Language Compiler

The invention relates to a method for compiling computer code.

A computer can understand and interpret only machine language which is in binary form and thus very difficult to write. Assembler language is a symbolic programming language that allows the programmer to code instructions instead of coding directly in machine language. Because the assembler language allows the programmer to use meaningful symbols made up of alphabetic and numeric characters instead of merely the binary digits 0 and 1 used in machine language, it can make coding easier to read, understand and change.

FIG. 1 shows an overview of a computer system. A programmer enters code in assembler language at a terminal 20 which is the stored in a source file 30 in memory. When the source file 30 is completed it is processed by an assembler 40 which produces an object module 50 in machine language. The object code in the object module 50 is used as an input to another processing program, the linkage editor 60, which in turn produces a load module 70. The load module 70 can be loaded into main storage 80 of a computer, which then executes the program.

The assembler language is the symbolic programming language that lies closest to the machine language in form and content. It is made up of statements that represent instructions and/or comments. The instruction statements are the working part of the language and are divided into the following three groups:

1) Machine Instructions. These are symbolic representations of machine languages instructions of the processor instruction set.

2) Assembler Instructions. These are requests to the assembler program to perform certain operations during the assembly of a source module. Examples of such operations are defining data constants, defining the end of a source module or reserving main storage areas. Except for instructions that define constants, the assembler does not translate assembler instructions into object code.

3) Macro Instructions. These are requests to the assembler program to process a predefined sequence of code called a macro definition in a "pre-compile" step. The macro definition may contain variables and occurs once at the beginning of the source code. The macro must be invoked by a "macro call" which may set the value of macro variables following the definition. From the macro definition, the assembler generates machine and assembler instructions which it then processes as if they were part of the original input in the source module. Macro definitions can be programmed by the user or may be pre-programmed into the assembler.

A description of the IBM/370 Assembler Language is found in the IBM Manual GC33-4010, entitled "OS/VS-DOS/VSE-VM/370 Assembler Language" and in the book "Principles of Assembler Language Programming for the IBM 370" by Spotswood D. Stoddard, McGraw Hill, Inc., New York, 1985. The description of the/370 Assembler language from these publications is incorporated herein by reference.

When compiling computer code, the assembler uses a base register and usually one or more "work" registers. The base register in IBM/370 or IBM/390 architecture can address a range of 4096 bytes. Whilst this is generally sufficient for compilation purposes, it can at times become overloaded. On example of such overloading is when code allowing the diagnosis of software errors (called trace points) needs to be incorporated into the main program code. Incorporating the trace point descriptors occupies valuable computer program main storage in the code and in some cases it would not be possible to introduce the descriptors since the base register would not have room for the extra storage required.

One way of solving this problem is by changing the sequence of language statements produced by a language compiler compared to the actual sequence of code generation.

Changing the sequence of language statements can be done for example by causing compiled language code to appear earlier in a program. Such a technique would be used to overlay already existing code data with new data and uses the ORG Assembler Language Control statement. This statement can also be used when, for example, at the time that a main storage location is generated in a program, the actual data to be stored there is not known until some later time when the compiler program generation has proceeded to some new location--one can then cause the compiler to overlay the past location with the final value which is now known. An example of such a program is:

In this example, the first statement reserves space for a variable LABEL1. Only later in the program is the value of this variable known. Using the ORG and DC statements, one can overlay the reserved space with the value 5. This value is then used during execution for any statements which use the value of variable LABEL1.

The ORG statement also has the advantage that one main storage location can be referenced with a given label (label "a") at the time of language compiling, and later another label (label "b") can be generated and caused to overlay the same location as label "a", thereby causing the same main storage location to be able to be referenced by two different labels ("a" and "b").

Another example of changing the sequence of language statements is when one wishes to cause language code to appear later in a program. This technique uses the LTORG statement to enable the programmer to create and refer to needed data constants in the program that do not yet exist. The statement is used to generate data constants (referred to as "literals", see Stottard p.30) that one would have to generate at some accessible location to the program and create some label for each on one and collect the data constants (called the "literal pool") at some place in main storage (see Stottard p.43). As an example, without the LTORG statement one might program:

But using the LTORG statement one might program:

The LTORG statement when compiled will effectively create the following statements without any effort by the programmer:

There are several disadvantages with using the LTORG statement. Firstly, the final order of the data within the program can be different from the order intended by the programmer, since the compiler may try to optimize the computer main storage required (e.g. the compiler will juggle the data so that they fit nicely aligned into 4 or 8 byte blocks without wasting space). For example, one codes:

but the LTORG statement when compiled will effectively create the following statements:

From this example, it can be seen that the order of the data has been altered from that order in which the data appear within the program. This is a serious limitation to the generalized usage of the LTORG statement.

The second disadvantage to the LTORG statement is that the data constants can not be manipulated by the program, i.e. they are "read-only" (see Stottard, p.114). Finally the third disadvantage to the LTORG statement is that only special parts of the language can be used as a "literal", the assumption being that these parts are to be used to generate data definitions, e.g. one cannot move statements forward that contain the ORG statement.

In principle, the ORG statement might also be used to cause language code to appear forward in a program. However in practice this does not occur for several reasons. Firstly, it is difficult generally to implement, so its use is discouraged and secondly there are theoretical limitations to the implementation as shown by the following example:

Here the Language Compiler is unable to figure out what to do because it can not calculate a location of LABELB without knowing the length of LABELC whose length depends in turn on the location of LABELB.

An object of the present invention is to provide a method for compiling computer code which overcomes the above disadvantages.

This object is solved by providing a macro definition which defines archive storage, pre-compiles and archives a first portion of the computer code into a code deposit in the archive storage, compiles in a language compiler a second portion of the computer code to form compiled computer code in the base register range, retrieves from said code deposit said pre-compiled first portion of said computer code and inserts in said compiled computer code said pre-compiled first portion at the required location.

The deposit is created by defining an assembler storage array of sufficient size to hold said portion of the pre-compiled code. In one embodiment of the invention the said step of placing a portion of said pre-compiled code into said deposit is carried out by calling a macro.

In another embodiment of the invention, the deposit is created by inserting statements of said pre-compiled code into a scaffolding comprising non unique statements of computer code and placing the unique statements of said computer code and said scaffolding into the code deposit.

In the preferred embodiment of the invention, the computer code placed into the code deposit comprises trace point descriptors. These trace point descriptors are, in the preferred embodiment, between 62 and 4096 bytes in length. In addition to the pre-compiled computer code placed into the deposit, a fixed length of computer code, comprising 46 bytes, is maintained in the base register range. This means that the computer code compiled in the base register range can have a maximum number of trace points equal to 4096/46=89. Without the invention the base register could only compile code with between 4096/(46+62)=37 and 4096/(46+4096)=0 trace points.

FIG. 1 shows an overview of a computer system with assembler.

FIG. 2 shows an overview of the assembler according to the current invention.

FIG. 3 shows a block diagram outlining the steps of the invention.

An overview of the assembler according to the current invention is shown in FIG. 2. The source code in assembler language is shown as source module 200. This code incorporates two type of macro instructions which make use of the invention and are shown in the source module 200. The macro instructions are defined by the macro definitions MACRO.sub.-- DEFA and MACRO.sub.-- DEFB. They are called by the macro instructions MACRO.sub.-- INST1 or MACRO.sub.-- INST2 and MACRO.sub.-- LAST respectively. These instructions are merely illustrative and other instructions could equally be well used. Block 210 represents the language compiler which compiles the source code from a source module 200 and outputs it as object code into an object module 220.

Block 230 represents the macro facility supplied by the assembler language compiler. Full details of the macro facility are supplied in part IV, pages 237-376, of the IBM Book GC33-4010-5 as referenced above. The macro facility 230 allows the programmer to write a macro instruction in the source module 200 which then tells the assembler 40 to process a particular macro definition. The instructions MACRO.sub.-- INST1, MACRO.sub.-- INST2 and MACRO.sub.-- LAST are three examples of macro instructions which use a macro definition. Macro instructions MACRO.sub.-- INST1 and MACRO.sub.-- INST2 use the macro definition MACRO.sub.-- DEFA. Macro instruction MACRO.sub.-- LAST uses the definition MACRO.sub.-- DEFB. The assembler 40 produces assembler language statements from these macro definitions every time the macro instructions call the macros defined by the macro definitions. Within the block 230 representing the macro facility, two sub blocks 235 and 237 are shown which represent macro definitions MACRO.sub.-- DEFA and MACRO.sub.-- DEFB respectively.

Block 240 represents archive storage which can be used to temporarily store parts of code and variables by the language compiler 210. Within the block 240 representing the working storage is shown a sub block 245 which represents an assembler storage array. The macro facility 230 allows the programmer to define for use during compilation of a source module 200 "storage array" tables for text statements. The tables vary in size depending on the capabilities of the language compiler. In the implementation of the invention, the table may contain between 4096 and 32768 entries. Each entry is able to hold text statements of between 1 and 256 characters.

The method in which the invention is implemented is shown in FIG. 3. In block 310, the programmer firstly defines at the beginning of his program in the working storage 240 an assembler storage array 245 of adequate size to hold the code "deposit" or "deposits". The maximum size of the deposit is determined by the compiler as mentioned earlier. For example it could be 1000 entries. In the source module 200 the following statement will appear:

GBLC is a macro command in/370 assembler language which is contained in both macro definitions MACRO.sub.-- DEFA and MACRO.sub.-- DEFB in the macro compiler 230. It is called by the first macro instruction MACRO.sub.-- INST1 defined by MACRO.sub.-- DEFA and is referred to by subsequent calls of the macro instructions defined by MACRO.sub.-- DEFA (i.e. MACRO.sub.-- INST2 in our example). It is also called by the macro instruction MACRO.sub.-- LAST. GLBC creates a character array 245 in the archive storage 240. In this example a character array 245 named &DEPOS is created with a maximum of 1000 entries.

At the point within the source module 200 where the code "deposit" is to be created, the programmer inserts a macro instruction call which pre-compiles and archives each code deposit statement to an entry within the assembler storage array 245 (block 320) by writing the following series of statements:

Each of the instructions SETC is a macro instruction whose variable content is `code statement x`. The series of SETC instructions is represented by the macro instruction MACRO.sub.-- INST1 in the source module 200 in FIG. 2. The macro instruction MACRO.sub.-- INST1 calls the macro defined by the macro definition MAC.sub.-- DEFA to pre-compile and archive the various code statements, represented by `CODE STATEMENT 1` etc, as entries &DEPOS(1) etc., into the created assembler storage array 245.

In the example illustrated a second code deposit is created by the macro instruction MACRO.sub.-- INST2. This also comprises a series of statements.

The second code deposit is stored in the created assembler storage array 245 at locations immediately following the location occupied by the first code deposit.

Later in the program when the programmer decides that he wishes to pass the code deposit to the language compiler (block 330), another macro is called by the macro instruction MACRO.sub.-- LAST in the source module 200. The called macro is defined by the second macro definition 237 MACRO.sub.-- DEFB in the macro facility 230. The macro instruction MACRO.sub.-- LAST causes the contents of the entries in the assembler storage array 245 to be passed to the language compiler 210:

The code statements in these entries are then compiled and placed in the object module 220 (block 340) at the required places in the object code. The code statements placed in the code deposit can of course refer to variables and labels which were first defined before, during or after the code had been placed into the deposit 245.

It is also possible to use the macro facility 230 to generate code in other ways. Suppose, for example, that the object module 220 contains sections of code generated from source code which had the same basic structure but in which only part of the statement was unique. An example of such a structure would be the following series of code statements:

This series of code statements appears repeatedly throughout the program with different labels and different variables. The assembler instruction DC remains in the same place throughout the repeated series of code statements (i.e. is "non-unique"). The macro definition MACRO.sub.-- DEFA 235 is then defined to contain a "scaffolding" which comprises the non-unique portions of the code statements:

The macro definition MACRO.sub.-- DEFA 235 also comprises a series of SETC instructions to indicate that the variables &VAR0, &VAR1 etc. given when the macro instruction is called are placed in the corresponding locations in the assembler storage array 245:

When the macro is called, e.g. by the macro instruction MACRO.sub.-- INST1, the values of the variable &VARO etc. are given:

The effect of calling the macro defined by MACRO.sub.-- DEFA is therefore to store in the assembler storage array the following statement:

The macro defined by the second macro definition 237 MACRO.sub.-- DEFB is then called later by the source module 200 using the macro instruction MACRO.sub.-- LAST. The called macro uses the statements in the assembler storage array 245 together with the scaffolding given by MACRO.sub.-- DEFA to produce following source code which is then passed to the language compiler 210:

In one particular embodiment of the invention, the code placed into the deposit 245 is code used for tracing routines, known as trace point descriptions. These trace point descriptions do not contribute to the actual operation carried out by the computer program, but are used by some diagnostic programs to trace errors in the computer program when it is running. By placing the tracing code into deposits 245, more space is found in the base register range for the computer program source code which is to be compiled.

An example will serve to illustrate the advantages of doing this. The implementation of the invention split the trace point description into two parts, one of which was placed within the base register range, and the other of which was placed into the code deposit stored beyond the base register range and referenced by use of a work register. Calling the macro defined by the macro definition MACRO.sub.-- DEFA will produce two parts for the trace point description. The first part of the trace point description is stored in the base register range and is fixed in size to 46 bytes. The second part of the trace point description is variable in size, in the current implementation from 62 to 462 bytes, in future implementations up to 4096 bytes.

Thus within one base register, there can be a maximum of 4096/46=89 trace points, as opposed to 4096/(46+462)=8 trace points. In future implementations of the tracing routines, it would not be possible to have incorporated all the code relating to one trace point description within one base register.

Based on the foregoing, a system and method for compiling computer code have been disclosed. However, numerous modifications and substitutions can be made without deviating from the scope of the present invention. Therefore, the present invention has been disclosed by way of illustration and not limitation and reference should be made to the following claims to determine the scope of the present invention.