If you wish to focus on program development outside the GT.M environment, skip the next section and continue with the section "Compiling from the Shell".
GT.M compiles M source code files and produces object files for complete integration into the UNIX enviroment. The object modules have the same name as the compiled M source file with an .o file extension, unless otherwise specified. The object files contain machine instructions and information necessary to connect the routine with other routines, and map it into memory. An M routine source file must be compiled after it is created or modified. You can compile explicitly with the ZLINK command or implicitly with auto-ZLINK. At the shell command line, compile by issuing the mumps command.
The compiler checks M code for syntax errors and displays error messages on the terminal, when processing is complete. Each error message provides the source line in error with an indicator pointing to the place on the line where the error is occurring. For a list and description of the compiler error messages, refer to the GT.M Message and Recovery Procedures Reference Manual.
You can generate a listing file containing the compile results by including the -list qualifier as a modifier to the argument to the ZLINK command in Direct Mode. This can also be done by redirecting the compiler messages to a file by adding >filename 2>&1 to the mumps command when compiling a program from the shell. See “Compiling from the Shell” for an explanation of the M command describing -list, and other valid qualifiers for the M and ZLINK commands.
The compiler stops processing a routine line when it detects an error on that line. Under most conditions the compiler continues processing the remaining routine lines. This allows the compiler to produce a more complete error analysis of the routine and to generate code that may have valid executable paths. The compiler does not report multiple syntax errors on the same line. When it detects more than 127 syntax errors in a source file, the compiler ceases to process the file.
In Direct Mode, GT.M provides access to the compiler explicitly through the ZLINK and ZCOMPILE commands, and implicitly through automatic invocation of ZLINK functionality (auto-ZLINK) to add required routines to the image. ZCOMPILE is a GT.M routine compilation command, it compiles the routine and creates a new object module. The primary task of ZLINK is to place the object code in memory and "connect" it with other routines. However, under certain circumstances, ZLINK may first use the GT.M compiler to create a new object module.
The difference between ZCOMPILE and ZLINK is that ZCOMPILE creates a new object module on compiling, whereas the ZLINK command links the object module with other routines and places the object code in memory.
ZLINK compiles under these circumstances:
ZLINK cannot locate a copy of the object module but can locate a copy of the source module.
ZLINK can locate both object and source module, and finds the object module to be older than the source module.
The file-specification portion of the ZLINK argument includes an explicit extension of .m.
Auto-ZLINK compiles under the first two circumstances, but can never encounter the last one.
When a command refers to an M routine that is not part of the current image, GT.M automatically attempts to ZLINK and, if necessary, compile that routine. In Direct Mode, the most common method to invoke the compiler through an auto-ZLINK is to enter DO ^routinename at the GTM> prompt. When the current image does not contain the routine, GT.M does the following:
Locates the source and object
Determines whether the source has been edited since it was last compiled
Compiles the routine, if appropriate
Adds the object to the image
By using the DO command, you implicitly instruct GT.M to compile, link, and execute the program. With this method, you can test your routine interactively.
For complete descriptions of ZLINK and auto-ZLINK, see Chapter 6: “Commands” .
Example:
GTM>do ^payroll GTM>do ^taxes
This uses the M DO command to invoke the GT.M compiler implicitly from the GTM> prompt if the routine requires new object code. When the compiler runs, it produces two object module files, payroll.o and taxes.o.
If you receive error messages from the compilation, you may fix them immediately by returning to the editor and correcting the source. By default, the GT.M compiler operates in "compile-as-written" mode, and produces object code even when a routine contains syntax errors. This code includes all lines that are correct and all commands on a line with an error, up to the error. Therefore, you may decide to tailor the debugging cycle by running the program without removing the syntax errors.
Caution | |
---|---|
The DO command does not add an edited routine to the current image if the image already includes a routine matching the DO argument routine name. When the image contains a routine, GT.M simply executes the routine without examining whether a more recent version of the module exists. If you have a routine in your image, and you wish to change it, you must explicitly ZLINK that routine. |
Example:
GTM>zlink "payroll" GTM>zlink "taxes.m"
The first ZLINK compiles payroll.m if it cannot locate payroll, or if it finds that payroll.m has a more recent date/time stamp than payroll.o. The second ZLINK always compiles taxes.m producing a new taxes.o.
For more information on debugging in GT.M Direct Mode, see Chapter 4: “Operating and Debugging in Direct Mode”.
From the shell, invoke the compiler by entering mumps file-name at the shell prompt.
Example:
$ mumps payroll.m $ mumps taxes.m
This uses the mumps command to invoke the GT.M compiler from the shell prompt, and creates .o versions of these files.
Use the mumps command at the shell prompt to:
Check the syntax of a newly entered program.
Optionally, get a formatted listing of the program.
Ensure that all object code is up to date before linking.
The mumps command invokes the compiler to translate an M source file into object code.
The format for the MUMPS command is:
MUMPS [-qualifier[...]] pathname
Source programs typically have an extension of .m. GT.M accepts other file extenions for explicit compilation commands, but not for auto-ZLINK.
Each pathname identifies an M source program to compile. If you do not specify a .m file extension, the mumps command assumes the .m file extension.
Qualifiers determine characteristics of the compiler output.
Qualifiers must appear after the command, but before the file name to be properly applied.
GT.M allows the UNIX * and ? wildcards in a file name.
The MUMPS command returns a status of 1 after any error in compilation.
The * wildcard accepts any legal combination of numbers and characters including a null, in the position the wildcard holds.
The ? wildcard accepts exactly one legal character in its position.
For example, mumps *.m compiles all files in the current default directory with an .m extension. mumps *pay?.m compiles .m files with names that contain any characters followed by pay, followed by one character. Unlike when using ZLINK or ZCOMPILE, the filename must be fully specified when compiling from the shell.
Caution | |
---|---|
When forming routine names, the compiler truncates object filenames to a maximum length of 31 characters. For example, for a source file called Adatabaseenginewithscalabilityproven.m the compiler generates an object file called Adatabaseenginewithscalabilityp.o. Ensure that the first 31 characters of source file names are unique. |
The mumps command allows qualifiers that customize the type and form of the compiler output to meet specific programming needs. MUMPS command qualifiers may also appear as a modifier to the argument to the GT.M ZLINK and ZCOMPILE commands. The following section describes the mumps command qualifiers. Make sure the arguments are specified ahead of file name and after the command itself.
Invokes a small GT.M image that immediately initiates Direct Mode.
-direct_mode does not invoke the M compiler.
The -direct_mode qualifier is incompatible with a file specification and with all other qualifiers.
Compiles certain data structures associated with literals used in the source code in a way that they are dynamically loaded and unloaded from the object code. The dynamic loading and unloading of these data structures:
Supersedes any specification of -NOINLINE_LITERALS.
Reduces the amount of private memory required by each process which in turn allows more processes to execute with the same memory.
In some circumstances, increases application performance by making more memory available for file system buffers.
Increases the CPU and stack costs of local variable processing.
With no -DYNAMIC_LITERALS specified, these data structures continue to be generated when a routine is linked and stay in the private memory of each process. As the use of -DYNAMIC_LITERALS increases object code size, and as the dynamic loading and unloading only saves memory when the object code is in shared libraries, FIS recommends restricting the use of -DYNAMIC_LITERALS to only when compiling object code to be loaded into shared libraries or executed from an auto relink enabled directory.
Instructs GT.M to embeds routine source code in the object code. The default is NOEMBED_SOURCE. Like other GT.M compilation qualifiers, this qualifier can be specified through the $ZCOMPILE intrinsic special variable and gtmcompile environment variable. EMBED_SOURCE provides $TEXT and ZPRINT access to the correct source code, even if the original M source file has been edited or removed. Where the source code is not embedded in the object code, GT.M attempts to locate the source code file. If it cannot find source code matching the object code, $TEXT() returns a null string. ZPRINT prints whatever source code found and also prints a TXTSRCMAT message in direct mode; if it cannot find a source file, ZPRINT issues a FILENOTFND error.
Instructs the compiler to produce an object file even when the compiler detects errors in the source code (-ignore), or not to produce an object file when the compiler encounters an error (-noignore).
When used with the -noobject qualifier, the -ignore qualifier has no effect.
Execution of a routine that compiles with errors produces run-time errors when the execution path encounters any of the compile time errors.
This compile-as-written mode facilitates a flexible approach to debugging and expedites conversion to GT.M from an interpreted environment. Many M applications from an interpreted environment contain syntax abnormalities. This feature of compiling and later executing a routine provides the feel of developing in an interpreted environment.
By default, the compiler operates in -ignore mode and produces an object module even when the source routine contains errors.
Controls the page length of the listing file.
The M compiler ignores the -length qualifier unless it appears with the -list qualifier.
By default, the listing has -length=66.
Instructs the compiler to produce a source program listing file, and optionally specifies a name for the listing file. The listing file contains numbered source program text lines. When the routine has errors, the listing file also includes an error count, information about the location, and the cause of the errors.
When you do not specify a file name for the listing file, the compiler produces a listing file with the same name as the source file with a .lis file extension.
The -length and -space qualifiers modify the format and content of the listing file. The M compiler ignores these qualifiers unless the command includes the -list qualifier.
By default, the compiler operates -nolist and does not produce listings.
Compiles routines to use library code in order to load literals instead of generating in-line code thereby reducing the routine size. At the cost of a small increase in CPU, the use of -NOINLINE_LITERAL may help counteract growth in object size due to -DYNAMIC_LITERALS.
Important | |
---|---|
Both -DYNAMIC_LITERALS and -NOINLINE_LITERNALS help optimize performance and virtual memory usage for applications whose source code includes literals. As the scalability and performance from reduced per-process memory usage may or may not compensate for the incremental cost of dynamically loading and unloading the data structures, and as the performance of routines vs. inline code can be affected by the availability of routines in cache, FIS suggests benchmarking to determine the combination of qualifiers best suited to each workload. Note that applications can freely mix routines compiled with different combinations of qualifiers. |
Instructs the compiler to produce an output object file and optionally specifies a name for the object file using the optional filename argument.
When you do not specify a file name, the compiler produces an object file with the same file name as the source file and an .o file extension.
When forming routine names, the compiler truncates object filenames to a maximum length of 31 characters. For example, for a source file called Adatabaseenginewithscalabilityproven.m the compiler generates an object file called Adatabaseenginewithscalabilityp.o. Ensure that first 31 characters of source file names are unique.
The -noobject qualifier suppresses the production of an object file and is usually used with the -list qualifier to produce only a listing file.
Compilation with -OBJECT without -NAMEOFRTN implicitly names the first routine to match the name specified with -OBJECT.
By default, the compiler produces object modules.
Instructs the compiler to produce an output object file with the specified name. You can use -NAMEOFRTN and -OBJECT to create two object files with different names from the same .m source file.
Instructs the compiler to suppress error output; the default is -warning.
When used with the -list qualifier, the -nowarning qualifier does not affect errors in the listing.
Note | |
---|---|
When used with the -noobject qualifier, the -nowarning qualifier instructs the compiler to produce no object with no indication of the fact or the cause of any errors. |
Invokes GT.M in Autostart Mode.
The next argument is taken to be an M entryref. That routine is immediately executed, bypassing Direct Mode. Depending on the shell, you may need to put the entryref in quotation marks (""). This qualifier does not invoke the M compiler and is not compatible with any other qualifier.
Controls the spacing of the output in the listing file. -space=n specifies n-1 blank lines separating every source line in the listing file. If n<1, the M command uses single spacing in the listing.
If this qualifier appears without the -list qualifier, the M compiler ignores the -space qualifier.
By default, listings use single spaced output (-space=1).