional approach in placement is to construct an initial solution by using constructive heuristic algorithms. A final solution is then produced by using iterative improvement techniques where a modification is usually accepted if a reduction in cost occurs, otherwise it is rejected. The solution generated by constructive algorithms may be far from optimal. Thus an iterative improvement algorithm is performed next to improve the solution the computation time of such algorithm is also large.

   This paper explains about the combination of techniques used for low power approach in integrated circuits (IC) or Chip and wire length minimization of a chip using 3D integrated circuits (ICs)

2. POWER MANAGEMENT

   Power consumption is an important issue the semiconductor ecosystem is trying to address. It concerns electronic product providers across all segments. For example the power consumption in the CMOS device is major concern with Static power , Dynamic power and Short Circuit Power. Mainly two components determine the power consumption in a CMOS circuit Static power consumption and Dynamic power consumption, since Short Circuit Power Consumption is rarely occurred one. So the total power consumption for the a CMOS device is given by

   Pdissipation = Pstatic +Pdynamic+Pshort circuit -2.1

   The power Consumption for a single CMOS is approximately 5V. InVLSI fabrication billions of gate fabricated assume that how much power we needed to operate a single Chip.From this above analysis we know the awareness of power reeducation needed for the Integrated Circuits(IC).

   Today, power and performance have become the predominant concerns for chip designers. Low power consumption is vital for handheld mobile and wireless devices as most of them are independent battery as the power source. Power management is becoming an increasingly urgent problem for almost every category of design, as power density (measured in watts per square millimeter) rises at an alarming rate.

   The static power dissipation is due to reverse saturation, sub-threshold and leakage current and occurs especially when the device is in idle mode. The dynamic power dissipation occurs due to charging and discharging of load capacitance and short-circuit current when both the NMOS and PMOS devices remain on for a short time duration

   Pswitching= afCeffVdd -2.2

   where a is switching activity, f is switching frequency, Ceff is effective capacitance and Vdd is supply voltage.

   Pshort-circuit = IscVddf — 2.3

   where Isc is short-circuit current during switching, Vdd is supply voltage and f is switching frequency. Dynamic power can be lowered by reducing the switching activity and clock frequency, which affects performance, and also by reducing the capacitance and supply voltage. Leakage power is a function of supply voltage Vdd, switching threshold voltage Vthand the transistor size

   PLeakage= f (Vdd, Vth, W/L) -2.4

   where Vddis supply voltage, V this threshold voltage, W is transistor width and L is transistor length

   The performance of the IC is based on the circuit speed In general, "small area" and "high performance" are two conflicting constraints. The IC designers' activities have been involved in trading off these constrain. Power dissipation issue was not design criterion but an afterthought. In fact, power considerations have been the Ultimate design criteria in special portable applications such as wrist watches and pacemakers for a long time. The objective in these applications was minimum power for maximum battery life time.

   There are so many numbers of ways to overcome this power consumption problem. In this further section of paper deals with the possible techniques available for the power management in an IC for long battery life .

3. METHODS OF POWER MANAGEMENT

   Due to the increased circuit density and speed, the power dissipation has emerged as an important consideration in circuit design. A lot of efforts on power reduction have been made at various levels of design bstraction. Considering the fact that the charging/discharging of capacitance is the most significant source of power dissipation in well-designed CMOS circuits

   Power consumption grows exponentially at 90nm and beyond technologies. At smaller geometries, aggressive management of leakage current can greatly impact design and implementation choices. Indeed, for some designs and libraries, leakage current exceeds switching currents, thus becoming the primary source of power dissipation in CMOS Depending upon the architecture of the design, the designers can choose from a wide range of options for reduction in power consumption of VLSI circuits. Some of the techniques to reduce the power consumption are briefed below.

   There are several methodology are available for low power management for IC. Some of them are listed below and briefly discussed (1)Logic restructuring (2)Clock tree optimization and clock Gating (3)Logic resizing (transistor resizing).(4)Transition rate buffering.(5)Pin swapping (6)Using multi-threshold voltage (7)Multi-supply voltage (voltage islands)(8)Dynamic voltage scaling (9)Dynamic voltage and frequency scaling (DVFS)(10)Power shutoff or power gating (11)Memory splitting (12)Substrate biasing (body-biasing or back-biasing)(13)Operand isolation

   From the above mention power management technique most of the popular used techniques are discussed and compare in this paper with respect to the power management issues in the 3D integration circuits

   (i)Dynamic voltage scaling (ii)Dynamic voltage and frequency scaling (DVFS )(iii)Power shutoff or power gating

   1. Dynamic Voltage Scaling

      Dynamic voltage scaling (DVS) is a standard technique for managing the power consumption of a system. It is based on the fact that the dynamic (switching) power P of CMOS circuits is strongly dependent on the core voltage V and the clock frequency f according to

      P f v2 3.1

      Under the assumption that the number of clock cycles required for a computation is independent of the core frequency, the execution time is inversely proportional to the frequency. The total energy E for the computation is then proportional to the square of the voltage:

      E v 2 — 3.2

      Note that the total energy for a computation is not depend on the frequency, but a reduced core voltage requires a reduction of the clock frequency and therefore implies a longer overall execution time. The assumptions behind Equation (2.8.2)are highly uncertain as they ignore other system components in particular the bus and memory. The other components impact the execution time of a program leading to a much more complex dependence on the processor frequency

   2. Dynamic Voltage and Frequency Scaling

      Dynamic Voltage and Frequency Scaling allows a host to dynamically switch its CPU frequency dependent on its load requirement. It do this by continuously monitoring the CPU utilization with the DVFS algorithm determining any necessary adjusts to the CPUs frequency with the goal being to run the CPU at a lower frequency so that it consumes less power. It uses processor performance states (P-states) presented to the VM kernel through an ACPI interface to achieve this. For example: If your 2GHz CPU is sitting at 30% utilization then DVFS will reduce the frequency of the CPU so it will operate nearer to its 600MHz frequency requirement including enough headroom to accommodate a sudden increase in CPU requirement.

   3. Power shutoff or power gating

      Power Gating is effective for reducing leakage power. Power gating is the technique in circuit temporarily turned off the sub blocks to reduce the overall leakage power of the chip. This temporary shutdown time can also call as "low power mode" or "inactive mode". When circuit blocks are required for operation once again they are activated to "active mode". These two modes are switched at the appropriate time and in the suitable manner to maximize power performance while minimizing impact to performance. Thus goal of power gating is to minimize leakage power by temporarily cutting power off to selective blocks that are not required in that mode. Power gating affects design architecture more compared to the clock gating. It increases time delays as power gated modes have to be safely entered and exited. The possible amount of leakage power saving in such low power mode and the energy dissipation to enter and exit such mode introduces some architectural trade-offs. Shutting down the blocks can be accomplished either by software or hardware. Driver software can schedule the power down operations. Hardware timers can be utilized. A dedicated power management controller is the other option.

      An externally switched power supply is very basic form of power gating to achieve long term leakage power reduction. To shutoff the block for small interval of time internal power gating is suitable. Power gating uses low-leakage PMOS transistors as header switches to shut off power supplies to

      parts of a design in standby or sleep mode. NMOS footer switches can also be used as sleep transistors. Inserting the sleep transistors splits the chip's power network into a permanent power network connected to the power supply and a virtual power network that drives the cells and can be turned off.

      The quality of this complex power network is critical to the success of a power-gating design. Two of the most critical parameters are the IR-drop and the penalties in silicon area and routing resources. Power gating can be implemented using cell- or cluster-based approaches or a distributed coarse-grained approach.

   4. Power-gating parameters

      Power gating implementation has additional considerations than the normal timing closure implementation. The following parameters need to be considered and their values carefully chosen for a successful implementation of this methodology .

      • Power gate size

      • Gate control slew rate:

      • Simultaneous switching capacitance

      • Power gate leakage

4. PLACEMENT PROBLEM DESCRIPTION

   (2) E = i) i = 1,,M} U ) i = 1,,M} U

   {(bi,bj) bi bj}

   If existing, edge (bi,bi+1) edge (bi+1,bi+2) and edge

   (bi,bi+2) then (bi,bi+2) omitted

   3) Vertex Weight equals the width of the block bi, but zero for Sh and th similarly the vertical constrain graph (VGH)

   Vertical constrain graph Gv(V,E) is constructed using above constrain and the height of the each block. The corresponding constrain graph Gh(V,E) and Gv(V,E), Both Gh(V,E) and Gv(V,E) are vertex weighted acyclic graph so longest path algorithm can be applied to find the x and y coordinates of each block. The coordinates of the block coordinate of the lower left corner of the block.

   1. Wire length estimation

      We are addressing the problem of VLSI standardCell placement with the objectives of minimizing wire length , power consumption, and timing performance (delay), while considering the layout width as a constraint.

   2. Wire length measurement and Cost

   In VLSI placement the cells present in the module are connected by wire. The estimation of wire length [2] required for connection is calculated by the formula

   Wire length = wi,j ((xi xj)2 + (yi – yj)2) I > j

   1. Area Estimation

      Given a set of module M1, M2,..Mn and set of n interconnects N1,N2Nn the objective of the placement is to obtain the non overlapping package of all the modules which achieves some optimization objective such as minimizing the area of package[3], the interconnection length as show in the below figure 1.

      4

      0

      2

      5

      1

      3

      H

      Figure 1. Placement Area Estimation

      Horizontal constrain:

      If (-X,Y) =(. . . a. . .b. . .,a. . . b. . .) Block b is is at right side of the block a

      Vertical constrain:

      If (X,Y) =(. . . a. . .b. . .,b .. . a. . .) Block b is at the below side of the block a

      Basd on "left of' constraint of (X, Y), a directed and vertex-weighted graph GH (V, E) (V: vertex set, E: edge set), called the horizontal-constraint graph (HCG) is constructed as follows:

      (1) V= {Sh} U {th} {bi i = 1, . . . , M},

      Where bi correspond to the block

      Sh is the source node representing the left boundary

      th is the target node representing the right boundary

      Where

      wi,j weight of the connection between cell x and y

      (xi xj) distance in X direction

      (yi – yj) distance in Y direction

      Inter connect wire length of each net in the circuit is estimate during Steiner tree and then total wire length is computed by adding the individual estimates

      Cost wire = l i

      i M

      Where l is the wire length estimation for net i and M denotes total number of nets in circuit.

5. PLACEMENT ALGORITHMS

   The philosophy of our tool follows that of its 2D counterpart, VPR .The flow of the TPR placement and routing CAD tool is shown in Figure 2. The placement algorithm first employs a partitioning step using the hMetis algorithm to divide the circuit into a number of balanced partitions, equal to the number of tiers for 3D integration. The goal of this first min-cut partitioning is to minimize the connections between tiers, which translates into reducing the number of vertical (i.e., inter-tier) wires and decreasing the area overhead associated with 3D switches as discussed in the previous section. After dividing the netlist into tiers, TPR continues with the placement of each tier using a hybrid approach that combines top-down partitioning and simulated annealing. The annealing step moves cells mostly within tiers. Finally, the cells are routed to obtain a placed and routed solution. The following sections describe these steps in more detail.

   Figure 2. Flow of 3D TPR Tool

   1. Partitioning the Circuit Between Tiers: The TPR step that performs partitioning and tier assignment of the circuit is shown conceptually in Figure 3. After the netlist is partitioned using hMetis, a novel linear placement approach is used to arrange the tiers such that wire length and the maximum cutsize between adjacent tiers is minimized. This is achieved by mapping this problem to that of minimizing the bandwidth of a matrix3 , using an efficient matrix bandwidth minimization heuristic.

      Figure 3. Partitioning of netlist in to Tires

   2. Partitioning-based Placement Within Tiers: After the initial tier assignment, placement is performed on each tier starting with the top tier, proceeding tier after tier. The placement of every tier is based on edge-weighted quad- partitioning using the hMetis partitioning algorithm, and is similar to the approach in which has the same quality as VPR but at 34times shorter run times. Edge weights are usually computed inversely proportional to the timing slack of the corresponding nets. However, we also selectively bias weights of the most critical nets. The set of critical nets is comprised of edges on the current k-most critical paths. In order to improve timing, the bounding box of the terminals of a critical net placed on a tier is projected to the lower tiers and used as a placement constraint for other terminals. More details of the partitioning based placement phase can be found

   1. Transformations for 3D Placement

      The transformation schemes from a 2D placement to a 3D placement. Initial optimized 2D placement is done on a chip die of the 3D chip, where K is the number of dies, so that the total area of the 2D K times larger than a single placement is equal to that of the target 3D chip. Given this 2D placement with minimized wirelength, local stacking transformation can achieve even shorter wirelength for the same netlist with 3D integration. We also present two folding-based transformation schemes, folding-2 and folding-4, which can generate 3D placement with a very low TSV number. Moreover, TSV number and wirelength tradeoffs can be achieved by the

      window-based stacking/folding. All these transformation methods can guarantee wirelength reduction over the initial 2D placement.

   2. Local Stacking Transformation

      Local stacking transformation (LST) consists of two steps, stacking and legalization, as shown in Figure 4. The stacking step shrinks the chip uniformly but does not shrink cell areas so that cells are stacked in a region K times smaller and remain as the original relative locations. The legalization step minimizes maximum on-chip temperature and TSV number through the position assignment of cells. The result of LST is a legalized 3D placement.

      Figure 4. Local Staking Transformation

   3. Transformation through Folding

   LST achieves short wire length by stacking the neighboring nodes together. However, a great number of TSVs will be generated when the nodes of local nets are put on top of one another. If the target 3D integration technology only allows a limited TSV density, we need to use the transformations that generate fewer TSVs. Folding-based transformation is to fold the original 2D placement like a piece of paper without cutting off any parts of the placement. The distance between any two nodes will not increase and the total wire length is guaranteed to decrease. TSVs are only introduced to the nets crossing the folding lines With an initial 2D placement of minimized wire length, the number of such long nets should be fairly small, which implies that the connections between the folded regions should be limited, resulting in much fewer TSVs (compared to that of the LST transformation, where many dense local connections cross different dies). shows one way of folding, named folding-2, by folding twice at both x and yanother way of folding, named folding-4, by folding twice at both x and y directions. The folding results are legalized 3D placements, so no legalization step is necessary. After folding-based transformations, only the lengths of the global nets that go across the folding lines (dotted lines in Figure 5) get reduced. Therefore, folding-based transformations cannot achieve as much wire length reduction as LST. Furthermore, if we want to maintain the original aspect ratio, folding-based transformations are only applicable to the 3D design with an even number of dies.

   Figure 5. Folding Based Transformation schemes

6. CONCLUSION

By using the transformation based 3D placement algorithm. The components with a dashed boundary are the exiting tools that we use. A 2D wire length thermal driven placer, force driven placement or memetic algorithm based placer is first used to generate a 2D placement for the target design. the quality of the final 3D placement highly depends on the initial placement . the 2D placement is then transform in to legalized 3D placement according to the already discussed 3D technology in the above paper during the transformation wire length , TSV number must be consider. A refinement process through die reassignment will be carried out after 3D transformation to further reduce the TSV number and bring down the maximum on chip temperature. Finally a 2D detailed placer will further refine the placement result for each die

As technology shrinks feature sizes, reliability will dominate as a design consideration .Die area (manufacturing cost) and performance (heavily influenced by cycle time) are important basic design considerations. Power consumption has also come to the fore as a design limitation. As technology shrinks feature sizes, reliability will dominate as a design consideration. the sizes of die decrease in future , so do device sizes. Smaller device sizes result in reduced capacitance. Decreasing the capacitance decreases both the dynamic power consumption and the gate delays. As device sizes decrease, the electric field applied to them becomes destructively large. To increase the device reliability, we need to reduce the supply voltage V. Reducing V effectively reduces the dynamic power cosumption, but results in an increase in the gate delays. We can avoid this loss by reducing Vth. On the other hand, reducing Vth increases the leakage current, and therefore, the static power consumption. Determining performance is relatively straightforward when compared with the determination of overall cost. A good design achieves an optimum costperformance tradeoff between the power and area . This determines the quality of a processor design

1. Bo Hu and Malgorzata Marek-Sadowska, Multi level Fixed-Point- Addition-Based VLSI Placement; IEEE Transactionson Computer Aided Design Of Integrated Circuits And Systems, Vol.24, No.8,

   August2005

2. Ulrich Brenner and Jens Vygen, Legalizing a Placement with Minimum Total Movement; IEEE Transactionson Computer Aided Design Of Integrated Circuits Andsystems, Vol.23, No.12, December 2004

3. Shawki Areibi and Zhen Yang; Effective Memetic Algorithms for VLSI Design = Genetic Algorithms+ Local Search+ Multi-Level Clustering; in 2003 by the Massachusetts Institute of Technology Evolutionary Computation

4. Fred Glover and Gary A. Kochenberger, Handbook of Metaheuristics, kluwer academic publishers, New York, Boston, Dordrecht, London,

   Moscow @ 2003

5. Elizabeth M.Rudick, Pinaki Mazumder, Genetic Algorithm for VLSI Design Layout and Test Automation, Pearson Edition @ 1999

6. Sabih H.Gerez, Algorithm for VLSI Design Automation, Jhon Wiley student Edition @ 1998

7. Volker Schnecke and Oliver Vornberger; Hybrid Genetic Algorithms for Constrained Placement Problems; in IEEE 1998, appeared in IEEE Transactionson Evolutionary Computation Vol.1, No.4, November1997, pp.266277

8. K. Shahookar and P. Mazumder, Department Of Electrical Engineering And Computer Science, University Of Michigan, Ann Arbor, Michigan 48109, Vlsi Cell Placement Techniques, ACM Computing Surveys, vol. 23, no. 2, june 1991

9. Transistor and pin reordering for gate oxide leakage reduction in dual circuit

10. J.Kaoetal Transistor sizing issues and tool for multi threshold CMOS technology in proceedings of ACM/IEEE DAC pp.409-414,JUN 1997

11. QqiangMQ & Evangeline Voltage island Driven floor planning in proceeding of IEEE pp644-649 2007

12. Stephen, Mike, Hook Enhanced multi threshold (MTCMOS) circuit usinh variable well bias in ISPED01/ACM pp165-169 August 2001

13. Chang, Cheng Post layout logic restructuring using alternative wires in Proceedings of IEEE pp587-596 August 1996.

14. W.Chen, C.T Hsieh, M.Pedram Gate sizing with controlled Displacement in Proceedings of international symposium on physical design pp127-132 1999

15. O.Coudert, R.Haddad, New algorithm for gate sizing comparative study in proceedings of 33 Design Automation Confrence pp734-789 Jun 1996

16. M.Hashimoto, HiOnodera & K.Tumara A practical gate resizing technique considering glitch reduction for low power desing in proceedings of Design Automation Conference 1999

17. J.M.Cheng & M.Pedram Energy minimization usig multi supply voltage IEEE Transaction on VLSI System Vol 5: no: 4 pp 1- 8 December 1992

18. C.Chen & M.Sarrafzadeh A Efficient algorithm for gatelevel power delay trade of using two voltages International conference on computer Design pp222-227 October 1999

19. Qing Wu, M.Pedram & Xunwei Clock gating and its application to low power design of sequential circuits in IEEE Transction pp415-420 in Vol-47 no103 March 2000

20. Oliver Coudert, Gate Sizing for Constrained delay/power/area optimization in IEEE Transcation on VLSI Design September 1997

21. P.J.Shah, P.Patil, V.M.Deshmukh & P.H.Zope. Low Power VLSI Design using Dynamic Thersholdlogic in proceedings of SPIT IEEE Colloquium international Conference Mumbai Vol-2.121

    	REFERENCE		technology with multi threshold voltage CMOS in IEEE Journel of
    [1]	Chen1, and Yao-WenChang1'2 Tung-Chieh Chen2, Jiang1, Hsin-Chen		solid state circuits Vol-30 August 1995
    	Zhe-Wei; challenges and solution in modern VLSI placement 2007	[27]	Kai Chang Wu, Diana Marculescu,Joint logic restructuring and pin
    	IEEE 1-4244-0583-1/07		reordering against NBTI-induced performance degradation.
    [2]	Sadiq M. Sait, Mustafa Imran Ali, and Ali Mustafa Zaidi; Evaluating		Proceedings of the Conference on Design, Automation and Test in
    	Parallel Simulated Evolution Strategies for VLSI Cell Placement; in 1-		Europe European Design and Automation Association 3001 Leuven,
    	4244-0054-6/06 ©2006 IEEE		Belgium, ©2009 ,ISBN: 978-3-9810801-5-5
    [3]	Ning Xu,Feng Huang ,Zhonghua Jiang ; A Fast Algorithm for VLSI	[28]	Roy K., Prasad S.C., Circuit activity based logic synthesis for low
    	Building Block Placement; in IMACS Multiconference on		power reliable operations, IEEE Transactions on Very Large Scale
    	Computational Engineering in Systems Applications(CESA), October		Integration (VLSI) Systems, Volume 1, Issue 4, Dec. 1993, pp 503
    	4-6, 2006, Beijing, China.		513
    [4]	Natalio Krasnogor, and Jim Smith; A Tutorial for Competent Memetic	[29]	Monica Donno, Alessandro Ivaldi, Luca Benini Enrico Macii, Clock-
    	Algorithms: Model, Taxonomy and Design issues; IEEE		tree power optimization based on RTL clock-gating, Proceedings of the
    	Transactionson Evolutionary Computation, Vol.A, No.B, CCC200D		40th annual Design Automation , ACM New York, NY, USA ©2003, ISBN:1-58113-688-9

22. S.I.Mutuh, T.Douseki, I-V power supply high speed digital circuit

1. Weixiang Shen; Yici Cai; Xianlong Hong; Jiang Hu; An Effective Gated Clock Tree Design Based on Activity and Register Aware Placement , IEEE Transactions on Very Large Scale Integration (VLSI)

   Systems, Volume: 18 Issue:12, page(s): 1639 – 1648 ISSN: 1063-8210

2. Seongmoo Heo and Krste Asanovi´ c Dynamically Resizable Static CMOS Logic for Fine-Grain Leakage Reduction, MIT CSAIL Technical Report (MIT-LCS-TR-957), July 2004.

3. A. Khan, P. Watson, G. Kuo et all. A 90nm Power Optimization Methodology and its Application to the ARM1136JF-S Microprocessor, 2005 IEEE Custom Integrated Circuits Conference,

   San Jose, CA.September 21, 2005

4. Le, H.-B. Do, X.-D. Lee, S.-G. Compact low-power high-slew-rate buffer for LCD driver application,IEEE Electronics Letters ,Volume: 46 Issue: 20 ,On page(s): 1370 – 1371 ISSN: 0013-5194

5. Yanbin Jiang; Sapatnekar, S.S.; Bamji, C.; Juho Kim; Interleaving buffer insertion and transistor sizing into a single optimization IEEE Transactions on Very Large Scale Integration (VLSI) Systems, Volume: 6 Issue:4 On page(s): 625 633 ISSN: 1063-8210

6. T.Kapilachander, I.Hameem Shanavas, V.Venkataraman,Technical Study on Low Power VLSI methods I.J. Information Engineering and Electronic Business, 2012, 1, 60-70

7. Guojie Luo, Placement and Design Planning for 3D Integrated Circuits, 2011