Instruction Set Extension of a Low-End Reconfigurable Microcontroller in Bit-Sorting Implementation

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 8, No. 4, August 2018, pp. 2595~2601
ISSN: 2088-8708, DOI: 10.11591/ijece.v8i4.pp2595-2601  2595
Journal homepage: http://iaescore.com/journals/index.php/IJECE
Instruction Set Extension of a Low-End Reconfigurable
Microcontroller in Bit-Sorting Implementation
Sani Irwan Md Salim, Yewguan Soo, Sharatul Izah Samsudin
Centre of Telecommunication Research and Innovation, Faculty of Electronic and Computer Engineering,
Universiti Teknikal Malaysia Melaka, Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia
Article Info ABSTRACT
Article history:
Received Mar 12, 2018
Revised May 21, 2018
Accepted May 28, 2018
The microcontroller-based system is currently having a tremendous boost
with the revelation of platforms such as the Internet of Things. Low-end
families of microcontroller architecture are still in demand albeit less
technologically advanced due to its better I/O better application and control.
However, there is clearly a lack of computational capability of the low-end
architecture that will affect the pre-processing stage of the received data. The
purpose of this research is to combine the best feature of an 8-bit
microcontroller architecture together with the computationally complex
operations without incurring extra resources. The modules’ integration is
implemented using instruction set architecture (ISA) extension technique and
is developed on the Field Programmable Gate Array (FPGA). Extensive
simulations were performed with the and a comprehensive methodology is
proposed. It was found that the ISA extension from 12-bit to 16-bit has
produced a faster execution time with fewer resource utilization when
implementing the bit-sorting algorithm. The overall development process
used in this research is flexible enough for further investigation either by
extending its module to more complex algorithms or evaluating other designs
of its components.
Keyword:
Instruction Set Extension
Reconfigurable Processor
Bit-Sorting
Copyright © 2018 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Sani Irwan Md Salim,
Centre of Telecommunication Research and Innovation,
Faculty of Electronic and Computer Engineering,
Universiti Teknikal Malaysia Melaka,
Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia.
Email: sani@utem.edu.my
1. INTRODUCTION
Embedded systems are rapidly evolving with numerous applications now are executed on its various
platforms. The embedded system variant has evolved from the traditional PC-based integrated system to
stand-alone and self-sustaining system with equivalent processing power and features compare to the former.
As the system becomes more ubiquitous, the design process for the embedded system hardware must be in
line with its software applications counterpart in order to provide the most streamlined and satisfying
experience for the end-users. To achieve this, the designer must consider several performance parameters that
ultimately lead to the introduction of a cluster of hardware called the reconfigurable system.
The relevance of 8-bit microcontroller has been the subject of intense debate within the computing
community. Many predicted since the 90’s that the decline of the 8-bit microcontroller is imminent and all
applications would be based on 16-bit architecture [1] and later 32-bit architecture [2] with the emergence of
the ARM processor. However, despite all things happened in the world of technology, the 8-bit
microcontrollers are still in demand and being adopted in wide-ranging applications. For example, the
Microchip’s 8-bit device has the ability to maintain functionality with voltage varies from low-level 1.3V to
TTL 5V [3] which is beneficial due to the long sleep-mode required for most of the IoT device. It is

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 8, No. 4, August 2018 : 2595 – 2601
2596
important to note that most of the data processing tasks in an IoT system are executed in the cloud server
while the IoT board only responsible for data acquisition. The IoT board is connected to various sensors and
minimal pre-processing tasks are required before the data are relayed to the cloud server through the internet
connection. As the IoT device transmits its data serially in a protocol, having an optimized 8-bit architecture
would suit the requirement of the IoT board better than the 32-bit architecture, without all the unnecessarily
peripherals that are integrated into the 32-bit architecture.
A soft-core processor is a microprocessor fully described in software which can be synthesized in
programmable hardware, such as FPGAs. Soft-core processors implemented in FPGAs can be easily
customized to the needs of a specific target application. A microprocessor defined in software code usually in
hardware description language (HDL). As the processors are available in programming code, the developers
are able to modify and reuse pre-designed hardware components in the form of intellectual property (IP)
cores. This method would reduce design time at the expense of area and performance penalty.
In general, having a soft-core processor provide several advantages in embedded system
development. Soft-core processors have a higher level of abstraction that makes it easier to understand.
Flexibility is also offered to the developer that enable the core to be edited and changed the parameters.
Notable applications of the soft-core processor are in motor control system [4], pulse width modulation
(PWM) generation [5] and cryptographic algorithm [6]. However, there is very little detail of the
implementation and choices made during the development process i.e. design decisions and performance
trade-off [7], [8]. There is the need to established a methodology that promotes incremental design
improvement for a system that is based on the soft-core processor.
UTeMRISC0 [9, 10] is a softcore processor that is essentially based on the PIC architecture, that is
also the fundamental architecture of several other soft-core processors previously mentioned in this thesis.
The UTeMRISC0 soft-core processor was first introduced in [11] as a part of the experimental setup to
maximise the capability in delivering highest performance achievable by a processor core compared to their
physical IC counterpart. The previous PIC-based soft-core processors are heavily reliant to the MPlab to
compile and generate the processor’s bit file. For UTeMRISC0 processor, a retargetable assembler is
designed exclusively from the ground up to replace the MPlab as the de-facto assembler. In the early phase,
the assembler is designed to match the existing instruction set in the PIC data sheet within a simulated
computer architecture environment called the CPUSim.
One of the prominent paper with regards to the instruction set extension on reconfigurable
processors was published by [12]. The overall target for manipulating the instruction set is to achieve higher
parallelism to the instruction execution. Over the years, different techniques to increase the level of
parallelism have been introduced at instruction level: for instance, techniques such as instruction pipelining,
superscalar execution, out-of-order execution, and register renaming. Parallelism has also been exploited at
other levels: bit-level, data-level, and loop-level parallelism. Although the level of parallelism has been
increased over the years, it is still relatively limited for highly parallelizable applications, which become poor
candidates for implementation of these architectures.
2. RESEARCH METHOD
The instruction set extension technique applied in this research is not necessarily focused on adding
new instructions to the instruction set [13]. It is also could encompass modification to the current instruction,
altering the behavior of the available instruction to perform a command that is totally different from its
original purpose and also omitting the unrelated instructions that are not involved in the specific application.
These customizations are done through manual configuration [14].
Manual extensions involved a degree of human effort to identify and implement the instruction set
extensions. Human ingenuity in the manual creation of custom capabilities creates high-quality results which
within the context of the application’s result. As the design grows more complex and constraint on the time-
to-market requirement, automatic design flow is desirable for the use of these new capabilities. For a large set
of instruction candidates, selection of multiple custom instructions involves complex trade-off and can be
hard to execute manually. However, as this research is focusing on low-end digital signal controller platform,
the custom instructions that are manually identified and created. These instructions are considered sufficient
to perform DSP tasks successfully within the hardware limitation.
2.1. 12-bit ISA to 16-bit ISA
Using the original 12-bit ISA [15] as the reference point, the ISA is extended to become 16-bit in
width that in the end becomes the configured ISA for the UTeMRISC01 microcontroller. The customized
instruction set is created by expanding the available instruction tokens to vacate more space to execute more
data-transfer tasks such as register swapping. Figure 1 shows the instruction format for both the 12-bit ISA

Int J Elec & Comp Eng ISSN: 2088-8708 
Instruction Set Extension of a Low-End Reconfigurable Microcontroller in Bit-Sorting … (Sani Irwan)
2597
and the 16-bit ISA with three different modes; (1) Literal operation, (2) byte-oriented operation with
direction and (3) bit-oriented operations.
The new 16-bit ISA now comprised of 6-bit of the opcode and 10 bits of the operand. The total
number of 16 is chosen as it is byte addressable and could easily accommodate more bits for future
expansion. Opcode bit is set to 6 bits that could stretch the maximum number of instructions to up to 64
instructions. The file register address is extended to 7 bits that enable theoretically up to 128 registers that
can be addressed by each instruction. The literal value maintained at 8-bit of raw data with a 3-bit select bit
to point to any position in the 8-bit data. Finally, the single directional bit is used to indicate the destination
of the data either to the working register or the register that is addressed in the operand. Technically, the
16-bit ISA provide the best setup for the low-end digital signal controller whereby there a lot of room to
exploit in adding new instructions, expanding the number of registers and also offered extra bits for literal
values for precision and sign extended data.
(1)
(2)
(3)
(a) (b)
Figure 1. Conversions from 12-bit ISA in (a) to 16-bit ISA in (b) for all instruction types
Compiler generator or retargetable code generator is used to synthesize code for the particular
applications. Once the instruction set is generated, the retargetable assembler is developed to accommodate
all changes made in the instruction set. The assembler then is responsible to decode and generate the object
code of the program based on the new instruction set [16]-[18].
2.2. Workflow and related tools
Specific tasks and procedures are followed to ensure smooth transactions as well as getting the right
tools. The UTeMRISC01 processor is implemented in the Xilinx ML605 FPGA board and the retargetable
assembler is developed from the ground up using the Visual Basic software by Microsoft. Once
implemented, the performance parameters such as resource utilization and the execution times are obtained
through the synthesis and Post-PAR timing report. Further measurements are also conducted using the logic
analyzer to verify the execution times of the bit-sorting program.
During the early phases of the design, the new instruction is tested and simulated using the CPUSim
software. CPU Sim is a Java application that allows users to design simple computer CPUs at the microcode
level and to run the machine-language or the assembly-language programs on those CPUs through
simulation [19].
The tools used in the hardware implementation phase are exclusively done on Xilinx environment,
including its Integrated Synthesis Environment (ISE) Design Suite. The Xilinx ISE Design Suite is
responsible for the bulk of the work by providing the tools to develop HDL modules in the processor
architecture. The behavioral simulations are completed by using the ISE simulators, also known as ISim,
which its main purpose is to perform functional and timing simulations for Verilog designs. Xilinx ISE will
perform the logic synthesis and design implementation to generate the bit file. Using a software called
iMPACT (integrated with the Xilinx ISE), the bit file is configured and loaded to the FPGA board, in this
case, the ML605 FPGA board).
2.3. Bit-sorting algorithm
There are various methods of sorting, such as interchange or bubble sort, shell sort, bucket sort and
radix interchange sort. Bubble sort is one of the popular methods, although the execution is not very efficient.
The key idea of bubble sort is to take adjacent pairs of elements in the list of data and put them in order by
interchanging their positions when it is necessary. By continuously executing the interchanging operations,
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
(a) x x k k k k k k k k
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
(b) f f f f f f f x x d
direction
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
(c) f f f f f f f b b b
opcode
file register (f)
bit (b)file register (f)
opcode literal (k)
opcode

 ISSN: 2088-8708
2598
the data are sorted ascendingly within the data range. Figure 2 shows the step-by-step implementation flow of
the bit sorting algorithm.
Start
Set numbers
to be sorted
Set the starting
address
Compare value with
the adjacent address
Less than?
Set next
number
Swap value and
interchange address
End of entire array?
Re-check each
array value?
End
N
Y
Y
N
Y
N
Figure 2. Flowchart for bit-sorting algorithm
3. RESULTS AND ANALYSIS
Establishing the benchmark of the soft-core processor is vital at this stage. The performance
parameters for the modified architecture are evaluated and compared with the performance of the original
processor architecture. In this case, the 8-bit soft-core processor called UTeMRISC0 resemble greatly to its
packaged-IC counterpart which is the PIC16C57. The UTeMRISC0 will be the starting point for the
architecture expansion and modification.
The instruction set architecture extension is performed on the UTeMRISC0 architecture. Originally,
the ISA consisted of 12-bit and is expanded to become 16-bit. The expansion does not solely involve the
lengthening of the ISA register. It also required further modifications to other modules and registers such as
the instruction decoder module, the ROM module and instruction registers. New opcodes are also assigned to
the instruction set. At this point, the original list of instruction is maintained as per original while a single
new instruction called ‘swapfw’ is created to demonstrate the capability of the architecture in handling and
executing the new instruction. To reflect the major changes being made due to the ISA expansion, the
modified processor now is called UTeMRISC01 [20], [21].
3.1. Instruction set generation
A new instruction is created and called ‘swapfw’ that would take both of its operands as register
address. To optimize the instruction set, the ‘swapfw’ instruction would occupy the space for another
instruction called ‘movf f,1’. Originally, this instruction only functions to transfer data from a register back to
its own address which it impractical unless a certain condition bit is required to be generated. Therefore, the
new ‘swapfw’ instruction is assigned to the previous opcode allocation but now with different function, as
shown in Table 1.

2599
Table 1. New Instruction for Bit-Sorting
CPU Sim
Format 16-bit Opcode
Opcode Mnemonics
0B swapfw f 6 7 -3 0010_11ff_ffff_fxxx
3.2. Code synthesis
Referring to Figure 3, the new instruction the is simulated using the CPUSim software with the
UTeMRISC01 architecture is loaded as the main processor execution. The new instruction’s micro-operation
is defined in this process to iterate the sequence of the instruction execution.
Figure 3. Machine instruction editing in CPUSim
3.3. Hardware synthesis
The findings from the code synthesis stage are brought forward to the FPGA implementation by
replicating the instruction execution sequence, this time in Verilog code. The instruction’s opcode and ISA
are updated in the instruction decoder and ALU module. Once the architecture is successfully synthesized
and implemented, the performance parameters are observed and measured using the logic analyzer.
When the UTeMRISC01 architecture is implemented for the bit-sorting program, an increase of
registers and LUTs have been recorded. As the UTeMRISC01 are equipped with the new ISA and an
improved instruction register, the increase in the slice register number is expected, as shown in Table 2.
Further optimization of the ALU module, especially in reducing the redundant instructions also contributed
to a smaller usage of slice LUTs. For the bit sorting program, the new instruction is introduced by
overwriting an unused instruction and this ensures the overhead of resource utilization is kept to a minimum.
The new ‘swap’ instruction only involved register data transfer, which is already embedded in the original
architecture, hence there is no additional hardware is required to execute the instruction.
Table 2. Resource Utilization for Bit-Sorting Program
Parameter
UTeMRISC0
(12-bit ISA)
UTeMRISC01
(16-bit ISA)
Difference
Number of Slice registers 102 119 16.7%
Number of Slice LUTs 457 355 22.3%
Number of LUT Flip Flop pairs used 494 409 17.2%
In terms of execution times, the UTeMRISC01 architecture consistently produced lower execution
times throughout the bit-sorting program implementation. Therefore, the program is ended much quicker, as
shown in Figure 4. By calculations, the UTeMRISC01 architecture could finish executing the same bit-
sorting program faster than the UTeMRSIC0 architecture by 12%. This is clearly due to the lower instruction
cycle, resulted in the introduction of the new swap instruction.

 ISSN: 2088-8708
2600
Figure 4. Execution times for bit-sorting program
4. CONCLUSION
In the bit-sorting programs, the number of registers in the UTeMRISC01 is reduced due to several
optimization steps are done to the original UTeMRISC0 architecture. There is no new module added to the
architecture and the new instruction is introduced by occupying the existing instruction slot in the instruction
set. The execution times also improved that makes the program finished earlier. The reduced time is not
because of the less-complexity of the bit-sorting program or the architecture itself but it is due to the
efficiency of the new instruction that replaced the repetitive instruction previously existed in the original
architecture. It is demonstrated here that the adoption of instruction set extension technique on the
UTeMRISC01 architecture has optimized the processor capability in dealing with a more complex task. The
overall development process used in this research is flexible enough for further investigation either by
extending its module or to adopt more complex algorithms in the future. While the focus of this research is to
reduce the execution times and the reduce utilization, power consumption will be considered in the future
iterations of the processor.
ACKNOWLEDGEMENTS
The author would like to thank Universiti Teknikal Malaysia Melaka and Ministry of Higher
Education Malaysia for the financial support through the research grant number PJP / 2016 / FKEKK-CETRI
/ S01496.
REFERENCES
[1] R. Cravotta. (2012, 11 July). One Processor to Rule Them All? Available: http://www.edn.com/design/systems-
design/4398890/One-processor-to-rule-them-all
[2] J. Ganssle. (2012, 2 July). Is 8-bits dying? Available: http://www.embedded.com/electronics-blogs/break-
points/4389890/Is-8-bits-dying-
[3] Microchip, "8-bit PIC® Microcontroller Solutions," Microchip Technology Incorporated 2014.
[4] T. Sutikno, N. R. N. Idris, A. Z. Jidin, and A. Jidin, "A Model of FPGA-based Direct Torque Controller,"
Indonesian Journal of Electrical Engineering and Computer Science, vol. 11, pp. 747-753, 2013.
[5] A. Zemmouri, R. Elgouri, M. Alareqi, M. Benbrahim, and L. Hlou, "Design and Implementation of Pulse Width
Modulation Using Hardware/Software MicroBlaze Soft-Core," International Journal of Power Electronics and
Drive Systems (IJPEDS), vol. 8, pp. 167-175, 2017.
[6] C. Kannan, "NIOS II Based Secure Test Wrapper Design for Testing Cryptographic Algorithms," International
Journal of Reconfigurable and Embedded Systems, vol. 4, 2015.
[7] F. Plavec, B. Fort, Z. G. Vranesic, and S. D. Brown, "Experiences with Soft-Core Processor Design," in IPDPS,
2005, p. 167.2.
[8] F. Plavec, "Soft-core processor design," Master of Applied Science, Graduate Department of Electrical and
Computer Engineering, University of Toronto, 2004.
[9] A. J. Salim, S. I. M. Salim, N. R. Samsudin, and Y. Soo, "Conversion of an 8-bit to a 16-bit Soft-core RISC
Processor," International Journal of Electronics Communication and Computer Technology, vol. 3, pp. 393-397,
2013.
[10] N. R. Samsudin, S. I. M. Salim, and A. J. Salim, "Designing UTeMRISCII Processor for Multiply-Accumulate
Operation," in 3rd International Conference on Engineering and ICT (ICEI2012), 2012, pp. 88-91.
[11] L. E. Yong and A. J. Salim, "Implementation of an 8-bit RISC Microcontroller Chip," in 4th International
Symposium on Broadband Communication, 2010, pp. 1-4.

2601
[12] C. Galuzzi and K. Bertels, "The Instruction-Set Extension Problem: A Survey," ACM Trans Reconfigurable
Technol. Syst., vol. 4, pp. 1-28, 2011.
[13] A. J. Salim, S. I. M. Salim, N. R. Samsudin, and Y. Soo, "Instruction Set Extension Through Partial Customization
of Low-End RISC Processor," Australian Journal of Basic and Applied Sciences, vol. 7, pp. 678-687, 2013.
[14] D. Liu, Embedded DSP Processor Design, : Application Specific Instruction Set Processors: Morgan Kaufmann,
2008.
[15] T. Coonan. (1999, 4 January). RISC8 Verilog Core. Available:
http://pldworld.info/_hdl/2/_ip/mindspring/~tcoonan/risc8doc.html
[16] S. I. M. Salim, H. A. Sulaiman, R. Jamaluddin, L. Salahuddin, M. N. S. Zainudin, and A. J. Salim, "Two-pass
assembler design for a reconfigurable RISC processor," in IEEE Conference on Open Systems (ICOS), 2013,
pp. 77-82.
[17] S. I. M. Salim, H. A. Sulaiman, R. Jamaluddin, L. Salehuddin, M. N. S. Zainudin, and S. Yewguan, "Assembler
Design Techniques for A Reconfigurable Soft-Core Processor," Journal of Theoretical and Applied Information
Technology, vol. 64, 2014.
[18] S. I. M. Salim, H. A. Sulaiman, M. N. S. Zainudin, R. Jamaluddin, and L. Salahuddin, "One-pass assembler design
for a low-end reconfigurable RISC processor," in International Symposium onTechnology Management and
Emerging Technologies (ISTMET), 2014, pp. 492-496.
[19] D. Skrien, "CPU Sim 3.1: A Tool for Simulating Computer Architectures for Computer Organization Classes,"
Journal on Educational Resources in Computing (JERIC), vol. 1, pp. 46-59, 2001.
[20] A. J. Salim, N. R. Samsudin, S. I. M. Salim, and S. Yewguan, "Modification of Instruction Set Architecture in a
UTeMRISCII Processor," International Journal of Computer Trends and Technology (IJCTT), vol. 4,
pp. 1196-1201, 2013.
[21] A. J. Salim, N. R. Samsudin, S. I. M. Salim, and S. Yewguan, "Multiply-accumulate instruction set extension in a
soft-core RISC Processor," in 10th IEEE International Conference on Semiconductor Electronics (ICSE), 2012,
pp. 512-516.
BIOGRAPHIES OF AUTHORS
Sani Irwan Md Salim obtained his first degree in B.Eng (Electronic Engineering), from the Universiti
Teknologi Malaysia (UTM), in 2002. Following that, he obtained his M.EngSc (Computer &
Communication) from Queensland University of Technology (QUT), Australia. He is currently
working as senior lecturer in the Computer Engineering Department, Faculty of Electronic and
Computer Engineering, UTeM. His main research area is in reconfigurable computing with interest in
FPGA applications.
Dr Soo Yew Guan joins Universiti Teknikal Malaysia Melaka (UTeM) as academician since 2001.
Currently he is the Senior Lecturer in Microprocessor and Microcontroller Technology in the Faculty
of Electronics and Computer Engineering (FKEKK). As a professional trainer in Embedded System,
he has conducted CAN Bus Communication in in Embedded System training to Western Digital (M)
Sdn Bhd and Internet of Thing (IoT) programs in Faculty. His research interests include Embedded
System and Robotics, particularly on the myo-electric robotic prosthesis and sensory substitution for
amputees. In addition, his is also interested in software development for various mobile platforms
(iOS, Win8 Phone, and Android) and Internet Technologies (Linux, Apache, MySQL, PHP, and
JSON).
Dr Sharatul Izah Samsudin graduated for her his first degree in B.Eng (Electronic Engineering) at
Universiti Sains Malaysia (USM). Following that, she obtained her M.Eng. and PhD from Universiti
Teknologi Malaysia (UTM). She specializes on control system modeling with interest in water quality
control environment and IoT applications. She is currently working as a senior lecturer in the Industrial
Electronic Department, Faculty of Electronic and Computer Engineering, UTeM.

Instruction Set Extension of a Low-End Reconfigurable Microcontroller in Bit-Sorting Implementation

More Related Content

What's hot (20)

Similar to Instruction Set Extension of a Low-End Reconfigurable Microcontroller in Bit-Sorting Implementation (20)

More from IJECEIAES (20)

Recently uploaded (20)

Instruction Set Extension of a Low-End Reconfigurable Microcontroller in Bit-Sorting Implementation