IL142342A - Method and apparatus for managing the configuration and functionality of a semiconductor design - Google Patents

Description

rumtt v^n n ivai n^iwas^a ^a \? )pmm no \y Method and apparatus for managing the configuration and functionality of a semiconductor design ARC Cores, Ltd.

C. 132416 METHOD AND APPARATUS FOR MANAGING THE CONFIGURATION AND FUNCTIONALITY OF A SEMICONDUCTOR DESIGN | 1. I Held of the Invention The invention relate, generally to the field of semiconductor design and layout and computer automated design (CAD) for semiconductors. More specifically, the invention provides a mBthod for managing the configuration, design parameters, and functionality of an integrated circuit design in which custom instructions and other elements may be arbitrarily controlled by the designer. 2. Description of Related Technology Several types of computer aided design (CAD) tools are available to design and fabricate integrated circuits (IC). Such computer-aided or automated IC design tools can include modules or programs addressing both the synthesis and optimization processes. Synthesis is generally defined as an automatic method of converting a higher level of abstraction to a lower level of abstraction, and can include any desired combination of synthesis techniques which occur at various levels of abstraction. So-called "behavioral synthesis" is a design tool wherein the behavior (e.g. inputs, outputs, and functionality) of a desired IC are entered into a computer program to design a device that exhibits the desired behavior. Such tools permit IC designers to produce increasingly complex and capable devices, sometimes having logic gate counts in the tens of millions, with few or no errors and in a much shorter time period than would be otherwise possible with manual design techniques such as hand layouts.

Examples of synthesis processes which involve different levels of abstraction include architectural level synthesis and logic level synthesis, bath of which may be incorporated into the IC design process.

Architectural level synthesis is primarily concerned with the macroscopic structure of the circuit; it utilizes functional blacks (including information relating to their interconnections and internal functionality). Architectural level synthesis includes register transfer level (RTL) synthesis, which can have multi-bit components such as registers and operators.

Logic level, synthesis, on the other hand, is concerned with gate level design. Logic level synthesis determines a microscopic structure of a circuit and transforms a logic model into an interconnection of instances of library cells. The result of the logic level synthesis is a netlist of logic devices and their interconnections. Logic-levBl synthesizers (so-called synthesis "engines") are available from several commercial vendors.

The synthesis process generally begins with the designer compiling a set of IC specifications based on the desired functionality of the target device. These specifications are then encoded in a hardware description language (HDL) such as VHDL® (VHSIC hardware description language) available from IEEE of New York, NY, or Verilog® available from Cadenca Design Systems, Inc. of Santa Clara, CA. The specifications define an IC in terms of the desiied inputs information relating to the design hierarchy and HDL library files, and generates a computer program script based on these inputs. Custom instructions and other elements (such as special-purpose registers, new condition code choices, local scratchpad RAM, or a library of multimedia extensions for 3D or OSP applications) may be arbitrarily added to the existing HDL instruction set by the designer as well during the design process. Unlike adding an external ASIC or other component post-synthesis, these instructions become included within the processor instruction and register set so as to eliminate integration problems, and allow maximal optimization based on one or more selected attributes. Ultimately, the generated script is run, and a customized HDL model of the IC is produced based on the user-defined instruction set This customized HDL model is then used as the basis for further simulation and/or synthesis as desired by the designer. This method further affords the designer the ability to generate an HDL model of the entire IC, thereby greatly enhancing the efficiency of the design process. This ability is especially useful for complex CPU- based processor designs, although it may readily be applied to other types of components.

In a second aspect of the invention, the eforementioned method is further embodied in a menu-driven computer program which may be used in conjunction with a microcomputer or other similar device for designing and managing the configuration of integrated circuits. In one exemplary embodiment, this menu-driven computer program comprises a series of routines or 'modules* which perform various sets or groups of related functions. An interactive architecture module gathers information about the designer's system, available libraries, and the design configuration to be generated. A hierarchy generation module assists in ordering logical blocks or entities and routing signals within the design, as well as ultimately generating a netlist. An HDL generation module generates a merged HDL file descriptive of the design. In the exemplary embodiment, these modules are connected together via computer programs scripts or user initiated execution of the individual modules. In yet another embodiment, these modules may also be compiled into one comprehensive program, and bundled with other software tools that facilitate rapid and integrated design of the subject IC on a standard microcomputer.

In a third aspect of the invention, en apparatus for generating, simulating, and/or synthesizing an integrated circuit design using the aforementioned m-thod is disclosed. In a first embodiment, a microcomputer having a microprocessor, display, and input device is used to run the previously described menu-driven computer program, thereby allowing the designer to interact with the program during the design, simulation, and/or synthesis processes. The microcomputer further includes a storage device and network interface to allow for the storage, retrieval and sharing of information between two or more microcomputers configured with the aforementioned computer program.

In a fourth aspect of the invention, an integrated circuit design depicted in a hardware description language and synthesized using the aforementioned method of the invention is disclosed.

In a fifth aspect of the invention, an integrated circuit fabricated using the aforementioned synthesized design is disclosed. In one exemplary embodiment, the integrated circuit comprises a reduced instruction set CPU (RISC) processor.

In a sixth aspect of the invention, a data storage device adapted for use with a computer system end comprising in part the aforementioned computer program is disclosed.

Brief Description of the Drawings Figure 1 is a flow diagram illustrating a typical prior art integrated circuit design and synthesis method.

Figure 2 is a flow diagram illustrating the general integrated circuit design and synthesis method of the invention.

Figure 2a is a flow diagram illustrating one specific embodiment of the generalized method of Figure 2.

Figures 3a-3e collectively comprise a logical flow diagram illustrating one exemplary embodiment of the architecture module of the invention.

Figure 4 is a logical flow diagram illustrating one exemplary embodiment of the method of determining the memory and register configuration of the system under design as shown generally in Figure 3c herein.

Figures 5a-5h collectively comprise a logical flow diagram illustrating one exemplary embodiment of the method of configuring file extensions as shown generally in Rgure 3c herein.

Figure 6 is a logical flow diagram illustrating one exemplary embodiment of the method of setting up the cache structure as shown generally in Figures 5b, 5c and 5d herein.

Rgure 7 is a logical flow diagram illustrating one exemplary embodiment of the method of setting up the load/store interface structure as shown generally in Figure 5d herein.

Rgure 8 is a logical flow diagram illustrating one exemplary embodiment of the method of setting up the register transfer language (RTL) Gnk structure as shown generally in Rgure 3d herein.

Rgure 9 is a logical flow diagram illustrating one exemplary embodiment of the method of generating a new script according to the invention.

Rgure 10 is a logical flow diagram of one exemplary embodiment of the method of building a libraries Gst as shown generally in Rgure 9.

Rgure 11 is a logical flow diagram of one exemplary embodiment of the method of adding the libraries list to the system "makefile" as shown generally in Figure 9.

Rgure 12 is a logical flow diagram of one exemplary embodiment of the method of adding datafile dependency rules to the system "makefile" as shown generally in Rgure 9.

Rgure 3 is a logical flowchart illustrating one exemplary embodiment of the method of hierarchy file generation according to the invention.

Figure 14 is a logical flowchart illustrating one exemplary embodiment of the method of reading input files in accordance with the hierarchy file generation method of Rgure 13.

Figure 15 is a logical flowchart illustrating one exemplary embodiment of the method of reading hierarchy files as shown generally in Rgure 14.

Rgure 16 is a logical flowchart illustrating one exemplary embodiment of the method of reading top level entity data files as shown generally in Figure 14.

Rgure 17 is a logical flowchart illustrating one exemplary embodiment of the method of reading primary block data files as shown generally in Figure 14.

Figure 18 is a logical flowchart illustrating one exemplary embodiment of the method of identifying and integrating new signals into the block data files as shown generally in Figure 17.

Rgure 19 is a logical flowchart illustrating one exemplary embodiment of the method of verifying data integrity as shown generally in Rgure 13.

S Figure 20 is a logical flowchart illustrating one exemplary embodiment of the method of adjusting signal types relative to the block data files as shown generally in Figure 13.

Rgure 21 is a logical flowchart illustrating one exemplary embodiment of the method of checking for "in" or "out" signals as shown generally in Rgure 20.

Figure 22 is a logical flowchart illustrating one exemplary embodiment of the method of generating the merged 10 hardware description language (HDL) file according to the invention.

Figure 23 is a logical flowchart illustrating one exemplary embodiment of the method of reading the HDL control file as shown generally in Rgure 21 Rgure 24 is a logical flowchart illustrating one exemplary embodiment of the method of reading the HDL top level entity data file as shown generally in Rgure 22. ] 5 Rgure 25 is a logical flowchart illustrating one exemplary embodiment of the method of reading the HDL dependency data file as shown generally in Rgure 21 Rgure 26 is a logical flowchart illustrating one exemplary embodiment of the method of processing the signal record as shown generally in Figure 25.

Rgure 27 is a logical flowchart illustrating one exemplary embodiment of the method of processing the signal list 0 into groups as shown generally in Rgure 22.

Figure 28 is a logical flowchart illustrating one exemplary embodiment of the method of verifying data integrity as shown generally in Rgure 21 Figure 29 is a logical flowchart illustrating one exemplary embodiment of the method of reading the library file as shown generally in Rgure 21 Figure 30 is a logical flowchart illustrating one exemplary embodiment of the method of writing the HDL file as shown generally in Rgure 21 Rgure 31 is a perspective view of one exemplary embodiment of an integrated circuit design apparatus according to the invention.

Rgure 32 is a top plan view illustrating the layout of one exemplary embodiment of a processor-based integrated 30 circuit designed using the method of the invention.

Detailed Description of the Invention Reference is now made to the drawings wherein like numerals refer to like parts throughout.

As used herein, the terms "computer program," "routine," "subroutine," and "algorithm" are essentially synonymous, with "computer program" being usBd typically (but not exclusively) to describe collections or groups of the latter three elements. In general, however, all of the aforementioned terms as used herein are meant to encompass any series of logical steps performed in a sequence to accomplish a given purpose.

Also, while the following description relates explicitly to the VHDL environment it can be appreciated that the general principles and functionality of the present invention may be embodied in other description language environments including, for example, Verilog®.

Background of "ARCn" In the following discussion, the term "ARC™" (ARC RISC Core) refers to a microprocessor-like central processing unit (CPU) architecture. While the method of present invention can be applied to other types of integrated circuits including, for example, application specific integrated circuits (ASICs ) and field-programmable gate arrays (FPGAs), the microprocessor-like CPU design is chosen to more clearly illustrate the operation and capability of the invention.

As discussed in greater detail below, the present invention takes in one embodiment the form of a computer program having a synthesizable and customizable (i.e., "soft") macro with a complementary suite of software tools for configuration and hardware-software co-design. This computer program employs the concept of a "system builder" to accomplish much of the desired functionality. In the present context, the ARC system builder refers to that portion or module of the computer program whereby the designer controls the generation of the subject CPU design. In one particular embodiment, the system builder directs the assembly of a series of predefined VHDL based designs along with design elements created by the user (also in VHDL, or whatever other description language is chosen) to produce a new, custom CPU specific to the user's specifications and needs. Hence, the system builder controls the creation and testing of HDL-based synthesizable CPUs. This approach further provides users with a great deal of flexibility in configuring the specific attributes of the resulting CPU.

The ARC System builder is embodied in a series of "scripts" that allows users to build customized ARC systems along with support files for both design simulation and synthesis. A script is mare specifically a computer program, often written in a special computer programming language designed for the purpose. Such script languages include for example the "perl" and language commonly employed in UNIX based computer systems as a "scripting language." There are other languages available to write scripts compatible with the present invention. It is noted that for the purposes of the present discussion, the term "scripts" refers to any series of logical instructions or actions of a computer program which are executed in a predefined order.

When executed, the ARC system builder script produces a series of questions, answered primarily through menus (as illustrated in greater detail below in the exemplary menu structure of Appendix I), the answers to which are used to build the VHDL simulator and synthesis files. An installation script allows several different ARC features to be selected, such as processor cache size and cache line length, size of external memory space to be cached, and clock period/skew. The script creates a working directory for the user, and further generates various VHDL files necessary to support further customized VHDL development.

The term 'makefile' as used herein refers to the commonly used UNIX makefile function or similar function of a computer system well known to those of skill in the computer programming arts. The makefile function causes other programs or algorithms resident in the computer system to be executed in the specified order, in addition, it further specifies the names or locations of data files and other information necessary to the successful operation of the specified programs. It is noted, however, that the invention disclosed herein may utilize file structures other than the "makefile' type to produce the desired functionality.

Central to the method of the invention is the concept that one computer program may write another program, which is subsequently executed by the computer system. For example, one script may write a second script that is tailored by user input to perform a specific task. The task is then performed when the second script is executed. This 'dynamic scripting" is employed in various aspects of the invention, as further described herein.

Detailed Description of Method Referring now to Figure 2, one embodiment of the generalized method of the invention is described. While this description is presented in terms of an algorithm or computer program running on a microcomputer or other similar processing device, it can be appreciated that other hardware environments (including minicomputers, workstations, networked computers, "supercomputers", and mainframes) may be used to practice the method 100. Additionally, one or more portions of the computer program may be embodied in hardware or firmware as opposed to software if desired, such alternate embodiments being well within the skill of the computer artisan.

Initially, the hierarchy file specific to the processor design under consideration is edited in the first step 102. Specifically, desired modules or functions for the design are selected by the user, and instructions relating to the design are added, subtracted, or generated as necessary. For example, in signal processing applications, it is often advantageous for CPUs to include a single "multiply and accumulate" (MAC) instruction. This instruction commonly improves system performance and reduces the amount of computer code required to perform signal processing tasks; two desirable characteristics in such systems. Conversely, applications that are essentially logical control oriented in nature have little or no need for such an instruction. Eliminating the instruction from a purpose-built CPU reduces the CPU die size and gate count, also a generally preferred advantage. In still another case, an instruction that is only of value to one specific application may be defined end implemented by designers. In all of these examples, the present invention permits designers a greater degree of control over the CPU design and more flexibility in making design tradeoffs and optimizations.

It should also be recognized that when using the system builder of the present embodiment in its most common mode, the hierarchy of design is determined by the script, and a "hierarchy builder" is called by the script builder.

In step 104 of the method of Figure 2, the technology library location for each VHDL file is defined by the user. The technology library files in the present invention store all of the information related to cells necessary for the synthesis process, including for example logical function, input/output timing, and any associated constraints. In the present invention, each user can define his/her own library name and location(s), thereby adding further flexibility.

A makefile is then generated in the third step 106 using the hierarchy file information, library information previously described, as well as additional user input information relating to the desired configuration and attributes of the device. For example, in the ARC system builder, the user is asked to input the type of "build" (e.g., overall device or system configuration}, width of the external memory system data bus, different types of extensions, cache type/size, etc., as described in additional detail below with respect to Figures 3a-3e. It will be recognized that such user input may be in the farm of an interactive software prompt, or alternatively may occur using command line parameters, by accessing a pre-built configuration file, or a concurrently running application or other computer program resident on the designer's system or another system linked thereto. Many other methods of input are also possible, all of which are considered to be within the scope of the invention.

Next, the makefile generated in the third step 106 is run in a fourth step 108 to create a customized VKOL model. As part of this step, the user is presented with a variety of optional response menus and prompts to further define the VHDL model based on his/her desired attributes.

At this point in the program, a decision is made whether to synthesize or simulate the design (step 1 0). If simulation is chosen, a separate script ("MTI make" in the present embodiment) is run in step 2 to generate a separate simulation makefile for use in the simulation. Simulation scripts and makefiles are generally well known to those of ordinary skill in the art, and accordingly will not be discussed further herein. Alternatively, if synthesis is chosen, the synthesis script(s) (e.g., "synopsysjnake") are run in step 14 to generate corresponding synthesis makefiles. After completion of the synthesis/simulation scripts, the adequacy of the design is evaluated in step 116. For example, a synthesis engine may create a specific physical layout of the design that meets the performance criteria of the overall design process yet does not meet the die size requirements. In this case, the designer will make changes to the control files, libraries, or other elements that can affect the die size. The resulting set of design information is then used to re-run the synthesis script.

Note that there are many different criteria that may be used to determine whether or not a design is acceptable. In general, each design will have a different set of criteria used to determine the acceptability of the final design. Die size, power consumption, clock speed, and numerous other factors may constitute alonB or collectively design constraints or targets. This great variability in acceptance criteria is one factor that demands the flexibility of the present invention.

If the generated design is acceptable, the design process is completed. If the design is not acceptable, the process steps beginning with step 102 are re-performed until an acceptable design is achieved. In this fashion, the method 100 is iterative. Note that if the simulation path 112 is chosen in order to simulate the operation of the generated design, the user may subsequently run the synthesis scripts per step 114 (dotted line in Rgure 2) if desired to synthesize the logical design into a physical structure.

Appendix I illustrates the operation of an algorithm utilizing the method of Figure 2. In Appendix I, several explicit examples illustrating various menu structures, design considerations, and user input options are presented.

Note that these examples are merely illustrative of the general process performed by the user in practicing the method of the present invention using the VHDL software embodiment set forth herein.

Appendix II is a list of the VHDL script files used in conjunction with the VHDL embodiment of the algorithm of the present invention.

It is noted that while the source code for the embodiment of the computer program set forth herein is written using AWK, a programming language commonly found on UNIX workstations, but also available on other personal computers, the program may be embodied using any of a number of different programming languages such as, for example, C*\ The source code directed to the exemplary AWK embodiment of the invention described herein is set forth in Applicant's aforementioned Provisional U.S. Patent Application number 601104,271, which is incorporated herein by reference in its entirety.

Referring now to Figure 2a, one specific embodiment of the general method of Figure 2 is described. In this embodiment, certain steps of the method depicted in Figure 2 are separated into constituent parts for illustration purposes. For example, step 102 of Figure 2 is separated into an interactive ("Wizard™") component and a hierarchy generator ("hiergen") component (not shown). The interactive component of step 102 provides substantially all of the direct user interaction control. Through a series of questions answered by the user, the program selects the relevant design elements necessary to realize the user's design. Portions of this information are used by the hierarchy generator to create a makefile script that is executed to build the HDL hierarchy.

Similarly, the step 104 of defining the library location for each VHDL file in Figure 2 corresponds to the steps of (i) creating a working directory 204, (ii) copying files from e master database 206, and merging the selected extension VHDL modules into placeholder files 208. The remainder of method 00 depicted in Figure 2a generally parallels that of Figure 2.

It will be appreciated by one skilled in the relevant art that there are a large number of alternative partitionings or structures of the flowchart of Figure 2, each of which results in the same or similar sets of scripts, makefiles, and other design data for a given set of input data. Further, it may be advantageous for additional data or partitionings to be selected in order to utilize commonly available tools for executing portions of the method 100.

Also, the order of performance of several of the individual steps present in the mBthod of Figure 2 (and Figure 2a) may be permuted without departing from the invention as disclosed herein.

Referring now to Figures 3a-3e, one embodiment of the architecture functionality of the invention ("ARChitect™") is described in detaiL The architecture module essentially gathers information about the designer's system, including for example, specifications from the designer's libraries, and assembles this information for later use by other modules within the program. The output of the ARChitect ™ process is an HOL synthesizable model of the IC under design.

As shown in Figure 3a, the process 300 specifically employed by the ARChitect"1 module comprises a series of steps, the first of which is initializing the program and setting default values 301 Next a destination directory is created in step 304 for the HDL fin this case VHDL) files to be generated in subsequent steps. In step 306, the technology/feature size is obtained from the user (or another source, as previously described). For example, the user may desire a 0.18 micron process for their design.

In step 308, the system clock, memory configuration, and cache default values are set based on the technology chosen in step 306 above. If the technology chosen by the user allows a choice of clock speed, the user is prompted (or the selection otherwise obtained) in steps 310 and 312. The clock period is then set per step 314.

Referring now to Figure 3b, the process 300 continues with steps 316 and 318, wherein the clock skew is selected (or otherwise defaulted). The extensions logic library name/location is then obtained from the user in step 320. As used herein, the term "extensions" refers to a sat of predetermined logical functions. For example, extensions within the ARC system may include rotate, arithmetic and logical shifts within the barrel shifter, small multi-cycle multiply, MAC function, swap function (for swapping upper and lower bytes), timer interrupt, and the like. Lastly, in steps 322-328, the need or user desire for a memory subsystem such as a scratchpad RAM is analyzed (based on the user's technology choice), and the memory subsystem values are set accordingly.

Referring now to Figure 3c, the method 300 continues with step 330, wherein the user's manufacturer code (specific to each user entity) and version are obtained from the user. In step 332, the memory bus and register file configuration required for the design are determined; see the discussion of Figure 4 herein for further details of this process. Once the memory/register configuration is determined, the user is prompted in step 334 to determine if the use of fast load returns is desired. As used herein, the term "fast load returns* refers to the condition in which the processor initiates a load operation but continues other execution before the load has been confirmed as being completed. Steps 336-338 allow the user to enable/disable this functionality, respectively, as desired.

Next in step 340, the system configures the extensions used with the design as specified in step 320 described above. Figures 5a-5h depicts this process in greater detail, as discussed further below. The user's choice of technology (step 306) is then examined in step 342 to determine if clock gating is allowed; if so, the user is prompted for a clock gating flag in step 344.

Referring now to Figure 3d, the user is next prompted for a series of inputs including the desired reset action (Le., "halt" or "run") per step 346, simulator choice (step 348), and pipeline display options (step 350). In step 352, the setup parameters for the register transfer link (RTL) and SeeCode Application Link ("RASCAL™") are defined, as discussed in greater d.tail below with reference to Figure 8. RASCAL1" is a utility module available from ARC Cores, Ltd. that connects the debugger module to the MTI simulator. This connection allows the debugger to control the simulation of code on the IC system model in the MTI simulator for hardware/software co-verification. The output directories for the files are then created in step 354.

Referring now to Figure 3e, the configuration is compiled, the core configuration registers and configuration register test code are set up per step 356. Next, in step 358, the functional HDL files are generated by (i) copying prototype files; (ii) substituting chosen and calculated values where required; and (iii) merging in the HDL where required. Note that in the illustrated embodiment, only the prototype files actually required, for the selected configuration are copied.

In step 360, the structural HDL files are generated, along with the synthesis scripts and simulation makefiles. Lastly, the test code required by the chosen extensions is compiled in step 361 Referring now to Figure 4, one exemplary embodiment of the method of determining the memory and register configuration of the system under design as shown generally in step 332 of Figure 3c is described. The method 400 comprises an initial determination in step 402 of whether the bus parameters are dictated by the technology previously chosen in step 306 (figure 3a). If the chosen technology requires specific bus values, these values are set in step 404. If not, (or alternatively, after setting the bus values in step 404), it is next determined whether core verification was requested by the user (steps 406, 407). If not, and the technology selected does not dictate the bus values, then the user is prompted to provide the desired bus widths in step 408. The memory values appropriate to the selected bus width are then set in step 410, and the user prompted for the desired register file configuration in step 4 2.

If core verification was requested by the user in steps 406 or 407, the memory values required for such verification are set in step 414, and the register values required by the selected technology and the core verification are set in step 416.

In the case that the technology selected does dictate the bus values, and core verification is not selected, then the register values are set per step 416 previously described.

Referring now to Figures 5a-5h, one exemplary embodiment of the method of configuring file extensions as shown generally in step 340 of Figure 3c herein is described. As illustrated in Figure 5a, the method 500 comprises first choosing one or more extensions library files in step 502 based on the technology initially selected and the memory subsystem configuration previously discussed. In step 504, the chosen extensions library file(s) are read. A list of available extensions compatible with the selected technology, memory subsystem configuration, and core verification choices is then built in step 506. Next, the user is prompted for specific core extension choices and associated parameters in step 508. Specifically, extensions include core registers, ALU extensions that provide more operations, auxiliary registers, condition codes, FIFO (first in-first out memory), increasing the number of pending load/stores that may be issued, and interrupts. After the user input has been received in step 508, each of the user core extension choices is analyzed to determine if it is compatible with the other choices made by the user (step 510). If so, the values required by those user selected core extensions are set in step 512. Note that the selection of some extensions will require the selection of other extensions which are needed for support functions. If not, the extension algorithm 500 is aborted in step 514, and restarted. It will be recognized that step 510 herein may be accomplished interactively, such that when the user is making selections in step 508, they will be immediately apprised of the compatibility of tha extensions chosen. Alternatively, the method 500 may be structured such that the system will not allow incompatible choices, either by prompting the user to make another selection, or only presenting the user with compatible choices to choose from, (such as on a display menu). Many other approaches are possible as well.

Referring now to Figure 5b, the method 500 continues with step 516, wherein the user's choice of technology is examined to determine whether it permits the choice of memory extensions (versus core extensions, as previously described). If no choice of memory extensions is allowed, the required memory extension values are set per step 518, the cache values set up per step 520, and the load/store interface configuration set up per step 522.

If a choice of memory extensions is allowed per step 516, the method of Figure 5c is utilized. As shown therein, the existence of a memory subsystem is next determined in step 524. If no memory subsystem exists, values consistent with no memory interface are set in step 526. Next, the user is queried if an instruction cache is desired in step 528; if no, the values for a core build only (i.e., no memory interface) are set in step 530. If yes, a cache random access memory (RAM) and direct mapped cache controller (DMCC) are added (step 532), nstruction fetch* extension is added to list of selected extensions (step 534), and the cache is set up (step 536), as further described with reference to Figure 6 herein. If a memory subsystem does exist in step 524, the desired memory extensions are obtained from the user in step 538.

Referring now to Figure 5d, the method 500 continues by setting up the load/store interface in step 540. The method of step 540 is described in greater detail with respect to Figure 7 herein. Upon completion of step 540, the algorithm determines in step 542 whether an instruction fetch was selected by the user. If so, the algorithm again determines whether a memory subsystem exists (step 544); if so, the user is queried in step 546 as to the type of instruction fetch system desired. If the type of instruction fetch system is a "cache", the algorithm then adds a cache RAM and DMCC to the configuration per step 548, and sets up the cache in step 550 as in step 536 previously described. If the type of instruction fetch selected in step 546 is a "dummy", then the algorithm proceeds to step 552 of Figure 5e. Note that if no instruction fetch was selected in step 542, or no memory subsystem exists in step 544, the algorithm proceeds to step 552 as well.

Referring now to Figure 5e, the method 500 continues with step 552 in which the initial choice of technology by the user (step 306) is examined to determine whether the choice of a memory arbiter is allowed. If yes, the existence of a memory subsystem is then determined in step 554. if no choice of memory arbiter is allowed, or if there is no memory subsystem, the algorithm then determines whether the choice of RAM sequencer is allowed for the chosen technology in step 556. Note that in step 544, if a memory subsystem is allowed, then the algorithm queries the user to specify the memory arbiter and channels desired.

Returning to step 556, if a RAM sequencer choice is permitted, then the algorithm determines in step 558 whether a synchronous RAM (SRAM) was chosen by the user. If so, the user is prompted in step 560 for the number of RAM sequencer wait states associated with the SRAM. Upon obtaining that information, the algorithm proceeds to step 562 of Figure 5f. Similarly, if no choice of RAM sequencer is permitted by the technology in step 556, the algorithm proceeds to step 562.

Referring now to Figure 5f, the extensions selection algorithm continues at step 562, wherein again the existence of a memory subsystem is determined. If a memory subsystem does exist, the existence of a memory arbiter is queried in step 564. If a memory arbiter exists, then the existence of a memory sequencer is determined in step 566. If a memory sequencer exists, the existence of an instruction fetch is determined in step 568. After step 568, the algorithm executes a branch return back to step 342 of Figure 3c. If no memory subsystem exists per step 562, a branch return to step 336 of Figure 3c is also executed, if no sequencer exists per step 566, the algorithm proceeds to step 574 of Figure 5g. if no instruction fetch exists per step 568, then the algorithm proceeds to step 576 of Figure 5g.

If no memory arbiter exists per step 564, then the existence of a sequencer is again determined per step 570. If a sequencer exists, the aforementioned return is executed. If no sequencer exists, the existence of an instruction fetch is determined in step 571 If an instruction fetch exists per step 572, the algorithm proceeds to step 580 of Figure 5h. If not, then step 582 of Figure 5h is selected.

Referring now to Figure 5g, step 574 comprises adding a dummy SRAM sequencer. After the dummy sequencer has been added, the algorithm determines whether an instruction fetch exists in step 578. If yes, a branch return is executed. If no, a dummy instruction fetch interface is added per step 579. If no instruction fetch exists per step 572 of Figure 5f, an instruction fetch channel is added to the arbiter per step 576 of Figure 5g, and a dummy instruction fetch interface added per step 579.

Referring now to Figure 5h, step 580 comprises adding an rniud) (external memory instruction fetch) interface for addressing external memory during instruction fetch. Alternatively, in step 582, an miu(4) interface and cache default values are added. Upon completion of step 580 or step 582, the algorithm determines in step 584 whether slide pointers or a scratch RAM was chosen by the user. If either was chosen, a scratchpad RAM is added per step 586, and slide pointers added to the build (as appropriate per steps 588 and 590.

Referring now to Figure 6, one exemplary embodiment of the method 600 of setting up the cache as previously discussed with respect to Figures 5b, 5c, and 5d herein is described. In a first step 602, the type of instruction cache to be used is obtained from the user. In the illustrated embodiment, the user may specify each, size, memory width, line size, and operation mode (such as debug and bypass), although other choices are possible. Next, in step 604, the values or the chosen cache type are read and stored as appropriate for later use.

Referring now to Figure 7, one embodiment of the method 700 of setting up the load/store interface per Figure 5d herein is described. First, the load/store memory controller size is obtained from the user per step 702; the values associated with the selected controller size are then obtained from the appropriate file and stared in step 704.

Referring now to Figure 8, one exemplary embodiment of the method 800 of setting up the register transfer language structure (e.g., RTL and SeeCode Link, or RASCAL™) as shown generally in Figure 3d herein is described. The method 800 first determines in step 802 whether a core verification was selected by the user; if a core verification was selected, a branch return to step 354 (Figure 3d) is executed. If no core verification was selected, the method 800 then determines what simulator choice has been made by the user in step 804. If no choice has been made, the system "testbench" function (a simulation environment with test vectors) is set to "standalone" mode, and the RTL and SeeCode Link disabled per step 806. If some choice (other than "none") has been made, the user is prompted in step 808 to determine whether the aforementioned RTL link is desired; if not, the testbench function is set per step 806. If the RTL link is desired, the existence of a memory subsystem is again determined per step 810, and the values for either a generic buiid with the RTL link (step 812) or the core build with RTL link (step 814) set as appropriate. Upon completion of the RTL setup method 800, a branch return to step 354 is executed.

Figure 9 illustrate, one exemplary embodiment of the method 900 of generating a new script according to the invention. First the hierarchy file representative of the physical hierarchy of the design is read per step 90Z Next, the applicable library file is read in step 904. In step 906, a ist of the libraries used in conjunction with the design is built as described in greater detail with reference to Figure 10 herein. After the libraries list is built an output makefile is created in step 908, and the constant headers written to the created makefile. In step 910, the aforementioned libraries Est is added to the makefile as described in detail with reference to Figure 11 herein. Next the "top-level" rule is added to the makefile in step 912. As used herein, the top-level rule relates to those rules specified by the hardware description language (such as VHOL or Verilog), although it will be appreciated that others rule bases may be used.

After the top-level rules have been edded, the datafile dependency rules are added to the makefile in step 914. This step 914 is described in greater detail with reference to Figure 12 herein. Lastly, the physical hierarchy generation control file ride is added to the makefile to complete the latter.

Referring now to Figure 10, one exemplary embodiment of the method 1000 of building a libraries list as shown generally in step 906 of Figure 9 is described. In a first step 002, each block present in the design is analyzed; if the read block's library is presently within the libraries list (step 1004), the next block is read and similarly checked. Note that as used herein, the term 'block* refers to a logical entity having a particular function, such as an adder or multiplier. If the required block library is not in the list that library is added per step 1006. This process 1000 is repeated until all blocks have been analyzed and the presence of their required libraries verified in the libraries list Referring now.to Figure 11, one exemplary embodiment of the method 1100 of adding the libraries list to the system "makefile* as shown generally in step 910 of Figure 9. Specifically, the method 1100 comprises reading each library per step 1102, and writing the library name and path to the aforementioned makefile per step 1104. The algorithm returns to step 912 of Figure 9 when all of tha required library information has been added to the makefile.

Figure 12 illustrates one embodiment of the method of adding datafile dependency rules to the system 'makefile* as shown generally in step 914 of Figure 9. Specifically, each block is read in step 1202, and the dependency rules are written to a file in step 1204 to extract that block from its library. When all dependency rules have been written, the algorithm returns to step 916 of Figure 9. This completes the script generation process 900.

Referring now to Figure 13, one exemplary embodiment of the method 1300 of hierarchy file generation according to the invention is described. In step 1302, the library file name and top level block name are obtained. Next in step 1304, the input files are read as described in detail with respect to Figure 14 herein. Next, the integrity of the data read is verified in step 1306; see Figure 19 for additional detail. Then, for each intermediate block, a subsumption map is built (step 1308}. In step 1310, an 1/0 list is built for each intermediate block as further described with reference to Figure 20 herein.. In step 1312, the "in*, 'out', and "inout" signals are written to a new data file for each block; the block ID and architecture type are also written thereto. Next, the HDL (e.g., "VHDLgen") control file for each block is written in step 1314. The foregoing three steps 1310, 1312, 1314 are subsequently performed for each intermediate block until all blocks have been processed; at this point the HDL control file for the top level is written in step 1316.

Next, for each intermediate black, the HOL is generated in step 1318, and when all intermediate blocks have been generated, the HOL for the top level generated in step 1 20. This completes the hierarchy file generation method 1300.

Figure 14 illustrates one embodiment of the method 1400 of reading input files (step 1304) in accordance with the hierarchy file generation method of Figure 13. First, the hierarchy file is read in step 1401 Next, the data file for the top level entity is read per step 1404. Then, for each primary block, the data file is read per step 1406. Figures 5, 16, and 7 illustrate these steps 140 404, 406 in detail Figure 15 illustrates one exemplary embodiment of the method 1500 of reading hierarchy files as shown generally in step 1402 of Figure 14. For each line in the hierarchy files, the associated block and list of dependent blocks is read in step 1502 and each added to the block list in step 1503. When all such lines have been read, the black list is then split into a separate (i) primary block list and IS) intermediate block list, in step 1504.

Referring now to Figure 16, one exemplary embodiment of the method 600 of reading top level entity data files as shown generally in step 1404 of Figure 14 is described. First, each One of the top level data file is analyzed in step 1602 to determine whether it describes a signal. If so, then the signal data is extracted from each fine per step 1604. In the illustrated embodiment, signal data includes, without limitation, "id," 'upper bound," "lower bound," "type," and "direction" information. Next, in step 1606, the direction of the signal in each line determined to have a signal present is analyzed; if the signal is outbound, the top level of the hierarchy is added as the destination of the signal, and the signal source set to "unknown" per step 1608. If the signal is not outbound, the signal's destination list is cleared, and the source of the signal set to the top level per step 1610. This process 1600 is repeated for each line in the top level data file until all lines have been analyzed.

Figure 17 illustrates one exemplary embodiment of the method 1700 of reading primary block data files as shown generally in step 1406 of Figure 14. Initially, for each primary block, each line of the associated top level data file is analyzed in step 1701 If the line under analysis is classified as a "signal" line, then the signal data is extracted from that line in step 1704 as previously described with reference to Figure 16. If the signal has been read before by the algorithm, and the values associated with the signal match those previously read, the signal is connected to the primary block per step 1710. If the signal has not been previously read, it is added to the signal list per step 708, and then connected to the primary block. Note also that if the signal has been read before but the current values do not match those previously read, then the algorithm aborts per step 1716.

If the line under analysis is an "architecture" line, the architecture of the primary block is set to "data" per step 1718, and the next fine analyzed.

If thB line under analysis is neither an "architecture" or "signal" line, no action is taken, and the next line in the top level data file is read.

Referring now to Figure 18, one embodiment of the method 1800 of identifying and integrating new signals into the block data files as shown generally in Figure 17 is described. Initially, the signal type is determined in step 1801 Signal types in the present embodiment include In", "out*, and "inout*. If the signal is an "in" signal the signal's source is set to "unknown" in step 1804; the block is then added as the destination of the signal per step 1805. If an "out" signal, the signal is examined to determine if a source already exists (step 1806); if yes, the algorithm 1800 is abortel If no, the signal's destination list is cleared and the signal's source set to "block" per step 1808. If the signal's type is Inout", then the block is set as the destination in step 1805.

Referring now to Figure 19, one exemplary embodiment of the method 1900 of verifying signal data integrity as shown generally in Figure 13 is described. First, in step 1902, each signal is examined to determine if it has an associated destination. If so, the existence of an associated source is then checked in step 1904. If both source and destination exist for all signals, the routine is terminated. If no destination is present for one or more signals in step 902, a list of signals without destinations is printed per step 1906. If all signals have destinations but one or more signals do not have a source, a list of signals without sources is printed per step 1908.

Referring now to Figure 20, one exemplary embodiment of the method 2000 of building an I/O list relative to the block data files as shown generally in Figure 13 is described. For a given block, the lists of "in", "out", and "inout" signals are first cleared per step 2002. Each signal is then checked to determine if it is an "inout" signal relative to the block in step 2004, as described in greater detail with reference to Figure 21 herein. For each signal with a -source inside the subsumption, the signal is checked in step 2008 to determine if it has a destination outside the subsumption. If so, the signal is added to the "out" list per step 2008, and the next signal analyzed. If the signal under analysis has no source inside the block, or all of its destinations are inside the subsumption, then the signal is checked to determine if it has an source outside of the subsumption in step 2010. If so, and it has at least one destination inside the subsumption (step 2012), then the signal is added to the "in" list per step 2014, and the next signal subsequently analyzed.

Figure 21 is a logical flowchart illustrating one embodiment of the method 2100 of checking for "in" or "out" signals as shown generally in Figure 20. For a given block and signal, the signal type is checked in step 2 02 to determine whether the signal is an "inout" type'; if the signal is not an Inout" type, the algorithm returns to Figure 20. If it is an "inout type, the signal's source is checked in step 2104 to determine if it is outside of the block, then whether it has destinations inside the block (step 2106). If both conditions are met, the signal is added to the inout list per step 2108. If either one of the conditions is not mat, the signal's source is checked in step 21 0 to determine if it is inside of the block; if so, then the destination is checked in step 2 2 to determine if it is outside of the block.

Referring now to Figure 22, one exemplary embodiment of the method 2200 of generating the merged hardware description language (HDL) file according to the invention is described.

It is again noted that while the following discussion is cast in terms of VHSIC Hardware description Language (VHDL), other languages may be used to practice the method of the invention with equal success VHDL is chosen for the illustrated examples simply for its comparatively wide-spread recognition and ease of use.

As used herein, the term "vhdlgen* refers to VHDL generation, a process within the invention by which the individual VHDL files relating to various aspects or components of the design are merged together to form a common VHDL file descriptive of the overall design.

As shown in Figure 21 the vhdlgen method 2200 comprises first reading a control file (step 22021, followed by reading the top level data file (step 2204). These steps 2202, 2204 are described in greater detail with reference to Figures 23 and 24, respectively. After the top level data file is read, each dependency present is analyzed by reading the dependency data file (step 2206), as further illustrated in Figure 25. After each dependency data file has been read, the signal fist is processed in step 2208 into a plurality of groups (Figure 27), the data integrity verified in step 2210 (Figure 28), the associated library file read in step 2212 (Figure 29), and the "merged" VHDL file written in step 2214 (Figure 30).

Referring now to Figure 23, one exemplary embodiment of the method 2300 of reading the HDL control file as shown generally in step 2202 of Figure 22 is described. Initially, the library file name is read from the control file in step 2302. Next, the top level entity name is read from the control file in step 2304. The control file is then examined in step 2306 to determine if it is at end of file (eof); if so, the routine is terminated and returns to step 2204 of Figure 21 If not. the first dependency entity name is read from the control file in step 2308 and added to the Est of dependent blocks in step 2309. If additional dependencies exist, they are subsequently read and added to the list of dependent blocks until the eof condition is met, at which point the routine returns to step 2204 of Figure 21 Figure 24 illustrates one exemplary embodiment of the method 2400 of reading the HDL top level entity data file as shown generally in step 2204 of Figure 21 Specifically, each record in the top level data file is examined in step 2402 to determine if it is a "signal" record; if so, the signal's upper bound, lower bound, direction, and type are set from the input record in step 2404. The signal's destination Gst is also cleared. Next each signal's direction (i.e., "in", "out", or "inout") is determined in step 2406; if the direction of a given signal is "out", then the signal's source is set to "unknown", and "PORT" is added to the signal's destination list in step 2410. If the signal is "in* or "inout", then the signal's source is set to "PORT" in step 2411 Lastly, in step 2414, the signal is added to the signal list, and the next record in the top level data file subsequently read.

Figure 25 illustrates one exemplary embodiment of the method 2500 of reading the HDL dependency data file as shown generally in Figure 21 For a given entity, each record in the data file is read and the type of record determined in step 2501 If the record is a "signal", data from the signal record is then processed in step 2504 (see Figure 26 for additional detaP on this step). After the data has been processed, the signal is added to the entity's signal list in step 2506, and the next record read. Conversely, if the record read is an "architecture" record, then the architecture of the dependency for that record is set per step 2510, and then a subsequent record read. When all records have been read, the routine is terminated in favor of step 2206 of Figure 25.

Figure 26 illustrates one embodiment of the method 2600 of processing signal record as previously discussed with respect to Figure 25. For the chosen entity, the algorithm 2600 first determines in step 2602 whether the signal being examined has been read before; if so, the signal's type and bounds are next examined to determine whether they match the previous occurrence of the signal (step 2604). If so, the signal's direction is next analyzed in step 2606 to determine its type, if the sipal is an "out* signal, the source of the signal is then analyzed in step 260S if a source exists, the routine is aborted per step 2610. If not, the signal's source is set to "entity" per step 2612. Conversely, if the signal is an "out" or Inout" signal, then the entity is added to the signal's destination Gst per step 2616.

In the case where the signal under examination has not been previously read, the signal's upper and lower bounds, direction, and type are set in step 2620 using information from the input line. The signal's destination Est is also cleared. If the signal's direction is "inout*. then the signal's source is set to "entity" in step 2622, and the signal added to the signal list in step 2624. If not, the signal is added to the signal Gst directly. After the addition of the signal to the signal Est in step 2624, the routine proceeds to step 2606 for analysis of the signal's direction as previously describe! Figure 27 illustrates one exemplary embodiment of the method 2700 of processing the signal list into groups as shown generally in Figure 22. Specifically, for each signal, the signal's source is first determined 'm step 2702; if the signal source is "PORT, then no further analysis of that signal is required, and the next signal is analyzed. If the signal's source is "unknown", then the signal is added to the list of signals with no source in step 2704, and the next signal analyzed. If the signal source is ""other", the signal is then analyzed in step 2706 to determine if any destinations are specified. If there are no destinations specified, the signal is added to the list of signals without destinations per step2710. If the signal does have one or more destinations, these destinations are then examined in step 2712 to determine if any are destinations other than "PORT". If so, then the signal is added to the list of intermediate signals per step 2716. Next, in step 271 B, the destinations of the signal are again examined to determine if any of these destinations are "PORT"; if yes, the signal is further added to the list of signals requiring output drives in step 2720.

Referring now to Figure 28, one exemplary embodiment of the method 2800 of verifying signal data integrity as shown generally in Figure 22 is described. First, in step 2802, each signal is examined to determine if it has an associated destination. If so, the existence of an associated source is then checked in step 2804. If both source and destination exist for all signals, the routine is terminated. If no destination is present for one or more signals in step 2802, a list of signals without destinations is printed per step 2806. If all signals have destinations but one or more signals do not have a source, a list of signals without sources is printed per step 2808.

Figure 29 illustrates one embodiment of the method 2900 of reading the library file as shown generally in Figure 22 herein. For each record present in the given library file, a determination is first made in step 2902 as to what type of record is under analysis; if the record is a "library id", then the status of setting the "worklib" flag is determined in step 2904. If the flag has not been set, it is set from the present record per step 2906. If it has been set, then the next subsequent record is read from the library. If the record is an "entity" record, the "entity id" and library id" are extracted from the record in step 2908, end the entity's library name is set to "library" in step 2910.

Lastly, if the record is classified as "other" in step 2902, then no action is taken and the next subsequent record is read.

After each record has been read from the selected library and analyzed as to type, each entity is examined to determine whether it has an associated entity library per step 2914. If a given entity does have a library, that library is next analyzed in step 2916 to determine if that library has been read before. If so, the next sequential entity is analyzed. If not, the library is added to the library Est in step 2918, end then the next subsequent entity analyzed. If the entity has no associated library in step 2914, then the program is aborted per step 2920.

Referring now to figure 30, one exemplary embodiment of the method 3000 of writing one or more HDL files as shown generally in figure 22 is described. The VHOL file is written in a sequence of steps that conform to the specification for VHDL format. It is noted that other HDL languages may employ differing formats that result in some information becoming unnecessary, or required in a different order. The first step 3002 of the illustrated method 3000 comprises writing the VHDL file header to a VHDL file. Next the list of libraries and packages used is written to the VHDL file in step 3004. Next the entity declarations are written to the VHDL file in step 3006. Subsequently the architecture information is written to the VHDL file in step 3008. In the next step 3010, VHOL component declarations are written to the VHDL file. The list of intermediate signals is then written to the VHDL file in step 30 2, and VHDL configuration statements are written to the VHDL file in step 3014. Then, in step 3016, all component instantiations are written to the VHDL file. Finally, output drives are written to the VHDL file in step 3018. This completes the hardware description language file generation process.

It is noted that while the foregoing description of Figures 2-30 refers to specific embodiments of the various aspects of the method of the invention, various substitutions, alterations, permutations, additions, and deletions may be made to the disclosed method. For example, certain routines disclosed herein may be selectively employed, or the order in which they are employed with respect to other routines varied. Similarly, individual steps within each of the routines may be modified as required by the particular application.

Description of the IC Design Apparatus Referring now to figure 31, an exemplary apparatus capable of implementing the method of Figures 2 -32 is described. A stand-alone microcomputer system 3100 of the type well know in the computer arts, having a processor 3102, internal or external storage device 3104, display device 3106, internal memory 3108, and input device 3110, is used. The algorithm 00 of Figure 2, embodied in the form of a computer program reduced to machine readable object code as is also well known in the art (not shown), is loader! intn the storage device 3104 and memory 3108 of the system 3100 and run on the processor 3102 based on inputs provided by the user via the input device 3110. Alternatively, the computer program may reside on a removable storage device (not shown) such as a floppy disk or magnetic data cartridge of thB type well known in the art. The display device 106, which may comprise for example a cathode ray tube (CRT), liquid crystal display (LCD), thin film transistor (TFT), or plasma display, provides the user with a variety of visual information including representations of the program's menu structure, and the status and output of various routines or modules running within the program. See Appendix I for one exemplary embodiment of the menu structure used in conjunction with the aforementioned computer program. While the illustrated embodiment is a microcomputer, it will be recognized that other computer types and architectures may be used with equal success. For example, a networked minicomputer system or client/server arrangement wherein data or libraries relating to the design are not resident on the local system 3100 may be employed, or alternatively, a plurality of different computers may ba used to run parts of the computer program in parallel. Many varying architectures and equipment/software configurations may be substituted, consistent with the general aims of providing an automated, interactive design process.

Description of the Integrated Circuit Referring now to Figure 32, an exemplary integrated circuit developed using the above-described method of the present invention is disclosed As shown in Figure 34, the integrated circuit 3200 is an ARC microprocessor-like CPU device having, inter alia, a processor core 3202, on-chip memory 3204, and an external interface 3206. The device is fabricated using the customized VHDL design obtained using the present invention (step 108 of Figure 2) which is subsequently synthesized into a logic level representation, and then reduced to a physical device using compilation, layout and fabrication techniques well known in the semiconductor arts. As previously noted, the designs for a wide variety of other types of integrated circuits (including, for example, ASICs and FPGAs) can be synthesized using the method and apparatus set forth herein. Hence, the CPU-based device of Figure 32 is merely exemplary of such devices.

It will be appreciated by one skilled in the art that the Integrated Circuit of Figure 32 may contain any commonly available microcontroller peripheral such as serial communications devices, parallel ports, timers, counters, high current drivers, analog to digital (A/D) converters, digital to analog converters (O/A), interrupt processors, LCD drivers, memories and other related peripherals. Further, the Integrated Circuit may also include custom or application specific circuitry that is specifically developed to solve a specific applications problem or meet the needs of a single application. The present invention is not limited to the type, number or complexity of peripherals and other circuitry that may be combined using the method and apparatus. Rather, any limitations are imposed by the physical capacity of the extant semiconductor processes which improve over time. Therefore it is anticipated that the complexity and degree of integration possible employing the present invention will further increase as semiconductor processes improve.

It is noted that power consumption of devices such as that shown in Figure 34 is significantly reduced due in part tn a lower gate count resulting from better block and signal integration. Furthermore, the above-described method provides the user with the option to optimize for low power (or use a low power silicon process such as Xemics for example). The system may also be run at a lower clock speed, thereby further reducing power consumption; the use of one or more custom instructions and/or interfaces allows performance targets to be met at lower clock speeds.

It is also noted that many IC designs currently use a microprocessor core and a DSP core. The OSP however, might only be required for a limited number of DSP functions, or for the IC's fast DMA architecture. The invention disclosed herein can support many DSP instruction functions, and its fast local RAM system gives immediate access to data. Appreciable cost savings may be realized by using the methods disclosed herein for both the CPU & DSP functions of the IC.

Additionally, it will be noted that the computer program as previously described herein can readily be adapted to newer manufacturing technologies, such as 0.18 or 0.1 micron processes, with a comparatively simple re-synthesis instead of the lengthy and expensive process typically required to adapt such technologies using 'hard" macro prior art systems.

While the above det-iied description and Appendices have shown, described, and pointed out novel features of the invention as applied to various embodiments, it will be understood that various omissions, substitutions, and changes in the form and details of the device or process illustrated may be made by those skilled in the art without departing from the spirit of the invention. The foregoing description is of the best mode presently contemplated of carrying out the invention. This description is in no way meant to be limiting, but rather should be taken as illustrative of the general principles of the invention. The scope of the invention should be determined with reference to the claims.

APPENDIX I The following describes and illustrates the operation of the ARC System Builder Program: Installation Script Installation script allowing the following ARC features to be selected: ?Extensions required TCache size ?Cache line length ?Si:e of external memory space that is to be cached.

?Clock period 7 Clock skew TSynthesized d-latches or 3-port RAM for register file TType of technology being used.

TManufacturer code and version number information The script creates a working directory for the user, and in it creates the following: VHDL for the extensions selected VHDL for a direct mapped i-cache configured to user's specification VHOL test-bench for testing basecase operation and external interfaces VHDL structure to link together all required modules Configuration files for Synopsys Design Compiler Memory image file containing basecase test code Synthesis script for Synopsys Design Compiler v3.4b or above Makefile - set up for the Model Technologies VSystem/VHDL simulator, but can be altered for use with other simulation environments.

The user can select from 4 different types of build: • Core Build • Generic System Build • Altera Build (for ARCAnger* development board) • Core Verification System After setting up his ARC area, before the user can use the ARC system builder, several environment variables need to be set In .cshrc Set ARCHOME environment variable, base of the arc install tree. Add bin directories to path. Set other environment variables such as ARC MANCODE and ARC MANVER and ARCASIC. (If these variables are not set the build script will ask for them at run time.) e.g. setenv ARCHOME fappslARC Cores setenv ARC MANCODE < Users Manu acturer code > setenv ARC MANVER < Users selected version number > set path = ($ARCHOME/arc/bin tpath) ARC MANCODE is the unique manufacturer code that ARC Cores gives to each ARC licensee, and the ARC MANVER is that manufacturers additional identity number. Both numbers go on to be stored as part of the identity register.

After setting up the environment, the ARC system builder can be run just by typing system_builder, which is an executable script. A log file (sys_bld.log), which keeps a record of the build selections the user make, is created in the directory that the user specify for the user working directory.

The ARC System builder script when run produces a series of text menus/questions.

Example MENU 1 1. Continue. 2. Valid selection. 3. * Mutually exclusive. 4. Mutually inclusive.

■ * Permanently selected.

N/A Invalid selection. 7. * Valid selection Please select : The above menu shows an example menu. The '" next to a menu entry indicates that it is selected.

Continue On all the menus that allow multiple selections 1 is always continue.

Valid selections In this example menu entry 2 is a valid selection which is currently unselected. Menu entry 7 is also a valid selection, which is selected. Valid selections can be selected and deselected by entering their corresponding menu number. (When menus are initially displayed some valid selections may be pre-selected; these are just recommended selections for this build.) Mutually exclusive selections (On the same menu] Selections are mutually exclusive, for example two types of the same extension cannot be selected at the same time. If one menu entry, of a mutually exclusive set, is already selected and the user then selects the another, the first is deselected.

Mutually inclusive selections (Between menus) Mutually inclusive relationships can exists between msnu entries on two separate menus. If this is so, the entry on the second menu is either permanently selected or invalid depending on the state of the first entry.

Permanent selections In certain circumstances menu entries usually valid select/deselect are permanently selected, e.g. Example menu entry 5. This usually because on a previous menu the user selected a menu entry to which this one is mutually inclusive. Permanent selections can also be active due to the build type.

Invalid selections As with permanent selections, invalid selections are normal valid selectideselect menu entries, however in this case thesB options are not available. This is usually due to either mutually inclusive selections on previous menus that were not selected, or this selection is not available on this type of build. Menu entry 6 is an example of an invalid selection.

Example MENU 2 1. Selection 1. 2. Selection 2. 3. Selection 3. 4. Selection 4.

Please select(3) : Menu Default Some menus in the ARC system builder have default selections. On such menus entering nothing, Le. just pressing enter, will select the default. The default menu entry number is specified next to the requestor. Menu defaults can be more than just numbers, on some questions they can be file paths, y/n etc.

The following describes examples for each of the four types of build available, Base case. Generic, ARCAngel and Core verification.

The first menu the user will encounter for all installations is the one below: ARC System Build Tool - ARC System install, (c) 1998 ARC Cores Technologies Ltd.

Generic install This program is capable of creating a basecase ARC or a complete ASIC for simulation, using MTi and synthesis using Synopsys Design Compiler/Design Expert.

We're going to make : 1. The directory structure 2. The structural VHDL levels 3. A personalized version of the cache config package 4. An optional external memory interface . Synopsys RAM models for the cache RAMs 6. Synthesis scripts 7. A simulation makefile But first, you must answer some questions : Please select the type of system you wish to build. 1. Core build with selected extensions 2. Generic System Build with selected core/memory extensions 3. Altera Build for ARCAngel Development Board 4. Core Verification System.

Please select (1) : This menu allows the user to select from four types of build Core (Base case) build, Generic, ARCAngel and Core Verification. We will now example an install of all four types, showing the menus the user will encounter and 'typical' responses. 1. Base ease From the first menu we select 1, which is also the default The ARC system builder will then prompt the user for details about the Core build, Le. Standard extensions (not including any scratchpad extensions), Direct mapped instruction cache, LoadJST interface to external memory, Host interface and model of external memory.

Instruction/Data byte Address selections -----■——-—·----— Instruction Fetch memory system byte address bus width 1.22 bit (4Mb) 2. 23 bit (8Mb) 3. 24 bit (16Mb) 4. 25 bit (32Mb) . 26 bit (64Mb) Please select (3) : This menu asks for the number of bits for the instruction fetch address. For the purpose of this example we will select the default of 24 bit ( 6Mb), by pressing ENTER.

The external memory system data bus width is 32 bits wide The external data bus on the ARC is 32-bits wide.

External memory system address bus width (22 < address < 32](24) : The external memory system data bus width can range from 22 to 32 bits, 4Mb to 4Gb address range, the default on this menu is set to the value of the l-f etch address bus width. We are going to select 29 bits, an address range of 512 Mb. NOTE: Even though the bus width is set to an arbitrary value the memory model for a core build simulation is 128Kbytes.

Core Register File Selection The 32x32 register file can be implemented using a synthesized array of d-latches, or by using a three-port (2 read, 1 write) synchronous RAM cell.

Use a synchronous 2r1w RAM for the register file? (y) : For the 32 general purpose registers rO r31 the user can either use : Synchronous 3-port (2r1w) RAM, (a fast write-through is not required).

Synthesized arrays of d-latches that require a write pulse to be generated.

Standard synthesis script assumes a 90°-delayed clock signal, for the synthesized array of d-latches.

We will use a synchronous 3-port RAM and so we select y. _ . Fast Load Returns — Enable Fast load returns? (n) : Fast Load returns, if the current instruction does not use register write back then the pipeline is not stalled.

Extensions Builder Select the extensions to be added to the ARC 1. Continue 2. 32x32 Barrel shifter · rotate, arithmetic and logical shifts · 1 cycle 3. 32x32 Barrel shifter · rotate, arithmetic and logical shifts - Multi-cycle 4. Small multi-cycle 32x32 Multiply . Faster (& larger) 32x32 Multiply 6. 16x16 Multiply/Multiply Accumulate function with 36 bit accumulator · Fast 7. 16x16 Multiply/Multiply Accumulate function with 36 bit accumulator · Slow 8. Swap function - swaps upper and lower 16 bits of 32 bit word 9. MIN/MAX function · 32 bit precision 0. Normalize function - finds first bit set in 32 bit word 1 . 24-bit Timer which counts to limit and then generates an interrupt N/A A Scratchpad RAM N/A A Scratchpad RAM with Sliding Pointers Please select : The above menu shows all the standard extensions, the user can select different combinations depending on his needs.

We will select the single cycle barrel shifter and the 32x32 multiply. Entering 2 and then entering 4 does this. If we wanted to disable the multiply we could do this by entering 4 again.

Note the user cannot select both types of barrel shifters, in an install, selecting one automatically deselects the other. This also applies to both instances of the 32x32 multiplier, the MAC and both instances of the Scratchpad RAM. We will now enter 1 to continue.

- Technology Selection Feature size (drawn) of technology in use 1. 0.35um 2. 0.5um 3. 0.65um 4. 0.8um . 1.0um Please select (5) : The feature size provides timings for DATA / TAG ram and the core registers. Our chosen technology is 1.07m so we press ENTER to continue.

The default synthesis script is set-up for the 1.0?m ISM OK library supplied with the Synopsys synthesis tool.

These values are only used to set the timings for the synthesis of RAM models. If an unsupported feature size required the user can select an arbitrary value and following the build change the clock period for the simulation of the VHDL and synthesis scripts to support the new feature size, timings of RAM etc.

Target clock speed in MHz (20) We will use the default of 20MHz as we have specified 1.0 ?m technology. Clock speed has a minimum of 1 MHz and a Maximum of 150 MHz.

Clock skew for synthesis (+/-1.5ns) : The default of 1.5ns will be used for this example. Clock Skew has a minimum of 0 and a maximum less than the clock period.

Library for extensions logic? (user) : Here the user can specify a name for his user's library. This name will be used as the VHDL library names and also during the install when creating directories and copying subsequent files to these directories (The directories created are in upper case). The name entered is arbitrary so we will enter 'Alex'.

... . Memory Extensions Do you wish to include a Direct-Mapped Instruction Cache? (y) : If the user enters "no" here the user will build an ARC with no cache, if the user enters "Yes" he will be prompted to for further information about the cache he wants. We will enter "Yes" the default.

Ifetch Modes Select the instruction Cache Modes. 1. Continue 2. Standard Cache with debug and cache bypass capability. 3. Virtual Cache 4. Mixed Code RAM / lcache space & Cache line lock The first thing the script needs to know is what kind of cache do the user want. It's possible to have a : 1. Standard Cache 2. Virtual cache allows a user to dynamically resize the cache in software. 3. Mixed Code RAM / lcache space with Line locking NOTE: Standard Cache and Virtual Cache are mutually exclusive, and Mixed Code RAM is a superset of Standard Cache. We want to try a range of cache sizes to see what cache size best suits our application, therefore we are selecting Virtual Cache Please select the max size of the Direct Mapped Instruction Cache. 1.0.5k bytes, 128 words 2. Ik bytes, 256 words 3.2k bytes, 512 words 4.4k bytes, 1024 words . 8k bytes, 2048 words 6. 16k bytes, 4096 words 7. Other size Please select (3) : The Direct Mapped Instruction cache is user definable, using the script the user can select sizes from 4 bytes (1 word), up to 512k bytes (128k words). There are shortcuts for 512 bytes to 16K bytes, but other sizes (larger or smaller) can be entered manually using option 7. For our test 4K bytes (selection 4) is sufficient.

Please select the min size of the Direct Mapped Instruction Cache. 1. 0.5k bytes, 128 words 2. Ik bytes, 255 words 3. 2k bytes, 512 words 4. 4k bytes, 1024 words - 8k bytes, 2048 words • 16k bytes, 4096 words 7. Other size Please select (3) : Virtual Cache requires a range of possible caches and line lengths so a minimum cache size is required. We will enter 2 512 bytes.

Select a max Cache line, length 1. 2 instruction words 1 4 instruction words 3. 8 instruction words 4. 6 instruction words . 32 instruction words 6. Other Please select (3) : The cache line length is the number of wards fetched when a cache miss occurs. The script has options for 2, 4, 8, 6, or 32 word lines, the default being 8 words. Selecting 4, will give a cache line length of 16 words, suitable for our test Select a max Cache line length 1. 2 instruction words 2. instruction words 3. 8 instruction words 4. 16 instruction words • 32 instruction words 8. Other Please select (3) : Virtual Cache requires a range of possible caches and line lengths so a minimum cache line length is required. Entering 1, selects two instruction words. Cache size, line length and external address all determine the size of the tag RAM. Larger cache means more tag words and fewer tag bits. Longer line length means fewer tag words and more tag bits.

........ - Simulation — ..... .....

Selection of the state of the ARC on reset is important since it is preferable to have the system build start on reset while the core build should halt on reset since the host effectively clears the halt bit during simulation.

Do you wish to halt the ARC at address location 0 on reset (y) : We have entered V here, that means the ARC Is halted immediately on reset. Had we selected 'n' the ARC starts running the code at address 0 on reset.

The default for this is Y on core build and N on generic and ARCAngel builds.

Note you will have to set the status register to the appropriate start address and clear the halt bit through the host interface before running the simulation.

You can manually edit "arc_start" in xaux_regs.vhdl so that the ARC is running following a reset.

Having selected 'γ' the above information message is displayed.

Is your VHDL simulator Model Technologies VSystem? (y) : At ARC Cores we enter Y as we use TI VSystem to simulate our VHDL If the user use a different simulator, select no. The user may set his environment up for his simulator by editing the makefile and other support files, after the install has completed.

Da you wish to use the R.T.L And SeeCode Application Link (RASCAL)? (n) : RTL And See-Code Application Link, otherwise known as RASCAL™, is an extra utility that connects the Metaware debugger to connect to the Modeltech simulator. This allows the debugger to control the simulation of code on the ARC system model in the MTI simulator for hardware/software co-verification. The default is no which is what we want for this build. _ Install Directory -.

Full path to ARC usei working directory (/home/someone/basearc) : At this prompt the user have to specify a destination directory name (with full path).

The path must not include /rmp_mnt/ etc.

We will select the default, a directory named "basearc" created where we launched the ARC system builder.

What we have built l-fetch address bus width 24-bit external address bus width 29-bit register file 3-port RAM Fast Load returns 'no' Extensions Single Cycle Barrel shifter Small multi-cycle 32x32 multiply Clock speed 20 MHz .

Clock skew 1.5 ns Technology size 1.07m Users library name 'alex' Direct mapped instruction Virtual Max/ Min instruction cache size 4K bytes 5 2 bytes Max/ Min Cache line length 16 instruction words 2 instruction words Halt ARC on Reset? 'yes' Destination directory /home/user/basearc Simulator - MTI? 'yes' RASCAL 'no' 2. Generic The second selection from the first menu is the generic build. This is similar to the base case build however it allows the selection of memory extensions. As described in the generic build certain selections determine which menus the user will see, we are going to make selections in this example that will give access to all menus. -.-.»._._.»._...- instructionJData byte Address selections ------- -Instruction Fetch memory system byte address bus width 1.22 bit (4Mb) 2. 23 bit (8Mb) 3.24 bit (16Mb) 4. 25 bit (32Mb) . 26 bit (64Mb) NOTE: if you intend to use the PC parallel communications host port do not exceed 24-bits Please select (3) : This menu asks for the number of bits for the instruction fetch address. For the purpose of this example we will select 22 bit (4Mb), by entering 1.

NOTE: Later in the build the user will be asked to select a host port far communications with the ARC. The SUN/PC parallel communications host port has maximum of 24-bits, if the user select a value greater than 24-bits in this menu the PC host port will no longer b. available.

The external memory system data bus width is 32 bits wide The external data bus on the ARC is set to 32 bits wide.

External memory system byte address bus width (22 < address < 321(22) : The external memory system data bus width can range from 22 to 32 bits, 4Mb to 4Gb address range, the default on this menu is set to the value of the l-fetch address bus width. We are going to select 24 bits, 16Mb address range, which is the maximum valid selection that can be used with the PC/SUN host port NOTE: Even though the bus width is set to an arbitrary value the memory model for a generic build simulation is 512Kbytes.

Core Register File Selection The 32x32 register file can be implemented using a synthesized array of d-latches, or by using a three-port (2 read, 1 write) synchronous RAM celL Use a synchronous 2r1 w RAM f or the register file? (y) : For the 32 general purpose registers r0-r31 the user can either use : Synchronous 3-port (2r1w) RAM, where a fast write-through not required, or Synthesized arrays of d-latches that require a write pulse to be generated.

Standard synthesis script assumes a 90°-delayed clock signal.

We desire a synchronous 3-port RAM and so we select y.

Fast Load Returns Enable Fast load returns? (n) : Fast Load returns, if the current instruction does not use register write back then the pipeline is not stalled. Our design allows us to take advantage of fast return loads and therefore we enter 'yes'.

Extensions Builder Select the extensions to be added to the ARC I. Continue 2. 32x32 Barrel shifter · rotate, arithmetic and logical shifts ■ 1 cycle 3. 32x32 Barrel shifter · rotate, arithmetic and logical shifts - Multi-cycle 4. Small multi-cycle 32x32 Multiply . Faster (& larger) 32x32 Multiply 6. 16 16 Multiply/Multiply Accumulate function with 36 bit accumulator - Fast 7. 16x16 Multiply/Multiply Accumulate function with 36 bit accumulator - Slow 8. Swap function · swaps upper and lower 16 bits of 32 bit word 9. MIN/MAX function - 32 bit precision 0. Normalize function - finds first bit set in 32 bit word I I . 24-bit Timer which counts to limit and then generates an interrupt 11 A Scratchpad RAM 13. A Scratchpad RAM with Sliding Pointers Please select : The above menu shows alt the standard extensions, the user can select lots of different combinations depending on his needs.

We will select multi-cycle barrel shifter, Fast Mul/ AC, scratchpad RAM with sliding pointers and swap. Entering 3,7,12 and 8 does this. We really wanted a Normalize extension rather than Swap. We disable Swap by entering 8 again and then enter 0 to select normalize.

Note that the user cannot select both types of barrel shifters, in an install, selecting one automatically deselects the other. This also applies to both instances of the 32x32 multiplier, the MAC and both instances of the Scratchpad RAM.

We will now enter 1 to continue.

Technology Selection Featura size (drawn) of technology in use 1. 0.35um 2.0.5um 3.0.65um 4.0.8um . l.Oum Please select (5) : The feature size provides timings for DATA f TAG ram simulation & synthesis models and the core registers. Our chosen technology is 0.35?m so press 1 to continue, if vendor models are to be used the timings these settings produce are not used These values are only used to set the timings for the synthesis of RAM models. If an unsupported feature size required the user can select an arbitrary value and following the build change the clock period for the simulation of the VHDL and synthesis scripts to support the new feature size, timings of RAM etc.

Target clock speed in MHz (100) We will use the default of lOOMhz as we have specified 0.35 ?m technology.

Clock skew for synthesis (+ /-0.4ns) : The default of 0.4ns will be used for this example.

Library for extensions logic? (user) : Here the user can specify a name for his user library. This name will bB used as the VHDL library names and during the install when creating directories and copying subsequent files to these directories. The name entered is arbitrary so we wilt enter 'sa '.

Memory Extensions .......

Select the memory extensions to be added to the ARC interface 1. Continue • * Load/Store memory controller 3. * Instruction fetch memory controller 4. * Host interface for communications with ARC . * Arbitration unit for memory accesses 6. * RAM Sequencer Please select : Here the memory extensions for off chip memory are selected. The LD/ST memory controller is not optional, and is therefore permanently selected. To show the full use of the build script we ere selecting all memory extensions. To do this there must a '*' next to each number, if not type that number (and pressing enter) , finally enter 1 to continue.

The queue depth for the Load/Store Memory Controller is fixed, Le.4 deep with a max of 2 stores.

Having continued the above information message is displayed.

■ Ld/St Memory Controller -Select RAM for local load/store operations. 1. No Local RAM 2.512 x 32 bit of Local RAM 3. 1024 x 32 bit of Local RAM 4. 2048 x 32 bit of Local RAM .4096 x 2 bit of Local RAM Please select (3) : The size of RAM to be used for local memory can be specified. We want 8 Kbytes of 32 bit local ram and so enter 4. The local RAM is situated at the top of memory by default (determined by the external address width) and therefore has a base address of [Maximum RAM size - Local RAM size].

I-Fetch Memory Controller -Select type of Instruction Fetch memory system 1. Dummy l-fetch unit for User Definition L Direct Mapped Cache Please select (2) : There are several different ways in which the instruction fetch memory system can be defined. 1. Dummy l-f etch, creates a template VHOL file for the user to edit. 2. Direct mapped cache.

We are using the default, 2, Direct Mapped Cache.

Ketc Modes --- — Select the instruction Cache Modes. 1. Continue 2. Standard Cache with debug and cache bypass capability. 3. Virtual Cache 4. Mixed Code RAM I lcache space & Cache line lock I S Please select : The first thing the script needs to know is what kind of cache do the user want. It's possible to have a : 1. Standard Cache 2. Virtual cache allows a user to dynamically resize the cache in software. 0 3. Mixed Code RAM / lcache space with Line locking We have previously looked at using Virtual Cache to experiment with a range of cache configurations. Using this method we have found a standard cache configuration that gives the required performance to cost ratio for our application. We are therefore selecting Standard Cache.

Please select the max size of the Direct Mapped Instruction Cache. 1.0.5k bytes, 128 words 1 1k bytes, 256 words 3.2k bytes, 512 words 4.4k bytes, 1024 words 0 5. 8k bytes, 2048 words 6. 16k bytes, 4096 words 7. Other size Please select (6) : The Direct Mapped instruction cache is user definable, using the script the user can input select sized from 512 bytes (128 words), up to 16k bytes (4k words). Other sizes (larger or smaller) can ba manually input using option 7. For our test 8K bytes (selection 5) is sufficient.

Please select the min size of the Direct Mapped Instruction Cache. 1. 0.5k bytes, 128 words 2. k bytes, 256 words 3. 2k bytBS, 512 words 4. 4k bytes, 1024 words . 8k bytes, 2048 words - 16k bytes, 4098 words 7. Other size Please select (1) : As we are using a virtual cache, it is necessary to specify a lower limit for the cache size. When using a virtual cache it is possible to artificially set the cache to any size between the maximum and minimum using software. For this test we want to test a large range of cache sizes down to the very small, therefore we select 7.

Please enter the min size of the Direct Mapped Instruction Cache in bytes.

NOTE: Size must be a power of two also 4 < - Size < - 6384 Enter Size : As we are testing a large range of caches we havB decided to choose 128 bytes as our minimum cache size. Entering 128 and pressing enter sets this constant At this menu WB can use abbreviations, 16k for example is the same as entering 16384.

Select a max Cache line length 1. 2 instruction words 2.4 instruction words 3. 8 instruction words 4. 16 instruction words . 32 instruction words 6. Other size Please select (3) : The cache line length is the number of words fetched when a cache miss occurs. The script has options for 2, 4, 8, 16, or 32 word lines, the default being 8 words. Typing ENTER here, like all other menus, selects the default, 3, giving a cache line length of 8 words.

Select a min Cache line length 1.2 instruction words 2.4 instruction words 3. 8 instruction words - 16 instruction words - 32 instruction words 6. Other size Please select (3) : When using a virtual each., as with the cache size, the user must also specify a minimum cache line length. 1 will give a minimum line length of 2 instruction words, which is the choice we have made.

Cache size, line length and external address, all go to determine the size of the tag RAM. Larger cache means mere tag words and fewer tag bits. Longer line length means fewer tag words and more tag bits.

--- --- - -- - - - - Host Interface Communications - -- - - - - - - - - - - - - —Select from available host communications ports. 1. Dummy Host Communications Port 2. Standard PC/SUN Parallel Communications Port 3. JTAG Communications Port Please select (2) : Host port for I/O communications can be either a dummy VHDL file, which the user can edit to create a user defined communications port, standard PC/SUN or JTAG communications ports. We are selecting 2, PC/SUN, as we are not using our ARC with custom hardware. .... -. _ > -, _ .... Memory Arbitrator ...... ............ Select the channels that you would like the memory arbitrator to service. 1. Continue ■ * Load/Store channel for external memory accesses.

Host Interface channel for debugging and data downloading.

• * Instruction fetch channel for l-cache.

■ * Local Scratch-pad RAM channel for burst reads/writes. 6. Additional Channels Please select : How this menu looks is very dependent on the selections that have been made before, as most options are mutually inclusive with other selections. For test purposes we are adding some additional channels 6.

How many additional channels do you want? (1 - 10) : The user can select from between 1 and 10 extra channels. Enter 5 to give 5 additional channels. 1. Continue ■ * Load/Store channel for external memory accesses.

· * Host Interface channel for debugging and data downloading.

- * Instruction fetch channel for l-cache.

• * Local Scratch-pad RAM channel for burst reads/writes. 6. * Additional Channels(5) Please select : The number of additional channels is now displayed next to the option. We enter 1 to continue to the next menu.

-- - - - - - ---- - - - - - -- - - - - - RAM Sequencer —Select the number of wait states for the RAM Sequencer. The script does not allow the user to enter a value greater than 31. If you wish use more than 31 wait states then you will have to edit the appropriate files manually.

How many wa t states? (31) : Choose from 0 - 31 RAM sequencer wait states. The default and the value we are using is 31. This is a good default, as it will produce an ARC that will run with almost any speed memory. For simulation purposes, or if the user know that the ARC the user are building will always use a certain speed RAM, it may be useful to set the default value for the wait states to a smaller number than the maximum 31.

Simulation Selection of the state of the ARC on reset is important since it is preferable to have the system build start on reset while the core build should halt on reset since the host effectively clears the halt fait during simulation.

Do you wish to halt the ARC at address location 0 on reset (n) : We have entered 'n' here, that means the ARC starts running the code at address 0 on reset. Had we selected 'y' the ARC is halted immediately on reset.

The default for this Y on core build and N on generic and ARCAngel builds.

Is your VHDL simulator Model Technologies VSystem? (y) : ARC Cares Technologies we enter V as we use MTI VSystem to simulate our VHDl If the user use a different simulator, select no. The user may set his environment up for his simulator by editing the makefile and other support files, after the install has completed.

Do you wish to use the R.T.L And SeeCode Application Link (RASCAL)? (n) : RTL And See-Code Application Link, otherwise known as RASCAL™, is extra utility that connects the Metaware debugger to connect to the Modeltech simulator. This allows the debugger to control the simulation of code on the ARC system model in the MTI simulator for Hardware/Software Co-Verification.

To include RASCAL we enter y and press enter. -.- install Directory -·-Full path to ARC user working directory (/home/someone/systemarc) : At this prompt the user have to specify a destination directory name (with full path).

The path must not include /tmp_mnt/ etc.

We will select the default, a directory named "systemarc" created where we launched the ARC system builder.

What we have built I fetch address bus width 22-bit External address bus width 24-bit Register file 3-port RAM Fast load returns 'yes' Extensions Multi-cycle barrel shifter.

Fast MuliMAC.

Scratchpad RAM with sliding pointers.

Technology size 0.35 ?m Clock speed 100 MHz Clock skew 0.4 ns Users library name 'sam' Memory Extensions Load/Store memory controller Instruction fetch memory controller Host interface for communications with ARC Arbitration unit for memory accesses RAM Sequencer Ld/St memory size 8 Kbytes.

Type of I Caehe Direct Mapped Cache l-Cache mode Virtual Cache l-cache size. Max / Min 8K bytes 128 bytes Cache line length Max / Min 8 instruction words 2 instruction words Host communication port Standard Sun/PC Arbitration Channels Load/Store channel for external memory accesses Host l/f channel for debugging and data downloading Instruction fetch channel or l-cache Local Scratch-pad RAM channel for burst reads/writes Additional Channels (5) Ram Sequencer Wait states 31 Halt ARC on Reset? 'no' Destination directory /homefuser/systemarc Simulator - MT1? 'yes' RASCAL? 'yes' 3. ARCAngel The final build is the ARCAngel. A large number of the selections on the ARCAngel build are chosen automatically due to the nature of the ARCAngel development board. The ARC Angel development board is described in detail in ", which is incorporated herein by reference in its entirety. ... ..... _ ast (.pad Returns ...............

Enable Fast load returns? In) : Fast Load returns, if the current instruction does not use register write back then the pipeline is not stalled.

ARCAngel development allows the use of fast return loads and so will our final design, therefore we enter 'yes'.

. Extensions Builder Select the extensions to be added to the ARC I. Continue 2. 32x32 Barrel shifter · rotate, arithmetic and logical shifts - 1 cycle 3.32x32 Barrel shifter ■ rotate, arithmetic and logical shifts · Multi cycle 4. Small multi-cycle 32x32 Multiply N/A Faster (& Larger) 32x32 Multiply N/A 16x16 Multiply/Multiply Accumulate function with 36 bit accumulator · Fast 7. 6x 6 Multiply/Multiply Accumulate function with 36 bit accumulator - Slow 8. Swap function · swaps upper and lower 6 bits of 32 bit word 9. MIN/MAX function - 32 bit precision 0. Normalize function - finds first bit set in 32 bit word I I. * 24-bit Timer which counts to limit and then generates an interrupt N/A A Scratchpad RAM N/A A Scratchpad RAM with Sliding Pointers Note the multi-cycle barrel shifter has been selected along with the 24-bit timer.

These have been selected as part of the standard Altera configuration, however, the system will still function correctly without them.

Please select : The above menu shows all the standard extensions, the user can select different combinations depending on his needs.

Note: The scratchpad RAM and the single cycle barrel shifter are too large to fit on the current ARCAngel FPGA, as are the fast 32x32 multiplier and the fast MUUMAC. Also a build with all extensions selected will not fit on the current FPGA.

The 24-bit Timer is automatically selected, as it is a common selection on an ARCAngei build. We are using this extension however the user can deselect it by entering the corresponding number. We will also select the Multi-Cycle Barrel shifter and the Small 32x32 multiply. Entering 3 then 4 does this.

Note the user cannot select the fast type of MulfMAC, in an install, if the user attempt to select it the ARC system builder prompts the user for another response.

We will now enter 1 to continue. (fetch Modes Select the instruction Cache Modes. 1. Continue 2. 'Standard Cache with debug and cache bypass capability. 3. Virtual Cache 4. Mixed Code RAM / lcache space & Cache line lock Please select : The first thing the script needs to know is what kind of cache do the user want. It is possible to have a : 1. Standard Cache 2. Virtual cache allows a user to dynamically resize the cache in software. 3. Mixed Code RAM f lcache space with Line locking Selecting 3 allows standard Cache with the extra features of Code RAM and Line locking.

Please select the norm size of the Direct Mapped Instruction Cache. 1.0.5k bytes, 12B words 2. Ik bytes, 256 words 3.2k bytes, 512 words 4.4k bytes, 1024 words . Bk bytes, 2048 words 6. 16k bytes, 4096 words Please select (3) : The Direct Mapped Instruction cache is user definable, using the script the user can input select sized from 512 bytes (128 words), up to 16k bytes (4k words). Other sizes (larger or smaller) can be manually instantiated but for our test 4K bytes (selection 4) is sufficient.

The ARCAngel has onboard tag-RAM but uses external memory for the cache data-RAMS, so selecting larger cache sizes does not drastically affect the amount of RAM resources available on the Altera device.

Cache line length 1. 2 instruction words 2. instruction words 3. 8 instruction words 4. 16 instruction words . 32 instruction words Please select (3) : The cache line length is the number of words fetched when a cache miss occurs. The script has options for 2, , 8, 16, or 32 word lines, the default being 8 words. Entering 2 will give a cache line length of 4 words, suitable for our test.

- - - - - - - - - - - - - Host Interface Communications - - - - - - - - - - - - - - - —Select from available host communications ports. 1. Standard PC/SUN Parallel Communications Port 2. JTAG Communications Port PleasB select ( 1 : Host port for I/O communications can be either a standard PC/SUN or a JTAG communications port. We are selecting 2, JTAG, as we wish to have an ARCAngel that can communicate to a range of host machines via the industry standard JTAG.

Simulation Selection of the state of the ARC on reset is important since it is preferable to have the system build start on reset while the core build should halt on reset since the host effectively clears the halt bit during simulation.

Do you wish to halt the ARC at address location 0 on reset (n) : We have entered V here, that means the ARC starts running the code at address 0 on reset. Had we selected 'y' the ARC is halted immediately on reset.

The default far this Y on core build and N on generic and ARCAngel builds.

Is your VHDL simulator Model Technologies VSystem? (y) : At ARC Cores we enter 'y' as we use ΜΤΊ VSystem to simulate our VKDL If the user use a different simulator, select no. The user may set his environment up for his simulator by editing the makefile and other support files, after the install has completed.

Do you wish to use the R.T.L And SeeCode Application Link (RASCAL)? (n) : RTL And See-Cade Application Link, otherwise known as RASCAL™, is extra utility that connects the Metaware debugger to connect to the Modeltech simulator. This allows the debugger to control the simulation of code on the ARC system model in the MTI simulator for Hardware/Software Co-Verification.

The default of n is selected by simply pressing enter. install Directory Full path to ARC user working directory (/home/someone/arcangel) : At this prompt the user have to specify a destination directory name (with full path).

The path must not include /tmpjnnt/ etc We will select the default; a directory named "arcangel" created where we launched the ARC build script NOTE: Even though the bus width is set to 26-bits, the memory model for a ARCAngel build simulation is 512Kbytes.

What we have built Fast Load returns? 'yes' Extensions Multi-cycle barrel shifter. 24-bit timer.

Small 32x32 multiplier Direct mapped instruction cache size 4K bytes Cache line length 4 instruction words Host interface controller JTAG Ram Sequencer SRAM 0 Wait states Halt ARC on Reset? 'no' Destination directory /home/userfsystemarc Simulator - MTI? 'yes' RASCAL? 'no' 4. Core Verification system The Core verification is a build which creates an ARC system and appropriate test files to verify that the ARC VHDL behaves as described in the ARC Programmers Reference manual.

„ Core Register File Selection The 32x32 register file can be implemented using a synthesized array of d-latches, or by using a three-port (2 read, 1 write) synchronous RAM cell.

Use a synchronous 2r1w RAM for the register file? (y) : For the 32 general purpose registers rtk31 the user can either use : Synchronous 3-port (2r1w) RAM, where a fast write-through not required, or Synthesized arrays of d-latches that require a write pulse to be generated.

Standard synthesis script assumes a 90°-delayed clock signal.

We selected a synchronous 3-port RAM and so we select y.

- Technology Selection Feature size (drawn) of technology in use 1. 0.35um 2. 0.5um 3. 0.65um 4.0.8um . 1.Oum Please select (5) : The feature size provides timings for DATA I TAG ram simulation & synthesis models and the core registers. Our chosen technology is 0.657m so press 3 to continue.

Target clock speed in MHz (50) We will use the default of 50Mhz as we have specified 0.65 ?m technology.

Clock skew for synthesis (+ /-0.4ns) : The default of +1- 0.7ns will be used for this example.

Library for extensions logic? (user) : Here the user can specify a name for his user's library. This name will be used during the install when creating two directories and copying subsequent files to these directories. The string entered is arbitrary so we will enter 'torn'.

Simulation ———.-.—-.——.—·—— Selection of the state of the ARC on reset is important since it is preferable to have the system build start on reset while the core build should halt an reset since the host effectively clears the halt bit during simulation.

Do you wish to halt the ARC at address location 0 on reset (y) : We have entered V here, that means the ARC is halted immediately on reset. Had we selected 'η' the ARC starts running the code at address 0 on reset.

The defaults are Ύ for. a core build and 'N' for generic and ARCAngel builds.

Note you will have to set the status register to the appropriate start address and clear the halt bit through the host interface before running the simulation.

You can manually edit "arc_start" in xaux_regs.vhdl so that the ARC is running following a reset Having selected 'y' the above information message is displayed.

Is your VHDL simulator Model Technologies VSystem? (y) : At ARC Cores we enter 'y' as we use MTI VSystem to simulate our VHDL If the user use a different simulator, select no. The user may set his environment up for his simulator by editing the makefile and other support files, after the install has completed. install Directory Full path to ARC user working directory (Jhome/someone/verifyarc) : At this prompt the user have to specify a destination directory name (with full path).

The path must not include ftmpjnnt/ etc.

We will select the default, a directory named "verifyarc" created where we launched the ARC system builder.

What we have built: Technology Feature Size 0.65um Clock Speed in MHz 50 MHz Clock Skew in ns +/· 0.7 ns Halt ARC on Reset? 'yes' Destination directory /home/user/verifyarc Simulator - MTI? 'yes' Once the script has completed the install, the base case / ARCAngel build generates files for simulation and synthesis. The Generic builder can generate files for simulation provided all the required memory extensions are selected. For all types of build the VHDL that the user can modify has been included in the directory 'ΜάΙ' As default the test code is linked to the tests: -Core/ Generic & ARCAngel builds : init_mem.hex Core Verification build : testscript.hcs* Type of Tests Hex Directories ALU and interrupts Core b Extension Core Interfaces Core_x Extension interfaces Ext_x Host interface m Auxiliary logic Aux_x Base Case ' For simulation a makefile is auto-generated to compile all the selected VHDL for the MTI VSystem simulator. Type 'make' to build database, see makefile header for information on using the makefile with other simulators. To simulate, simply type 'coretest'. This runs the Model Technologies simulator from the command line. The behavior of the ARC core is displayed in the form of a pipeline diagram (SEE FIGURE 3) that shows the state of the processor as it processes each instruction. The user is made aware when host interface accesses are taking place, in addition to Load/Store from external memory. NOTE: The code will be executed twice for the default set-up since it is run in normal mode and also in single step mode where each instruction is single stepped through the pipeline. The single step feature is only executed upon successful conclusion of the test code in normal operational mode.

Upon conclusion the user is made aware whether the test being simulated passed or failed.

To synthesize, type 'syn_arc', which runs the synthesis scripts for Synopsys Design Compiler.

Simulation Display Example of Pipeline diagram. 8 12975ns : Z 000001 AO 7FFFFFFF 8 13025ns : Z 000001 AO 7FFFFFFF 8 13075ns : Z 000001 AO 601 F7C00 mov 8 13125ns : Z 000001A48000000080000000 mov — # 13175ns : Z 000001A848000100 rlc.f mov -8 13225ns : Z 000001 AC 601FFE00 mov rlc.f wb : rO < - 8Q00000Q 8 13275n : Z 000001 BO 48000100 rlc.f mov rlc.f ·· 8 13325ns : Z CV 000001 B457E07A01 sub.f rlc.f mov wb : rO <- OOOCOOOO 8 13375ns : Z CV 000001 B87FFFFFFF nop sub.f rlcf wb : rO < · 00000000 8 13425ns : 000001BC 7FFFFFFF nop nop sub.f wb : rO <· 00000001 # 13475ns : Z 000001C0 FFFFFFFF --■ nop nop - 8 13525ns : Z 000001 CO FFFFFFFF nop ■· Empty slots Long Immediate Stage 4 -Writeback to core register Condition Code Flags Data Word Program Counter Opcodes in pipeline stages 1 · 3 Simulation Time Stage 4 Generic build & ARCAngel builds The ARCAngel build allows the designer to simulate and synthesize an ARC to the FPGA on the ARCAngel development board so that extensions and other features that have been added can be debugged. The generic build is similar however memory extensions can be added in addition to the core extensions. This build allows the designer to simulate most variations although certain combinations provide 'Dummy' blocks for the designer to add their own functionality. The generic system can be synthesized from auto-generated scripts for the Synopsys Design Compiler. To simulate, simply type 'asictest'.

To run code in the Generic or ARCAngel builds First ■ we need to have produced a Mentor-quicksim format HEX file from the Metaware tools ■ This has already been done for basecase and extensions test code, since the basecase memory models read these files. The system builds are also capable of reading hex files, however, the user must be aware that test code generated must Upon conclusion (when the ARC has been halted) the user is made aware whether the test being simulated passed or failed. This is accomplished by reading the contents of the Status register via the host port to check whether the Zero flag has been set.

Core Verification system build The Core verification system defaults to settings that allow the user to simulate, synthesize and verify the ARC core. For more information see the ARC core verification overview document. 4.2 Running Extensions test code When an installation has been completed, the simulation is set up to run the basecase test code. The VHDL reads a file 'init_mem.hex' from the working directory that contains a memory image of the test program. Compiling the extensions test source code (testxalits) with the script mw.bat created these files.

Memory image Tests x_bshift.hex 32 bit barrel shift/rotate block x_minmax.hex MIN and MAX instructions x_mul64.hex 32 x 32 scoreboard multiplier with 64-bit result x_mulmac.hex 16-bit Multiply/Multiply-accumulate function with 36-bit accumulator x_norm.hex Normalize (find-first-bit) instruction x_swap.hex SWAP instruction x_ldstram.hex Local LD/ST RAM x_timer.hex 24-bit timer counter x_scratch.hex Scratch Pad RAM. x_slide.hex Scratch Pad RAM with sliding pointers. x_all.hex All extensions together.

For example, to test an ARC configured with the barrel shifter block, do the following from the working directory: rm initjnem.hex In -s $ARCHOME/exts|src/metawarelx_bshift.hex init_mem.hex Now run the VHOL simulation. To replace the link to the basecase test code: rm initjnem.hex In -s $ARCHQMEfarclsrcfcoretest.hex initjnem.hex Single step tests (For Core build) The standard testbench runs the code provided in initjnem.hex, and then runs it again, single stepping each of the instructions. Editing the file vhdlfmemconlvhdl can disable this. Altering the signal assignment of no_step_test to supply the value '1' does this.

Tests (For GenericfARCAngel build) The standard testbench runs code provided in initjnem.hex. Modifying the variable do_pc_test to 'V in glue.vhdl can also set the PC test, provided that the hex file $ARCHOMEfexts/srcfmetawarefaa.hex is used. Note that single step instruction mode for test code can be enabled by modifying the variable no_step_test to '0' in glue.vhdl. The single stepping feature is executed only when the test code has been successfully run in normal operation mode.

To create further additions to the test the user can use the METAWARE HighC compiler. The core build can read the output from the METAWARE tool set directly, i.e. HEX format files. The ARCAngel/generic builds also read the HEX format files output by HighC.

Cache test code Test code has been provided to allow the cache logic to be tested. It must be edited by the user to set the cache size which is in use. Make a subdirectory under his working area, and copy the appropriate files into it: llama > mkdir src llama > cd src llama > cp $ARCHOME/exts/src/metaware/cache.* . llama > cp $ARCHOME/exts|srclmetaware/macros.s . · llama > chmod +w cache.s Now edit the test code file cache.s to set the CACHES1ZE variable to the appropriate value, and recompile the code. This will require a certain amount of copying files between PC and Sun systems if the Metaware tools are not available on the Sun system. llama > cache.bat The user may now either link cache.hex to rnitjnem.ftex for his specific build, i.e. core or system. The designer can edit the files in the VHDL directory to modify or add features to the extension core logic: -xalu. vhdl Extension alu functions. xdefs.vhdl Extension instruction opcodes, auxiliary registers and other constants. xrctl.vhdl Extension control logic for pipeline flow. xaugs_regs.vhdl Auxiliary registers. xcore_regs.vhdl Core registers.

There are memory extension files that can be edited. Depending on the user's selection, files that can be edited are as follow: -sram_seq.vhdl RAM sequencer. mc_sys.vhdl Memory arbitration unit. mc_arc.vhdl Ld/st memory controller. dmccvhdl Direct mapped cache controller, ijetch.vhdl Dummy instruction fetch, miu.vhdl External memory i/f unit.

The synthesis scripts and support files can also be edited: · synopsys-dcsetup Technology-specific setup usar_synopsys_dc.setup ARC-specific setup. Included by above scripts/elaborate.dc Elaboration of VHOL scriptslanalyse.de Analysis of VHDL scripts/do_all.dc Complete synthesis of design.

Hierarchy Generation If additional files are added to the hierarchy then the makefile must be modified to reflect this.

Command line parameters : NAME system_builder · builds ARC VHDL files for simulation & synthesis.

SYNOPSYS system_builder ARCgui [-parameters] DESCRIPTION The system builder REQUIRED PARAMETERS •build_type [value] core, generic, altera, core_verification •if etch addr [value] 19 - 2B •ext addr [value] 19 - 32 -tech [value] Feature size. e.g. 0.35um •ckjpeed [value] Clock speed in MHz e.g. 100 •ck_skew [value] Clock Skew +|- value in ns. •extjib [value] Name to be used for ext library, •localjam [value] Amount of Local load/st RAM •ifetch_system [value] dummy, Direct Mapped Cache ■cache_size (value! Cache size -cachejinejength [valuel Cache Line length ■add chan [value) Additional channels -wait_states [value] Ram sequencer wait states -install dir [path] Where destination dir is to be created -installjiame [dir name] Name of the destination directory OPTIONAL PARAMETERS •registerjile Use synchronous RAM cell register file, ■cache Cache enabled •ldst_cont load/store controller •hostif_cont host interface controller •host_port [value] PCdummy ■ifetch_cont l-fetch memory controller •arbitrator Memory Arbitrator •ram_seq Ram sequencer •sram_chan Ram Sequencer Memory Arbitrator channel •halt Halt the ARC on reset ■mti Compile for use with MTI Vsystem.

APPENDIX II List of Script files sampieextoisioiixrctLvhdl Merged into placeholders.

Claims (40)

142342/2

1. A computer-implemented method of producing at least a portion of an integrated circuit design, said design including at least one of a processor or processor peripheral device, the method comprising the following acts performed by a computer process: receiving one or more inputs from a user for at least one customized parameter of the processor or processor peripheral; generating a customized description language model based on the least one customized parameter, said act of generating including the act of modifying at least one prototype description by substituting or modifying values in the at least one prototype description, or merging additional descriptions based on the at least one customized parameter; and generating test code associated with the customized description language model based at least in part on the at least one customized parameter.

2. The method of Claim 1 , wherein the customized description language model includes both functional and structural description language elements for the processor or processor peripheral.

3. The method of Claim 2, wherein said acts of generating both comprise generating via an automated process.

4. The method of Claim 2, wherein said one or more inputs comprise at least one input selected from the group consisting of (i) overall device or system configuration, (ii) width of the external memory system data bus, and (iii) cache type/size.

5. The method of Claim 2, wherein said act of generating test code comprises generating configuration register test code.

6. The method of Claim 2, wherein said act of generating test code comprises generating test code based at least in part on one or more extensions added to said customized description language model as part of said act of generating said model.

7. The method of Claim 2, wherein said processor or processor peripheral comprises a RISC processor, and said description language model comprises a VHDL model.

8. The method of Claim 2, wherein said act of receiving one or more inputs from a user comprises receiving one or more inputs via a menu driven user interface. 142342/2

9. The method of Claim 1 or 6, further comprising causing a first computer program to write a second computer program adapted to perform a specific task based at least in part on said user inputs during said act of generating a customized description language model.

10. The method of Claim 9, further comprising executing said second program to perform said task.

11. 1 1. The method of Claim 2, further comprising evaluating the acceptability of said description language model using at least said test code.

12. The method of Claim 1 , further comprising testing the customized description language through simulation or synthesis using said test code; and if a different testing outcome is desired, receiving one or more additional inputs and generating a second customized description language model.

13. The method of Claim 102, wherein the method further comprises conducting a hardware/software co-verification.

14. The method of Claim 13, wherein the act of conducting a hardware/software co-verification further comprises using a debugger to run software using at least a portion of the description language model.

15. The method of Claim 1 , wherein the method further comprises conducting a core verification as part of the act of generating a customized description language model.

16. Apparatus configured to produce at least a portion of an integrated circuit design, said design including at least one of a processor or processor peripheral device, the apparatus comprising: an input module adapted to receive one or more inputs from a user for at least one customized parameter of the processor or processor peripheral; a model generator adapted to generate a customized description language model based on the least one customized parameter, said generating of said model including modifying at least one prototype description by substituting or modifying values in the at least one prototype description, or merging additional descriptions based on the at least one customized parameter; and a test code generator adapted to generate test code associated with the customized description language model based at least in part on the at least one customized parameter.

17. The apparatus of Claim 16, wherein the customized description language model includes both functional and structural description language elements for the processor or processor peripheral. 142342/2

18. The apparatus of Claim 17, wherein said model generator and test code generator each comprise a computerize module.

19. The apparatus of Claim 17, wherein said one or more inputs comprise at least one input selected from the group consisting of (i) overall device or system configuration, (ii) width of the external memory system data bus, and (iii) cache type/size.

20. The apparatus Claim 17, wherein said test code comprises configuration register test code.

21. The apparatus of Claim 17, wherein said test code is generated based at least in part on one or more extensions added to said customized description language model as part of said act of generating said model.

22. The apparatus of Claim 17, wherein said processor or processor peripheral comprises a RISC processor, and said description language model comprises a VHDL model.

23. The apparatus of Claim 17, wherein said input module comprises a menu-driven user interface program.

24. The apparatus of Claim 16, further comprising a first computer program which writes a second computer program adapted to perform a specific task based at least in part on said user inputs.

25. The apparatus of Claim 24, wherein said apparatus further automatically executes said second program to perform said task.

26. The apparatus of Claim 17, further comprising a module configured to evaluate the acceptability of said description language model using at least said test code.

27. The apparatus of Claim 16, further comprising a module adapted to test the customized description language through simulation or synthesis using said test code; wherein if a different testing outcome is desired, said apparatus is further configured to receive one or more additional inputs and cause generation of a second customized description language model.

28. The apparatus of Claim 16, further comprising a module adapted to conduct a hardware/software co-verification.

29. The apparatus of Claim 28, wherein the act of conducting a hardware/software co-verification further comprises using a debugger to run software using at least a portion of the description language model.

30. The apparatus of Claim 16, further comprising a module adapted to conduct a core verification as part of the generation of said customized description language model. 142342/2

31. The apparatus of Claim 16, further comprising a feedback module that enables a user to test the customized description language through simulation or synthesis using the test code generated; and wherein if a different testing outcome is desired, the input module enables the user to input an identification of one or more input for at least one customized parameter and the model generator generates a second customized description language model.

32. The apparatus of Claim 16, wherein said apparatus comprises a microcomputer system having a display device, microprocessor, input device, and storage device.

33. An integrated circuit fabricated according to the method comprising: creating a customized description language model of an integrated circuit design by: receiving one or more inputs from a user for at least one customized parameter of the integrated circuit; receiving an identification of a location of one or more library files that provide at least one prototype description and at least one extension logic description for the integrated circuit for which a model is being generated; and generating through an automated process a customized description language model based on at least one customized parameter, the at least one prototype description, and the at least one extension logic description.

34. The integrated circuit of Claim 33, wherein said automated process is further adapted to read the at least one prototype description and modifying the at least one prototype description by (i) substituting or modifying values in the at least one prototype description, or (ii) merging additional descriptions based on the at least one customized parameter.

35. The integrated circuit of Claim 34, wherein said method further comprises: generating a netlist which is descriptive of the circuitry of said integrated circuit; compiling said netlist and said hardware description model to produce a compiled integrated circuit design; fabricating at least one mask or configuration file representing said compiled integrated circuit design; and fabricating said integrated circuit using said at least one mask or configuration file.

36. The integrated circuit of Claim 34, wherein said method is performed at a high level of design abstraction. wherein said act of creating is performed at a high level of abstraction. 142342/2

37. The integrated circuit of Claim 34, wherein the act of receiving one or more inputs comprises receiving input of one or more parameters by a user selecting at least one of a plurality of input parameters associated with said design, said at least one parameter being selected from the group consisting of: (i) custom instruction sets; (ii) cache configurations; (iii) memory interface configurations; and (iv) system architecture configurations.

38. The integrated circuit of Claim 35, wherein the act of generating a netlist comprises generating a list of logic devices and their interconnections.

39. The integrated circuit of Claim 35, wherein the act of fabricating said integrated circuit comprises defining physical features on a semi- conductive substrate via a lithographic process.

40. A method of generating an integrated circuit design at a high level of abstraction, comprising: providing a user with a plurality of optional inputs, including the ability to generate a customized hardware description language code instruction; selecting at least one of said plurality of optional inputs; selecting at least one cache configuration; defining at least one memory interface; generating through an automated process a customized description language model based on at least one optional input, cache configuration, and memory interface customized parameter, the automated process including the acts of reading at least one prototype description, modifying the at least one prototype description by substituting values in the at least one prototype description or merging additional descriptions based on the at least one optional input, cache configuration and memory interface, and incorporating any customized hardware description language code instructions. For the Applicants AND PARTNERS