Tesi sul tema "Application specific instruction-set processor (ASIP)"

Embedded systems are becoming ubiquitous, primarily due to the fast evolution of digital electronic devices. The design of modern embedded systems requires systems to exhibit, high performance and reliability, yet have short design time and low cost. Application Specific Instruction set processors (ASIPs) are widely used in embedded system since they are economical to use, flexible, and reusable (thus saves design time). During the last decade research work on ASIPs have been carried out in mainly for single pipelined processors. Improving performance in processors is possible by exploring the available parallelism in the program. Designing of multiple parallel execution paths for parallel execution of the processor naturally incurs additional cost. The methodology presented in this dissertation has addressed the problem of improving performance in ASIPs, at minimal additional cost. The devised methodology explores the available parallelism of an application to generate a multi-pipeline heterogeneous ASIP. The processor design is application specific. No pre-defined IPs are used in the design. The generated processor contains multiple standalone pipelined data paths, which are not necessarily identical, and are connected by the necessary bypass paths and control signals. Control unit are separate for each pipeline (though with the same clock) resulting in a simple and cost effective design. By using separate instruction and data memories (Harvard architecture) and by allowing memory access by two separate pipes, the complexity of the controller and buses are reduced. The impact of higher memory latencies is nullified by utilizing parallel pipes during memory access. Efficient bypass network selection and encoding techniques provide a better implementation. The initial design approach with only two pipelines without bypass paths show speed improvements of up to 36% and switching activity reductions of up to 11%. The additional area costs around 16%. An improved design with different number of pipelines (more than two) based on applications show on average of 77% performance improvement with overheads of: 49% on area; 51% on leakage power; 17% on switching activity; and 69% on code size. The design was further trimmed, with bypass path selection and encoding techniques, which show a saving of up to 32% of area and 34% of leakage power with 6% performance improvement and 69% of code size reduction compared to the design approach without these techniques in the multi pipeline design.

There is a lot of literature already available describing well-structured approach for embeddeddesign and implementation of Application Specific Integrated Processor (ASIP) micro processorcore.

This concept features hardware structured approach for implementation of processor core fromminimal instruction set, encoding standards, hardware mapping, and micro architecture design,coding conventions, RTL,verification and burning into a FPGA. The goal is to design an ASIPprocessor core (Micro architecture design and RTL) which can perform DSP task, e.g., FIR. Thereport is a well structured approach of design and implementation of an ASIP DSP processor forDSP applications like FIR. This report contains design flow starting from Instruction set design,micro architecture design and RTL implementation of the core. Details of the power simulationsof FPGA are also listed and analyzed.

This thesis addresses two ubiquitous trends in the embedded system world - the increasing importance of design turnaround time as a design metric, and the move towards closing the design productivity gap. Adopting the right choice of design approach has been recognised as an integral part of the design flow in order to meet desired characteristics such as increasing software content, satisfying the growing complexities of an application, reusing off-the-shelf components, and exploring design metrics tradeoff, which closes the design productivity gap. The importance of design turnaround time is motivated by the intensive competition between manufacturers, especially makers of mainstream electronic consumer products, who shrinks the product life cycle and requires faster time-to-market to maximise economic benefits. This thesis presents a suite of design automation methodologies to automatically design embedded systems for an application in the state-of-the-art design approach - the extensible processor platform. These design automation methodologies systematise the extensible processor platform???s design flow, with particular emphasis on solving four challenging design problems: i) code segment identification; ii) instruction generation; iii) architectural customisation selection; and iv) processor evaluation. Our suite of design automation methodologies includes: i) a semi-automatic design system - to design an extensible processor that maximises the application performance while satisfying the area constraint. By specifying a fitting function to identify suitable code segments within an application, a two-level hierarchy selection algorithm is used to first select a predefined processor and then select the right instruction, and a performance estimator is used to estimate an application's performance; ii) a tool to match instructions - to automatically match the pre-designed instructions with computationally intensive code segments, reducing verification time and effort; iii) an instructions estimation model - to estimate the area overhead, latency, power consumption of extensible instructions, exploring larger design space; and iv) an instructions generation tool - to generate new extensible instructions that maximises the speedup while minimising power dissipation. A number of techniques such as system decomposition, combinational equivalence checking and regression analysis etc., have been heavily relied upon in the creation of the final design system. This thesis shows results at every stage to demonstrate the efficacy of our design methodologies in the creation of extensible processors. The methodologies and results presented in this thesis demonstrate that automating the design process for an extensible processor platform results in significant performance increase - on average, an increase of 4.74x (up to 15.71x) compared to the original base processor. Our system achieves significant design turnaround time savings (2.5% of the full simulation time for the entire design space) with majority Pareto points obtained (91% on average), and can lead to fewer and faster design iterations. Our instruction matching tool is 7.3x faster on average compared to the best known approaches to the problem (partial simulations). Our estimation model has a mean absolute error as small as 3.4% (6.7% max.) for area overhead, 5.9% (9.4% max.) for latency, and 4.2% (7.2% max.) for power consumption, compared to estimation through the time consuming synthesis and simulation steps using commercial tools. Finally, the instruction generation tool reduces energy consumption by a further 5.8% on average (up to 17.7%) compared to extensible instructions generated by previous approaches.

The design of specialized instructions for application specific processors is a challenging task. This thesis describes the issues of effective specification and use of specialized instructions for optimization of applications. It focuses on improvements of the outputs and usability of the semiatomatic method of selection of specialized instructions to allow the optimization of complicated applications. This method combines manual selection of instructions by marking a section of source code in the application and automatic generation of the instruction description in the modelling language.

This thesis introduces the concept of assertion-based verifi cation of application-specifi c instruction set processors (ASIPs). The proposed design is implemented in SystemVerilog Assertions language as a part of veri fication environment created using Codasip Framework. The implemented concept is simulated in QuestaSim tool using model of Codix RISC processor. Main outcome of this thesis is the verifi cation concept usable not only on other processors, but as a part of system that automates the processor design as well.

This thesis presents design automation methodologies for extensible processor platforms in application specific domains. The work presents first a single processor approach for customization; a methodology that can rapidly create different processor configurations by the removal of unused instructions sets from the architecture. A profile directed approach is used to identify frequently used instructions and to eliminate unused opcodes from the available instruction pool. A coprocessor approach is next explored to create an SoC (System-on-Chip) to speedup the application while reducing energy consumption. Loops in applications are identified and accelerated by tightly coupling a coprocessor to an ASIP (Application Specific Instruction-set Processor). Latency hiding is used to exploit the parallelism provided by this architecture. A case study has been performed on a JPEG encoding algorithm; comparing two different coprocessor approaches: a high-level synthesis approach and our custom coprocessor approach. The thesis concludes by introducing a heterogenous multi-processor system using ASIPs as processing entities in a pipeline configuration. The problem of mapping each algorithmic stage in the system to an ASIP configuration is formulated. We proposed an estimation technique to calculate runtimes of the configured multiprocessor system without running cycle-accurate simulations, which could take a significant amount of time. We present two heuristics to efficiently search the design space of a pipeline-based multi ASIP system and compare the results against an exhaustive approach. In our first approach, we show that, on average, processor size can be reduced by 30%, energy consumption by 24%, while performance is improved by 24%. In the coprocessor approach, compared with the use of a main processor alone, a loop performance improvement of 2.57x is achieved using the custom coprocessor approach, as against 1.58x for the high level synthesis method, and 1.33x for the customized instruction approach. Energy savings are 57%, 28% and 19%, respectively. Our multiprocessor design provides a performance improvement of at least 4.03x for JPEG and 3.31x for MP3, for a single processor design system. The minimum cost obtained using our heuristic was within 0.43% and 0.29% of the optimum values for the JPEG and MP3 benchmarks respectively.

High efficiency and low power consumption are among the main topics in embedded systems today. For complex applications, off-the-shelf processor cores might not provide the desired goals in terms of power consumption. By optimizing the processor for the application, or a set of applications, one could improve the computing power by introducing special purpose hardware units. The execution cycle count of the application would in this case be reduced significantly, and the resulting processor would consume less power. In this thesis, some research is done in how to optimize a software and hardware development for ultra low power consumption. A cardiac beat detector algorithm is implemented in ANSI C, and optimized for low power consumption, by using several software power optimization techniques. The resulting application is mapped on a basic processor architecture provided by Target Compiler Technologies. This processor is optimized further for ultra low power consumption by applying application specific hardware, and by using several hardware power optimization techniques. A general processor and the optimized processor has been mapped on a chip, using a 90 nm low power TSMC process. Information about power dissipation is extracted through netlist simulation, and the results of both processors have been compared. The optimized processor consume 55% less average power, and the duty cycle of the processor, i.e., the time in which the processor executes its task with respect to the time budget available, has been reduced from 14% to 2.8%. The reduction in the total execution cycle count is 81%. The possibilities of applying power gating, or voltage and frequency scaling are discussed, and it is concluded that further reduction in power consumption is possible by applying these power optimization techniques. For a given case, the average leakage power dissipation is estimated to be reduced by 97.2%.

This thesis describes the design of the universal assembler that represents a part of the Lissom project. You will be provided with the description of the assembler architectures and their usual tasks. Special attention is paid to GNU assembler. Designed assembler consists of the fixed and the generated part. The generated part is created automatically from the description of instruction set, that is defined using architecture and instructions set description language ISAC. Using this approach, it is possible to change assembler target architecture automatically. The second part of thesis describes the Parserlib2 library implementation that is a part of the Lissom project and provides the information about the target instruction set for an assembler generator.

Este trabalho propõe a geração de uma arquitetura dedicada a aplicações específicas, baseadas no microcontrolador MCS8051. Por ser utilizado na solução de problemas em indústrias locais, este processador foi escolhido para servir como base em um sistema dedicado. O 8051 dedicado gerado deverá permitir a integração completa do sistema, proporcionando um aumento do valor agregado e, conseqüentemente, a diminuição do custo. Busca-se com a otimização da arquitetura obter um conjunto de instruções reduzido, construído com as instruções mais utilizadas em cada aplicação. O objetivo principal da otimização do conjunto de instruções está relacionado ao fato de que os circuitos decodificadores e geradores de microcódigo da parte de controle ocupam uma área significativa do processador. Uma otimização no sentido de reduzir-se o conjunto de instruções, portanto, resulta numa economia de área, o que vem de encontro com a idéia da integração completa do sistema com o processador. Um processador dedicado a aplicações específicas (ASIP) irá possuir um custo maior do que a sua versão original, devido as otimizações realizadas. Para compensar este custo, uma alternativa a seguir é a integração completa do sistema. Um Sistema Integrado para Aplicações Específicas (SIAE) torna-se desejável, pois aumentando o valor agregado do circuito possibilita-se a redução do custo pela eliminação de conexões da placa, do encapsulamento de outros circuitos, entre outros motivos. Todavia, para que um SIAE possa ser construído com um custo aceitável, é necessário que seja construído em uma área que não exceda muito a área original do processador. Tenta-se fazer isto neste trabalho, através da implementação de aplicações com poucas instruções diferentes. Por ser uma arquitetura comercial, o 8051 possui um grande parque de software desenvolvido e resolvendo problemas. Isto pode ser considerado uma vantagem pois, software básicos como por exemplo, compiladores, já estão desenvolvidos. Outra vantagem é o grande número de engenheiros treinados na sua utilização. Desse modo, torna-se necessária a criação de uma compatibilidade de software, para preservar o que já está desenvolvido. Uma vez que a programação em nível de linguagem montadora tende a constituir-se em uma tarefa cansativa e sujeita a erros, é desejável que se tenha uma compatibilidade em alto nível, ou seja, através de um compilador. Para criar a compatibilidade de SW necessária é realizada a otimização de um compilador C desenvolvido para o 8051. A escolha pela linguagem C deve-se ao fato de sua grande utilização. O compilador C otimizado procura utilizar um conjunto de instruções reduzido para obter a economia de área. Quando uma instrução necessita ser utilizada e não está presente no conjunto de instruções desejado, o compilador tenta substituí-la por outra(s). Um conjunto de instruções é utilizado para cada aplicação, sendo constituído pelas instruções mais utilizadas por esta. Para determinar as instruções mais utilizadas de cada aplicação é realizada uma análise estática sobre um código em linguagem montadora previamente compilado. As instruções implementadas serão sempre parte do conjunto de instruções original do 8051, de modo que novas instruções não serão criadas.Um programa em linguagem montadora gerado com um conjunto de instruções reduzido (RISC) normalmente terá um número maior de instruções do que o seu 10 equivalente com o conjunto de instruções completo (CISC). Isto ocorre porque possivelmente algumas substituições de uma instrução por outras, terão que ser realizadas. Como as instruções que serão utilizadas nas substituições pertencem ao conjunto de instruções original, o programa gerado com o compilador otimizado poderá executar em um tempo maior do que se fosse compilado com o código CISC. Para compensar esse atraso foi implementado um pipeline de instruções para o 8051. Este trabalho apresenta resultados da Síntese Lógica em Standard Cell e FPGA da arquitetura otimizada. Além disso, resultados de programas em linguagem montadora gerados com o compilador otimizado, são também apresentados.
This work discusses a processor for specific applications architecture, based on the MCS8051 microcontroller. This processor is used as a solution for many local industry applications, being the base of dedicated systems. The dedicated 8051 generated should allow complete integration of the system, and with the added value to the chip, reduced costs. The architecture optimization will produce as result a reduced instruction set, made by the often used instructions for each application. The main instruction set optimization goal refers to the instrucions decoders and microcode generators in the control part, because a large area in the processor is needed to implement them. Thus, a reduced instruction set will allow area savings, making possible the complete system integration in a chip. An ASIP architecture will have a higher cost than the original one. An alternative to solve this problem is add value to the chip, creating an Application Specific Integrated System (ASIS). An ASIS can be made with a acceptable cost, if it’s possible to integrate other circuits to the chip without area increase. This can be done in the area saved by using fewer implemented instructions. Because the 8051 is a commercial architecture, there is a large amount of software developed for it. This can be considered an advantage because basic softwares like compilers are available, being not necessary to create them. Another advantage refers to the large number of engineers trained to use the 8051. To preserve the already developed applications it’s necessary to mantain software compatibility. Assembler level programming is very boring an error prone task, being desirable to have software compatibility at higher levels through the use of high level languages. To create the necessary SW compatibility, a C compiler developed for 8051 was optimized. The chose for C language refers to its large utilization. The optimized C compiler tries to use a reduced instruction set, formed with the most important instructions for each application, in order ro save area. When an instruction needs to be used in an application, and it’s not present in the instruction set, the compiler tries to replace it with other instructions. The compiler will not use instructions not present in the original 8051 instruction set. So, new instrucions will be not created. To create an instruction set formed with the most important instructions for each application, a static analysis is made on a precompiled assembler source. An assembler source generated with a reduced instruction set (RISC) will probably have more instructions than the same assembler generated with a full instruction set (CISC). This can be explained because of the replacements instruction. If one instruction is replaced by other two, and these are from the original instruction set, probably the time needed to execute them would be higher. In order to deal with this problem, an instruction pipeline was implemented to the 8051. This work presents Standard Cells and FPGA results of Logic Synthesis of the optimized architecture. Also, assembly programs generated by the optimized compiler are presented.

With commercial processor design tools, a designer can quickly design a C- programmable ASIP for a specific application domain. There are several such ASIPs available for both wireless (UWB baseband processing), encryption, and biomedical processing (particularly for ECG beat detection). In traditional CPUs and DSPs the impact of the instruction-set definition and the complexity of the instruction decoder can be substantial, especially in terms of power consumption. Fully orthogonal VLIW processors, do not incur the cost of an instruction decoder that severely. Instead the instruction word becomes very large, thereby shifting the (power-)cost to the program memory or instruction cache. For the purposes of this thesis a SIMD processor is developed and is compared to a soft-SIMD to observe its area, performance and energy efficiency for a bioimaging benchmark and how the processor description in the ASIP language nML, defines the generated HDL. This SIMD processor is turned into orthogonal and using iterative experiments it is investigated, what is the impact on power while manipulating the instruction-set architecture in combination with the program memory size. It is also investigated how instruction-set re-configuration can be exploited to improve power efficiency. Using this investigation guidelines for low-power ASIP design can be produced.
Με τη σύγχρονη τεχνολογία σχεδιασμού επεξεργαστών, ο σχεδιαστής μπορεί με ευκολία να σχεδιάσει ένα προγραμματιζόμενο Επεξεργαστή Συνόλου Εντολών Ειδικού Σκοπού (ASIP - Application-Specific Instruction-set Processor) για ένα συγκεκριμένο εύρος εφαρμογών. Υπάρχουν διάφοροι τέτοιοι επεξεργαστές διαθέσιμοι για ασύρματες εφαρμογές, κρυπτογράφηση και βιοϊατρικές εφαρμογές (π.χ. στον αλγόριθμο εντοπισμού χτύπου ηλεκτροκαρδιογραφήματος). Στους παραδοσιακούς επεξεργαστές και επεξεργαστές σήματος (DSP - Digital Signal Processor) ο ορισμός του συνόλου εντολών και η πολυπλοκότητα έχουν μεγάλη επίδραση, ειδικά στην κατανάλωση ισχύος. Μία πιθανή λύση σε αυτό το πρόβλημα είναι οι ορθογώνιοι επεξεργαστές μεγάλου μεγέθους λέξης εντολής (VLIW - Very Large Instruction Word). Με τον όρο ορθογώνιο επεξεργαστή, ορίζεται ένας επεξεργαστής οριζόντιου σύνολου εντολών, άρα ένας επεξεργαστής στον οποίο μπορεί να υπάρξει κάθε διαθέσιμος συνδυασμός μεταξύ των διαθέσιμων εντολών και των μεθόδων διευθυνσιοδότησης για πρόσβαση στη μνήμη και το αρχείο καταχωρητών. Οι ορθογώνιοι επεξεργαστές δεν επιβαρύνουν τόσο τον αποκωδικοποιητή εντολών. Αντί αυτού το μέγεθος της λέξης της εντολής γίνεται πολύ μεγάλο, και έτσι μετατίθεται το ενεργειακό κόστος στην μνήμη εντολών προγράμματος (program memory )ή την κρυφή μνήμη εντολών προγράμματος (instruction cache). Για τους σκοπούς αυτής της διπλωματικής εργασίας, αναπτύχθηκε ένας επεξεργαστής SIMD, ο οποίος συγκρίνεται με έναν soft-SIMD για να μελετηθούν η απαιτούμενη περιοχή στο ενσωματωμένο, επιδόσεις και κατανάλωση ενέργειας για μία βιοϊατρική εφαρμογή, καθώς και το πως η περιγραφή ενός επεξεργαστή στη γλώσσα περιγραφής επεξεργαστών ASIP nML ορίζει την παραγούμενη γλώσσα περιγραφής υλικού (HDL - Hardware Description Language). Ο επεξεργαστής αυτός μετατρέπεται σε ορθογώνιο, και με τη χρήση επαναληπτικών πειραμάτων μελετάται η επίδραση στην κατανάλωση ενέργειας κατά τη διάρκεια αλλαγών στην αρχιτεκτονική του συνόλου εντολών και του μεγέθους της μνήμης εντολών προγράμματος. Ακόμη μελετάται πως μπορεί να εκμεταλλευτεί ο σχεδιαστής την αναδιάρθρωση του συνόλου εντολών για να βελτιώσει την κατανάλωση ενέργειας.

The battery driven nature of wireless sensor networks, combined with the need of extended lifetime mandates that energy efficiency is a metric with high priority. In the current thesis we explore and compare the energy dissipation of di fferent processor architectures and how it is associated with performance and area requirements. The processor architectures are di erentiated based on the datapath length (16-bit, 32-bit, 64-bit and 128-bit) and the corresponding size of the data memories. Our study focuses on AES algorithm, and the indicated processor architectures support AES forward encryption, CCM (32/64/128), CBC (32/64/128) and CTR common modes of operation. In each processor architecture the instruction set is extended to increase the efficiency of the system.
-

Cheung Man Ting.
Thesis (M.Phil.)--Chinese University of Hong Kong, 2005.
Includes bibliographical references (leaves 69-71).
Abstracts in English and Chinese.
Chapter 1 --- Introduction --- p.1
Chapter 1.1 --- The Emergence of ASIP --- p.1
Chapter 1.1.1 --- Related Work --- p.3
Chapter 1.2 --- Motivation --- p.6
Chapter 1.3 --- ASIP Design Methodologies --- p.7
Chapter 1.4 --- Fundamentals of Speech Recognition --- p.8
Chapter 1.5 --- Thesis outline --- p.10
Chapter 2 --- Automatic Speech Recognition --- p.11
Chapter 2.1 --- Overview of ASR system --- p.11
Chapter 2.2 --- Theory of Front-end Feature Extraction --- p.12
Chapter 2.3 --- Theory of HMM-based Speech Recognition --- p.14
Chapter 2.3.1 --- Hidden Markov Model (HMM) --- p.14
Chapter 2.3.2 --- The Typical Structure of the HMM --- p.14
Chapter 2.3.3 --- Discrete HMMs and Continuous HMMs --- p.15
Chapter 2.3.4 --- The Three Basic Problems for HMMs --- p.17
Chapter 2.3.5 --- Probability Evaluation --- p.18
Chapter 2.4 --- The Viterbi Search Engine --- p.19
Chapter 2.5 --- Isolated Word Recognition (IWR) --- p.22
Chapter 3 --- Design of ASIP Platform --- p.24
Chapter 3.1 --- Instruction Fetch --- p.25
Chapter 3.2 --- Instruction Decode --- p.26
Chapter 3.3 --- Datapath --- p.29
Chapter 3.4 --- Register File Systems --- p.30
Chapter 3.4.1 --- Memory Hierarchy --- p.30
Chapter 3.4.2 --- Register File Organization --- p.31
Chapter 3.4.3 --- Special Registers --- p.34
Chapter 3.4.4 --- Address Generation --- p.34
Chapter 3.4.5 --- Load and Store --- p.36
Chapter 4 --- Implementation of Speech Recognition on ASIP --- p.37
Chapter 4.1 --- Hardware Architecture Exploration --- p.37
Chapter 4.1.1 --- Floating Point and Fixed Point --- p.37
Chapter 4.1.2 --- Multiplication and Accumulation --- p.38
Chapter 4.1.3 --- Pipelining --- p.41
Chapter 4.1.4 --- Memory Architecture --- p.43
Chapter 4.1.5 --- Saturation Logic --- p.44
Chapter 4.1.6 --- Specialized Addressing Modes --- p.44
Chapter 4.1.7 --- Repetitive Operation --- p.47
Chapter 4.2 --- Software Algorithm Implementation --- p.49
Chapter 4.2.1 --- Implementation Using Base Instruction Set --- p.49
Chapter 4.2.2 --- Implementation Using Refined Instruction Set --- p.54
Chapter 5 --- Simulation Results --- p.56
Chapter 6 --- Conclusions and Future Work --- p.60
Appendices --- p.62
Chapter A --- Base Instruction Set --- p.62
Chapter B --- Special Registers --- p.65
Chapter C --- Chip Microphotograph of ASIP --- p.67
Chapter D --- The Testing Board of ASIP --- p.68
Bibliography --- p.69

碩士
國立交通大學
電子工程系所
93
Programmable processors are dramatically attractive to amortize manufacturing costs and design efforts, as the system complexity grows. Besides, in order to satisfy the tight design constraints such as performance and power of today’s embedded systems, processor architectures are getting more specialized to some application domains (e.g. an application-specific instruction-set processor; ASIP). This thesis discusses the acceleration of system prototyping of new processor cores by reducing the software development time. Firstly, we propose a simple and effective high-level language compilation method by encapsulation new processor cores in compiler o friendly RISC shell. The native code translation form compiled RISC codes to the target processor is carried out by cooperating hardware and software. Secondly, we propose an efficient instruction set simulator with decoupled hazard checker and memory simulator. The simulation time is significantly reduced via native translation, while the cycle accuracy is maintained with proper instrumentation. Finally, we have constructed a C compiler with 6.98% hardware over head and a cycle-accurate ISS with 102~104 speed up for a proprietary DSP processor. Moreover, we have developed JPEG and H.264 encoding systems based on these software tools.

碩士
逢甲大學
資訊工程所
95
Embedded system designers sometimes need to optimize some specific time consuming application programs to speed up the system. Optimizing embedded processor for specific application is studied in recently years. To re-design whole processor makes a long design time, so the idea of changing instruction set of processor only is proposed. As some tool providers adding such a idea into their system design tool, ASIP is beginning to be discussed. In past days, ASIP researches almost focus on instructions only. There is a few researches which are considering instructions and the components outside of processor at the same time. However, system optimizing is not only involving the processor but also the memory or device controller. Single memory access or device I/O usually makes longer time than single computation operation. Because memory accessing may reduced by some compiler techniques, if we can make some flow to find new instruction, called Application-Specific Instruction(ASI), and reduce memory access at the same time, therefore system performance will be accelerated. In this thesis, we will propose a design flow to find new instructions for embedded processor and take memory access into account. At the beginning of our flow, we will compile the application from C to assembly without register allocation, and then produce data flow graphs (DFGs) from the assembly. We generate instruction templates from the generated DFGs. All isomorphic templates will be gathered as a candidate instruction. Finally, ASIs will be selected to satisfy all specified constraints. The experiment results will be evaluated by microprocessor simulator. Our result show that the program sha using ASIs generated by our flow obtains at most 22% performance improvement comparing to the ASIs ignoring memory factor.