HLA Language Reference and User Manual

Overview

HLA, the High Level Assembler, is a vast improvement over traditional assembly languages. With HLA, programmers can learn assembly language faster than ever before and they can write assembly code faster than ever before. John Levine, comp.compilers moderator, makes the case for HLA when describing the PL/360 machine specific language:

 

1999/07/1119:36:51,themoderatorwrote:

"There'sno reason that assemblers have to have awful syntax. About 30 years ago I used Niklaus Wirth's PL360, which was basically a S/360 assembler with Algol syntax and a a little syntactic sugar like while loops that turned into the obvious branches. It really was an assembler, e.g., you had to write out your expressions with explicit assignments of values to registers, but it was nice. Wirth used it to write Algol W, a small fast Algol subset, which was a predecessor to Pascal. ... -John"

 

PL/360, and variants that followed like PL/M, PL/M-86, and PL/68K, were true "mid-level languages" that let you work down at the machine level while using more modern control structures (i.e., those loosely based on the PL/I language). Although many refer to "C" as a "medium-level language", C truly is high level when compared with languages like PL/*. The PL/* languages were very popular with those who needed the power of assembly language in the early days of the microcomputer revolution. While it's stretching the point to say that PL/M is "really an assembler," the basic idea is sound. There really is no reason that assemblers have to have an awful syntax.

HLA bridges the gap between very low level languages and very high level languages. Unlike the PL/* languages, HLA really is an assembly language. You can do just about anything with HLA that you can do with a traditional assembler like MASM, TASM, NASM, or Gas. If you want to write low-level assembly code using x86 machine instructions, HLA does not get in your way; if you want to use compares and conditional branches rather than structured control statements, you can. On the other hand, if you prefer to use more readable high-level control structures, HLA allows this, as well. HLA lets you work at the level you are most comfortable with and at the level that is most appropriate for the task at hand.

Beyond supplying a "non-awful" syntax, HLA has one other important feature -- it's extensible. HLA provides special features that let you add new statements to the language. So if HLA is not "high level" (or "low level") enough for your tastes, you can extend it. This document will expend considerable effort describing exactly how to do this in a later section.

In addition to the HLA language itself, the HLA system provides one other very important component - the HLA Standard Library. This is a collection of hundreds of functions that you can use to write assembly language programs as quickly and easily as you would write C programs.

What is a "High Level Assembler"?

The name "High Level Assembler" and its abbreviation "HLA" is certainly not new1. Nor is the concept of a high level assembler. David Salomon in his 1992 text "Assemblers and Loaders" (Ellis Horwood, ISBN 0-13-052564-2) uses these terms to describe various assembly languages dating back to 1966. Furthermore, both IBM and Motorola have assembler products with very similar names (e.g., IBM's HLAsm, though it's somewhat debatable whether HLAsm is truly a high level assembler).

Salomon offers the following definitions for a High Level Assembler (or HLA):

A high-level assembler language (HLA) is a programming language where each instruction is translated into a few machine instructions. The translator is somewhat more complex than an assembler, but much simpler than a compiler. Such a language should not have features like the if, for, and case control structures, complex arithmetic, logical expressions, and multi-dimensional arrays. It should consist of simple instructions, closely resembling traditional assembler instructions, and of a few simple data types.

Since Salomon describes a couple of high level assemblers that exceed this definition, he offers a second definition for high level assemblers that is a bit higher-level:

A high-level assembler language (HLA) is a language that combines most of the features of higher-level languages (easy to use control structures, variables, scope, data types, block structure) with one important feature of assembler languages namely, machine dependence.

Neither definition is particularly useful for describing HLA/86 and other HLAs like Terse, MASM and TASM. Of course the term "High Level Assembler" is very nebulous and offers a fair amount of latitude. Almost any macro assembler could pass as an HLA on the basis that a macro-instruction expands into a few machine instructions.

David Salomon describes several different high level assemblers in his text. The examples he describes are PL/360, NEAT/3, PL516, and BABBAGE.

PL/360 and PL516 are products that conform to the second definition above. They allow simple arithmetic expressions and assignment statements, the use of high level control structures (if, for, while, etc.), high level data declarations, and block structure (among other things). These languages expose the underlying machine's registers and allow the use of machine instructions using a "functional" syntax.

The NEAT/3 language is a much lower-level language; basically it is an assembly language for the NCR Century computers that provide COBOL-style data declarations. Most of its "instructions" translate one-for-one into Century machine instructions, though it does automatically insert code to convert data types from one format two another if the data types of an instruction's operands are incompatible.

The BABBAGE assembly language is an expression-based assembly language (very similar to Terse). It allows simplified high level control structures like if and while. The interesting thing about this assembler is that it was the only assembler for the GEC4000 family of computers.

In addition to the HLAs that Salomon describes, there have been several other high level assemblers created over the years. PL/M and PL/M-86 was designed by Intel for their 8080 and 8086 CPU families. This was an obvious adaptation of the PL/360 style HLA for Intel's CPUs. PL/68 was also available for the Motorola 680x0 family. SL/65 was a similar adaptation of PL/360 for the 6502 family. At one point there was a product named "High Level Assembler" for the Atari ST system (68K based). Jim Neil has also created an expression-based high level assembler (similar in principle to Babbage) for Intel's x86 family. MASM and TASM (for the x86) also fall into the category of a high level assembler due to their inclusion of high level control structures and logical expressions.

So where does HLA/86 fit into these definitions? In truth, the definition of HLA/86 falls somewhere between these two definitions. So the following paragraphs will define the term "High Level Assembler" as it should apply to HLA/86 and similar high level assemblers.

The first definition above is overly restrictive. It implies that any language that exceeds these limits is a high level language, not a high level assembly or traditional assembly language. Obviously, this definition is too restrictive in the sense that by this definition many traditional assemblers would have to be considered as high level languages (even beyond a high level assembler). Furthermore, it elevates many traditional assemblers to the status of an HLA even though we wouldn't normally think of them as high level assemblers; i.e., most macro assemblers provide the ability to create instructions that translate into a few machine instructions. Macro facilities, however, are something we expect out of a modern assembly language; their presence doesn't make the language a "high level" assembly language in most people's mind. Furthermore, most modern assemblers provide a mechanism for declaring multi-dimensional arrays (even though you still have to use some sequence of instructions to index into said arrays).

The second definition David Salomon provides hits the other extreme. Arguably, languages like C could be called HLAs under this definition (yes, there are some machine dependent features in C, though probably not enough to satisfy David Salomon's original intent).

The definition of high level assemblers like Terse, MASM, TASM, and HLA/86 fall somewhere between these extremes. Therefore, this document will define a high level assembler as follows:

A "high level assembly language" (HLAL) is a language that provides a set of statements or instructions that practically map one-to-one to machine instructions of the underlying architecture. The HLAL exposes the underlying machine architecture including access to machine registers, flags, memory, I/O, and addressing modes. Any operation that is possible with a traditional assembler should be possible within the HLAL. In addition to providing access to the underlying architecture, the HLAL must provide some abstractions that are not normally found in traditional assemblers and that are typically found in traditional high level languages; this could include structured control statements (e.g., if, for, and while), high level data types and data structuring facilities, extensive compile-time language facilities, run-time expression evaluation, and standard library support. A "High Level Assembler" is a translator that converts a high level assembly language to machine code.

 

There is a very important difference between this definition and the ones that David Salomon provides. Specifically, a high-level assembly language must provide access to the underlying machine architecture. Within the HLAL you must be able to specify any (reasonable) machine instruction that is available on the CPU. The HLAL may provide other statements that do not directly map to machine instructions (e.g., an if statement), but it must, at least, provide a set of statements that practically map one-to-one with the machine instructions. The "practically" modifier appears here for two reasons. First of all, some assembly source statements may map to two or more different, but equivalent, machine instructions. A good example is the x86 "mov reg, reg" which can map to two different (though equivalent) opcodes depending on the setting of the direction bit in the opcode. Most assemblers will map the source statement to only one of these opcodes, hence there is not truly a one-to-one mapping (since there exist some opcodes that do not map back to some source instruction). Another allowable restriction is that the HLAL may not allow the use of special "protected mode instructions" if the language is intended only for user-mode programming (as is the case for HLA/86).

In addition to supporting the underlying machine architecture (which almost any traditional assembler will do), the HLAL must also provide support for some features normally found in a high level language. The definition does not require that a HLAL support all the features listed above, nor is it restricted to just the features listed, but a HLAL must support some of the features traditionally found in a high level language. The number and type of features the HLAL supports determines how "high level" the assembly language is. Like HLLs, we can have "low-level" HLALs, "medium-level" HLALs, "high-level" HLALs, and even "very high-level" HLALs. NEAT/3, for example, would be a low-level HLAL since it provides higher-level data types, conversions, and not much else.

MASM and TASM are probably best considered medium-to-high-level HLALs since they provide high level data structuring facilities, structured control statements, high level procedure definitions and invocations, a limited block structure, powerful compile-time language (macro) facilities, standard library support (e.g., the UCR Standard Library and many other available library modules), and other high level language features. In actual use, the programmer is expected to normally use standard machine instructions and rise up to the high level statements only as necessary.

The Terse language is a good example of a medium level HLAL since it uses an expression syntax but otherwise maps statements fairly closely to the assembly counterparts. It does provide some higher-level data structuring capabilities, though this is inherited from the underlying assembler(s) on which Terse is based.

PL/360 and PL516 are definitely high-level HLALs because they fully support simplified arithmetic expressions, control structures, high-level data types, and other features. These languages provide access to the underlying architecture, but the emphasis is to use these langauges as a high level language and drop down to the machine instructions only as necessary.

HLA/86 probably falls in the high-level-to-very-high-level range because it provides high level data types and data structuring abilities, high level and very high level control structures, extensive parameter passing facilities (more than most high level languages), a very extensive compile time language, a very extensive standard library, built-in parsing facilities for language extension, and many other features. As a general rule, HLA/86 has a larger feature set than the other HLALs described above, but there are a couple of design goals that limit the "high-levelness" of HLA/86: (1) with one exception, HLA never emits any code behind the programmer's back that modifies registers or flags (the one exception is object method invocation, and this is well documented), and (2) HLA doesn't support arithmetic expressions (it does support a limited form of logical/boolean expressions). One interesting aspect of HLA/86 is that it is extensible. Using features built into the language, you can extend HLA/86's syntax by adding new statements and other features. This feature gives you the ability to make HLA/86 as high level as you desire (though it may take some effort to achieve certain language features). The bottom line is this: in some ways, HLA/86 is lower level than languages like PL/360 and PL516; in other ways, it's higher level than these HLALs. However, as the definition requires, almost anything you can do with a traditional assembler is possible in HLA/86.

What is an "Assembler"

Because high level assemblers are clearly different that traditional assemblers, one might question whether a high level assembly language is truly an assembly language and whether translators for high level assembly languages can be properly called an assembler. Unfortunately, there is a consierable range of opinions as to exactly what consitutes an "assembler" versus other translators. This document will not attempt to get involved in this debate. Instead, this section provides a set of definitions that are useful for describing assemblers at various levels of abstraction.

Pure Assembler:

A "pure assembler" is a program that processes an assembly langauge source file and translates the source code using a direct mapping from source code instructions to individual machine instructions (each source instruction is mapped to exactly one machine instruction). The assembler only provides machine-primitive data types like bytes, words, double words, etc. A pure assembler does not provide macro facilities. A pure assembler always produces machine code as output.

Traditional Assembler:

A "traditional assembler" is a pure assembler plus macro facilities. The assembler may provides some "built-in macros" and instruction synonyms, but in general, the built-in statements should still map to individual machine instructions (note that the programmer may extend this by writing macros). There is no support by the assembler for run-time arithmetic or boolean expressions. A traditional assembler may also provide some simple data typing facilities (such as the ability to rename primitive data types as something else, e.g., byte->char). A traditional assembler always emits machine code as output.

High Level Assembler:

A high-level assembler is a macro assembler plus some additional high-level language-like facilities, such as high-level control constructs or high-level-like procedure calls. If a programmer elects to ignore these additional facilities, they still have all the capabilities of a macro assembler at their disposal.

Is HLA a True Assembly Language?

Some people are confused by HLA. On the one hand it looks like a High Level Language, employing syntax similar to Pascal and C/C++. On the other hand, it does support the machine instructions found in a typical assembly language. Many people accuse HLA of being a compiler rather than an assembler. What's the truth?

The truth is, assembly languages have evolved, just as high-level languages have evolved, and we can no longer use a definition for an assembler that made sense in the 1950s when describing modern assemblers such as MASM, TASM, and HLA. Today, the best definition we can use is that an assembler is a compiler for an assembly language. An assembler accepts a source file written in some sort of assembly language and produces an object file as its output.

The real question, then is not whether HLA is an assembler, but whether the HLA language is an assembly language. Some people argue that any compiler that includes any sort of statement that compiles into more than one machine instruction cannot be called an "assembler." However, such an argument immediately eliminates macro assemblers. Eliminating macro assemblers is unsatisfactory because almost every modern assembler provides, at the very least, some simple macro facilities. Whether you implement an "IF" statement with a macro (generally supplied by the assembler's author, as is the case, for example, with FASM) that you have to include into your source file, or via a `macro' that the assembler's author has provided as part of the assembler is really a matter of implementation. To the end user of the assembler, the "IF" statement is just as much a part of the language that they can employ regardless of the implementation. The fact that assemblers such as MASM, TASM, and HLA provide these high-level-like control structures in assembly language does not imply that the languages these products implement are not assembly languages.

Some people argue that "high-level assemblers" such as MASM, TASM, and HLA are not assemblers any more than C/C++ compilers could be considered assemblers if those C/C++ compilers support an in-line assembly capability. However, their arguments strengthen the case for calling a product like HLA an "assembler." After all, if we're going to continue to call C/C++ a high-level language even though it provides support for machine instructions, then there is no reason we cannot call a product like MASM, TASM, or HLA "assemblers" even though they provide a modicum of support for high-level-like control structures. Ultimately, it is the focus of the language that defines the type of language it is. C/C++'s focus is on writing high-level language programs, with a few machine instructions thrown in now and then when the high-level language doesn't quite handle everything. High-level assemblers, such as HLA, MASM, and TASM are focused on writing assembly language modules. They have some high-level control structures thrown in to simplify some tasks (e.g., in the case of HLA, the high-level control structures exists as a bridge between HLLs and assembly during the learning process), but the focus is mainly on writing assembly language code.

Some people feel that if you learn HLA (or some other high level assembler), then you're not really learning "assembly language." This is utter nonsense. If you thoroughly learn HLA, you'll know assembly language programming inside and out. Switching to a different assembler from HLA would be no different, say, than switching from Gnu's Gas assembler to MASM (or vice versa). One might bemoan the features lost in such a translation, but when going from HLA to some other assembler you're typically giving up features rather than gaining anything.

Still there is a pervasive argument that high level control structures like IF/WHILE/FOR/etc. don't belong in a true assembler. Well, HLA, MASM, and TASM users can elect to ignore these statements (as many old-time MASM programmers do; with HLA you can even disable these statements). As long as the rest of the assembler supports a language that allows one to write "pure" assembly language code, why would anyone question the validity of the title "assembly language" for the code? (Unless, of course, they have an ax to grind.) For those who are diametrically opposed to allowing any language that contains IF/WHILE/FOR/etc. statements to be called assembly lanugage, well, that's why we call these things "high level assembly languages." To note the fact that they are a little more powerful than traditional assembly languages.

The bottom line is this: if you learn HLA, you will learn assembly language programming. As long as you understand how to write the low-level code (within HLA) and don't rely exclusively on the high-level control statements in your programs, no one can truthfully question your assembly language programming knowledge.

HLA Design Goals

HLA was originally conceived as a tool to teach assembly language programming. In early 1996 I decided to do a Windows version of my electronic text "the Art of Assembly Language Programming" (AoA). After an attempt to develop a new version of the "UCR Standard Library for 80x86 Programmers" (a mainstay of AoA), I came to the conclusion that MASM just wasn't powerful enough to make learning assembly language really easy. I decided to develop an assembler with sufficient power, providing the tools for a good standard library as well as satisify some other requirements. Therefore, HLA has two important goals: provide a system that is powerful enough to develop code and macros to make learning assembly language, which simultaneously providing a system that is easy for beginners to learn.

The principle goal of HLA was to leverage student's existing programming knowledge. For example, a good Pascal programmer can get their first C/C++ program operational in a few minutes. All they've got to do is note the similarities between the two programming languages, make the appropriate syntactical changes, and they're up and running. Take that same Pascal programming and expect them to learn LISP or Prolog the same way, and you'll not meet with the same success. LISP and Prolog are completely different, they use a different "programming paradigm," so the student has to "start over from scratch" when learning these languages. Although assembly language is an imperative language (like Pascal and C/C++), there is a considerable "paradigm shift" when moving from one of these high level languages to assembly. In HLA, I wanted to create a language with high level control structures and declarations that made it possible for someone familiar with an imperative language like Pascal or C/C++ to get their first HLA program running in a matter of minutes (or, at worst, a matter of hours). Of course, to achieve this goal, I needed to add high-level data declarations and high-level control constructs to the HLA language.

The astute reader will quickly point out that high level control structures are not assembly language and letting the students use these types of statements is not really teaching them assembly language. This is quite true; since the purpose of teaching an assembly language course is to teach the students "assembly language programming" it is quite clear that HLA would fail if it only provided these high level control structures (e.g., like the PL/M language does). Fortunately, this is not the case. HLA supports all standard assembly language instructions including CMP and Jcc instructions, so you can still write "pure" assembly language programs without using those high level language control structures. However, it does take time to learn the several hundred different machine instructions. Traditionally, it's taken my students (using only MASM) about five weeks before they could really write any meaningful programs in assembly language (you have to cover things like numeric representation, basic CPU architecture, addressing modes, data types, and introduce the instruction set before any real programs can be written).

HLA lets students write meaningful programs within about a week of it's introduction (e.g., the first assignment I give in a typical quarter is to write an "addition table" program that computes the outer product [addition table] of the two vectors 0..15 and 0..15, printing the table formatted nicely). They achieve this by using statements they already know (like IF and WHILE) with the injection of just a few assembly language concepts (registers, and the MOV and ADD instructions) plus an introduction to the HLA Standard Library. Over the next several weeks, these students write more and more complex programs as they are introduced to new assembly language and HLA concepts (e.g., data representation, basic architecture, addressing modes, data types, and additional instructions). At about the sixth week, I begin "weaning" these students off the high level language statements and force them to use the low level machine instructions. It turns out that they learn how to simulate an IF statement at roughly the same point in the quarter as they did when they used only MASM, but the big difference is that they've written a lot more code up to that point proving out other concepts in machine organzation and assembly language programming. In my limited experience with classroom testing, I've found that students spend less time on the class, cover more material, and retain the knowledge better (by the time of the final exam) than they did when I only used MASM.

 

The general goal of reducing the learning curve for students is achieved several ways.

(1) As noted above, HLA allows a gradual transition from high level languages into pure assembly language. My favorite analogy here is the Nicoderm CQ smoking cessation system ("gradual steps are better."). Like the Nicoderm system, HLA lets students learn assembly language in gradual steps rather than throwing them into the water and shouting "sink or swim!"

(2) In addition to letting the students employ high level language statements in their assembly language programs, HLA contains several other familiar concepts and syntactical items that ease the transition from high level language programming to assembly language. For example, HLA uses the familiar (to C/C++ programmers) "/*" and "*/" comment delimiters (as well as the "//" comment delimiter). Statements generally end with a semicolon (just as in high level languages). Machine instructions use a functional notation rather than "mnemonic-operand" notation. Constant, type, and variable declarations should look very familiar to Pascal programmers. HLA's standard library should look comfortable to anyone who has used the C/C++ standard library.

In addition to syntactical similarities, well-written HLA programs share a similar programming style with modern high level languages. So a student who has learned how to write readable Pascal, C/C++, or Java programs will be able to write readable HLA programs with almost no additional study. Contrast this with the style guide I've written for (MASM) assembly language programmers that is quite a bit different than high level languages and takes a while to master.

Another factor many people don't consider is the evaluation of a programming project. At UCR we are given about 1.5-2 hours per student per quarter of reader (student grader) time to grade projects. Experienced readers who can grade (or want to grade) assembly language projects are few and far inbetween. Most readers get "stuck" with grading the assembly class rather than volunteer for the job. The fact that most student assembly language projects have a horrible programming style and are hard to read only exacerbates this situation. HLA helps solve this problem. Since good HLA programming style is very similar to good C/C++ style, UCR's readers have a much easier time reading the projects and evaluating their programming style. Also, since the students have (presumably) learned good programming style in the prerequisite course(s), they tend to write easier to read HLA programs than MASM programs. This lets me assign more projects without fear of exceeding my reader budget each quarter.

HLA's advantages are easily summed up by a complaint I had from a student once. She said "HLA drives me nuts. It's so similar to C++ that I often get confused and try out something that would work in C++ only have have the HLA compiler reject it." I agreed with this student that this was a bit of a problem, but I also mentioned "what about all the times you've tried something from C++ and it HAS worked?" She thought about it for a moment and walked away agreeing with my assessment of her complaint. Had this student been learning assembly the traditional way, she wouldn't have bothered to try anything. She would had to have spent extra time learning how to achieve what she wanted by reading an assembly text or she would have missed out on the opportunity to actually learn something new. HLA's similarity to C++ encouraged her to try something out on her own. The experiments weren't always successful, but in those cases where they were, she benefited greatly from this. This anecdote, more than any other, sums up what my goals with HLA were and describes the success I believe I have achieved with it.

How to Learn Assembly Programming Using HLA

Of course, a compiler without a language reference manual and tutorial is useless. This document will provide a reference to the HLA programming language. It is not, however, appropriate pedagogy for beginners (it's more suitable for those who already know assembly language programming and wish to learn HLA's syntax). A better text for beginners is "The Art of Assembly Language Programming/Win32 Edition." This provides a complete college level textbook that teaches assembly language programming from the ground up using HLA. You can find a copy of "AoA" on Webster at http://webster.cs.ucr.edu. Webster also contains the latest version of HLA as well as tons of HLA sample source code. That's the first place you should go for information on learning HLA.

Legal Notice

The HLA v1.xx implementation is a prototype intended to test language design and implementation features. I (Randall Hyde) have placed this code and language design in the public domain so others may benefit from this work. However, keep in mind that, as a prototype, HLA is not up to contemporary commercial standards for software quality. It is your responsibility to evaluate whether HLA is suitable for whatever purpose you intend its use.

At any given time there are several known and unknown defects in this software. Some may be corrected in later releases of HLA v1.x, some may never be corrected in the v1.x series. I (Randall Hyde) do not warrant or guarantee this software in any way. In particular, you cannot expect corrections of any given defect in the system. Obviously, I try to fix known problems (if possible), but I refuse to be held legally responsible for such defects in the software.

Note that defects will come in three general varieties: defects that cause the compiler to fail or generate bad code, defects in support code (e.g., the HLA Standard Library or other example code), and defects in the documentation accompanying this product. No guarantee applies to anything in HLA, especially in these three areas.

The purpose of developing a prototype implementation of the HLA language was to try out language design and implementation ideas. The prototype phase of HLA development is rapidly coming to an end and an "official" HLA language design will be forthcoming. HLA v2.0 will implement this new language. The only guarantees I make about compatibility between HLA v1.x and HLA v2.0 is that there will be some incompatibilities. The exact nature and magnitude of those incompatibilities is unknown at this point, but it is safe to assume that no HLA v1.x program will compile under HLA v2.0 without at least some minor source code changes. So please don't get the idea that any investment you make in HLA source code will be protected in v2.0 (note: after the release of v2.0 this is a relatively safe assumption to make, though there will still be no guarantees). The changes in the source language between HLA v1.25 and HLA v1.26, and between HLA v1.80 and v1.82 are but a small harbinger of the changes that will occur between v1.x and v2.0.

The HLA Standard Library may also undergo changes between v1.x and v2.0. So expect this to happen and plan accordingly if you intend to port your HLA code to v2.0 eventually.

Because HLA is constantly changing (typical of a prototype), it is very difficult to keep the documentation in phase with the language. You can expect this documentation (and all HLA documentation) to contain omissions (e.g., of new features that have yet to be documented), discussion of features removed from HLA, and incorrect descriptions of HLA features. Every attempt will be made to keep the documentation in phase with the software, but like so many free software projects, lack of time and motivation prevents perfection2.

This software is not fit for use in mission-critical or life-support software systems. This software is principally intended for evaluation and educational (i.e., learning assembly language) purposes only. It has been successfully used to develop commercial applications and it has been successfully used in educational environments, but again, you are personally responsible for determining the fitness of this software and documentation for your particular application and you must take responsibility for that choice.

HLA's current design makes use of other software tools that I (Randall Hyde) did not write. These tools include the FASM assembler, the MASM assembler, the Microsoft Linker, the Microsoft Librarian, the Pelles C linker, the Pelles C librarian, Borland's Turbo Assembler, Borland's Turbo Linker, Borland's Turbo Librarian, and the Free Software Foundations ld and as programs. Because some of these tools are commercial products and are covered by various license agreements, not all of these tools come with the HLA distribution. For example, if you want ot use the Microsoft or Borland tools, you'll have to obtain copies of them from some other source. Note that using HLA does not require the Microsoft or Borland tools; HLA is simply compatible with these tools if you already own them and would prefer to use them. HLA does ship with all the tools you need to effectively use HLA; the use of these non-free tools is optional. Licenses for all the products shipped with HLA are included in the package and you may view the licenses any time by specifying the "-license" command-line option when running the HLA program.

Installing HLA Under Windows

New Easy Installation:

As of HLA v1.58, you can find a program titled "hlasetup.exe" on Webster. Running this application automatically installs HLA on your system. You must first install MASM32 (described below) and then run hlasetup.exe. That's all there is to it. Those who wish to exercise complete control over the placement of the HLA executables can still choose to manually install HLA, but this is not recommended for first-time users.

Original Manual Installation:

HLA can operate in one of several modes. In the standard mode it converts an HLA source file directly into an object file like most assemblers. In other modes, it has the ability to translate HLA source code into another source form that is compatible with several other assemblers, such as MASM, TASM, FASM, and Gas. A separate assembler, such as MASM, can compile that low-level intermediate code to produce an object code file. Strictly speaking, this step (converting to a low-level assembler format and assembling via MASM/FASM/GASM/Gas is not necessary, but there are some times when it's advantageous to work in this manner. Finally, you must link the object code output from the assembler using a linker program. Typically you will link the object code produced by one or more HLA source files with the HLA Standard Library (hlalib.lib) and, possibly, several operating system specific library files (e.g., kernel32.lib under Win32). Most of this activity takes place transparently whenever you ask HLA to compile your HLA source file(s). However, for the whole process to run smoothly, you must have installed HLA and all the support files correctly. This section will discuss how to set up HLA on your system.

First, you will need an HLA distribution for your particular Operating System. These instructions describe installation under Windows; see the next section if you're using Linux. The latest version of HLA is always available on Webster at http://webster.cs.ucr.edu. You should go there and download the latest version if you do not already possess it.

The HLA package contains the HLA compiler, the HLA Standard Library, and a set of include files for the HLA Standard Library. It also includes copies of the FASM assembler, the Pelles C librarian and linker, and some other tools. These tools will let you produce executable files under Windows and Linux. However, it's still a good idea for Windows' users to grab a copy of the MASM assembler and linker. The easiest way to get all the MASM files you need is to download the "MASM32" package from http://www.pdq.com.au/home/hutch/masm.htm or any of the other places on the net where you can find the MASM32 package. Once you unzip this file, it's easy to install the MASM32 package using the install program it supplies.

Here are the steps I went through to install MASM32 on my system:

• I downloaded masm32v6.zip from the URL above (later versions are probably okay too, although there is a slight chance that the installation will be different.
• I double-clicked on the masm32v6.zip file (which runs WinZip on my system).
• I choose to extract "install.exe". I told WinZip to extract this file to C:\.
• I double-clicked on the "install.exe" icon and selected the "C:" drive in the window that popped up. Then I hit the install button and waited while MASM32 extracted all the pertinent files. This produced a directory called "MASM32". MASM32 is a powerful assembly language development subsystem in its own right; but it uses the traditional MASM syntax rather than the HLA syntax. So we'll use MASM32 mainly for the assembler, linker, and library files. MASM32 also includes a simple editor/IDE and several other tools that may be useful to an HLA programmer. Feel free to check this software out and see if it is useful to you. For now, note that the executable files you will ultimately need are ML.EXE, ML.ERR, LINK.EXE, and a couple of DLLs. You can find them in the MASM32\BIN subdirectory. Leave them there for the time being. The MASM32\LIB directory also contains many Win32 library files you will need. Again, leave them alone for the time being.
• Next, if you haven't already done so, download the HLA executables file from Webster at http://webster.cs.ucr.edu. On Webster you can download several different ZIP files associated with HLA from the HLA download page. The "Executables" is the only one you'll absolutely need; however, you'll probably want to grab the documentation and examples files as well. If you're curious, or you want some more example code, you can download the source listings to the HLA Standard Library. If you're really curious (or masochistic), you can download the HLA compiler source listings to (this is not for casual browsing!).
• I downloaded the 1_90HLA.zip file while writing this. Most likely, there is a much later version available as you're reading this. Be sure to get the latest version. I chose to download this file to my "C:\" root directory.
• After downloading 1_90HLA.zip to my C: drive, I double-clicked on the icon to run WinZip. I selected "Extract" and told WinZip to extract all the files to my C:\ directory. This created an "HLA" subdirectory in my root on C: with two subdirectories (include and lib) and two EXE files (HLA.EXE and HLAPARSE.EXE. The HLA program is a "shell" program that runs the HLA compiler (HLAPARSE.EXE), MASM (ML.EXE), the linker (LINK.EXE), and other programs. You can think of HLA.EXE as the "HLA Compiler".
• Next, I set some environment variables :

 

path=c:\hla;c:\masm32\bin;%path%

set lib=c:\masm32\lib;c:\hla\hlalib;%lib%

 

• HLA is a Win32 Console Window program. To run HLA you must open up a console Window. Under Windows 2000, Microsoft has hidden this away in Start->Programs->Accessories->Command Prompt. You might find it in another location. You can also start the command prompt processor by selecting Start->Run and entering "cmd".
• At this point, HLA should be properly installed and ready to run. Try typing "HLA -?" at the command line prompt and verify that you get the HLA help message. If not, go back and figure out what you've done wrong up to this point (it doesn't hurt to start over from the beginning if you're lost).
• Thus far, you've verified that HLA.EXE is operational. Now try the following command: "ML /?" This should run the Microsoft Macro Assembler (MASM) and display the help screen. You can ignore the information that appears; you will probably never need to know this stuff.
• Next, let's verify the correct operation of the linker. Type "link /?" and verify that the linker program runs. Again, you can ignore the help screen that appears. You don't need to know about this stuff.
• Now it's time to try your hand at writing an honest to goodness HLA program and verify that the whole system is working. Here's the canonical "Hello World" program written in HLA. Enter it into a text editor and save it using the filename "HW.HLA":

 

program HelloWorld;

#include( "stdlib.hhf" )

begin HelloWorld;

 

stdout.put( "Hello, World of Assembly Language", nl );

 

end HelloWorld;

 

• WARNING: if you are a notepad.exe user, note that in certain Windows modes notepad will always append ".txt" to the end of the filename you specify. This will cause HLA to fail if you attempt to compile such source files (because HLA expects the filename to end with ".hla" not ".txt". Be aware of this problem.
• Make sure you're in the same directory containing the HW.HLA file and type the following command at the "C:>" prompt: "HLA -v HW". The "-v" option tells HLA to produce VERBOSE output during compilation. This is helpful for determining what went wrong if the system fails somewhere along the line. This command should output similar to the following (this changes all the time, so don't expect it to be exact):

 

HLA (High Level Assembler)

Released to the public domain by Randall Hyde.

Version Version 1.59 build 8109 (prototype)

 

Files:

1: hw.hla

 

Compiling "hw.hla" to "hw.asm"

 

Assembling hw.asm via "ml /c /coff /Cp hw.asm"

 

Microsoft (R) Macro Assembler Version 6.14.8444

Copyright (C) Microsoft Corp 1981-1997. All rights reserved.

 

Assembling: hw.asm

Linking via "link -subsystem:console /heap:0x1000000,0x1000000 /stack:0x1000000,0x1000000 /BASE:0x3000000 /machine:IX86 -entry:?HLAMain @hw.link -out:hw.exe kernel32.lib user32.lib c:\hla\hlalib\hlalib.lib hw.obj"

Microsoft (R) Incremental Linker Version 5.12.8078

Copyright (C) Microsoft Corp 1992-1998. All rights reserved.

 

/section:.text,ER

/section:readonly,R

/section:.edata,R

/section:.data,RW

/section:.bss,RW

 

• If you get all of this output, you're in business.

Manually installing HLA is a complex and slightly involved process. Fortunately, the hlasetup.exe program automates almost everything so that you don't have to worry about changing registry settings and things like that. If you're a first-time HLA user, you definitely want to use this method to install HLA. Manual installation is really intended for upgrades as new versions of HLA appear. You do not have to change the environment variables to install a new version of HLA, simply extract the executable files over the top of your existing installation and everything will work fine.

The most common two problems people have running HLA involve the location of the Win32 library files and the choice of linker. During the linking phase, HLA (well, link.exe actually) requires the kernel32.lib, user32.lib, and gdi32.lib library files. These must be present in the pathname(s) specified by the LIB environment variable. If, during the linker phase, HLA complains about missing object modules, make sure that the LIB path specifies the directory containing these files. If you're a MS VC++ user, installation of VC++ should have set up the LIB path for you. If not, then locate these files (they are part of the MASM32 distribution) and copy them to the HLA\HLALIB directory (note that the ihla.bat file includes c:\hla\hlalib as part of the LIB path).

Another common problem with running HLA is the use of the wrong link.exe program. Microsoft has distributed several different versions of link.exe; in particular, there are 16-bit linkers and 32-bit linkers. You must use a 32-bit segmented linker with HLA. If you get complaints about "stack size exceeded" or other errors during the linker phase, this is a good indication that you're using a 16-bit version of the linker. Obtain and use a 32-bit version and things will work. Don't forget that the 32-bit linker must appear in the execution path (specified by the PATH environment variable) before the 16-bit linker.

For more information, please see the sections on HLA Internal Operation and Customizing HLA.

Installing HLA Under Linux

HLA is not a stand alone program. It is a compiler that translates HLA source code into either object code or a lower-level assembly language that Gas or FASM must process. Finally, you must link the object code output using a linker program such as the Linux ld linker. Typically you will link the object code produced by one or more HLA source files with the HLA Standard Library (hlalib.a). Most of this activity takes place transparently whenever you ask HLA to compile your HLA source file(s). However, for the whole process to run smoothly, you must have installed HLA and all the support files correctly. This section will discuss how to set up HLA on your system.

First, you will need an HLA distribution for Linux. Please see Webster or the previous section if you're attempting to install HLA on a different OS such as Windows. The latest version of HLA is always available on Webster at http://webster.cs.ucr.edu. You should go there and download the latest version if you do not already possess it.

Under Linux, HLA can operate in one of three modes: it can directly produce object files (.o files) that you can link with ld; it can produce a low-level assembly language output file that you can assemble using the Free Software Foundation's Gas assembler, or it can produce a low-level assembly language output file that you can assemble using FASM. The HLA package contains the HLA compiler, FASM, the HLA Standard Library, and a set of include files for the HLA Standard Library. If you write an HLA program want Gas to process it, you'll need to make sure you have a reasonable version of Gas available (Gas is available on most Linux distributions, so this shouldn't be a problem). Note that the HLA Gas output can only be assembled by Gas v2.10 or later (so you will need the 2.10 or later binutils distribution). Note that (apparently) HLA does not work with 64-bit versions of Gas, so make sure you're using a 32-bit version of Gas with HLA or use the object code output feature or use FASM.

Here's the steps I went through to install HLA on my Linux system:

• First, if you haven't already done so, download the HLA executables file from Webster at http://webster.cs.ucr.edu. On Webster you can download several different ZIP files associated with HLA from the HLA download page. The "Linux Executables" is the only one you'll absolutely need; however, you'll probably want to grab the documentation and examples files as well. If you're curious, or you want some more example code, you can download the source listings to the HLA Standard Library. If you're really curious (or masochistic), you can download the HLA compiler source listings to (this is not for casual browsing!).
• I downloaded the hla.tar.gz file for HLA v1.90 while writing this. Most likely, there is a much later version available as you're reading this. Be sure to get the latest version. I chose to download this file to my "/usr/hla" directory; you can put the file whereever you like, though this documentation assumes that all HLA files wind up in the "/usr/hla/..." directory tree. Note: the .tar.gz file downloads into /usr/hla. If you want the files placed somewhere else, unpack them to this directory and them move them.
• After downloading hla.tar.gz, I executed the following shell command: "gzip -d hla.tar.gz". Once decompression was complete, I extracted the individual files using the command "tar xvf hla.tar". This extracted several executable files (e.g., "hla" and "hlaparse") along with three subdirectories (include, hlalib, and hlalibsrc). The HLA program is a "shell" program that runs the HLA compiler (hlaparse), gas (as), FASM (fasm), the linker (ld), and other programs. You can think of hla as the "HLA Compiler". It would be a real good idea, at this point, to set the permissions on "hla" and "hlaparse" so that everyone can read and execute them. You should also set read and execute permissions on the two subdirectories and read permissions on all the files within the directories (if this isn't the default state). Do a "man chmod" from the Linux command-line if you don't know how to change permissions.
• If you prefer a more "Unix-like" environment, you could copy the hla and hlaparse (and other executable) files to the "/usr/bin" or "/usr/local/bin" subdirectory. This step, however, is optional
• Next, (logged in as a plain user rather than root or the super-user), I edited the ".bashrc" file in my home directory ("/home/rhyde" in my particular case, this will probably be different for you). I found the line that defined the "path" variable, it originally looked like this on my system
  "PATH=$DBROOT/bin:$DBROOT/pgm:$PATH"
  I edited this line to add the path to the HLA directory, producing the following:
  "PATH=$DBROOT/bin:$DBROOT/pgm:/usr/hla":$PATH
  Without this modification, Linux will probably not find HLA when you attempt to execute it unless you type a full path (e.g., "/usr/hla/hla") when running the program. Since this is a pain, you'll definitely want to add "/usr/hla" to your path. Of course, if you've chosen to copy hla and hlaparse to the "/usr/bin" or "/usr/local/bin" directory, chances are pretty good you won't have to change the path as it already contains these directories.
• Next, I added the following four lines to ".bashrc" (note that Linux filenames beginning with a period don't normally show up in directory listings unless you supply the "-a" option to ls):
  hlalib=/usr/hla/hlalib/hlalib.a
  export hlalib
  hlainc=/usr/hla/include
  export hlainc
  These four lines define (and export) environment variables that HLA needs during compilation. Without these environment variables, HLA will probably complain about not being able to find include files, or the linker (ld) will complain about strange undefined symbols when you attempt to compile your programs. Note that this step is optional if you leave the library and include files installed in the /usr/hla directory subtree.
• 
  Optionally, you can add the following two lines to the .bashrc file (but make sure you've created the /tmp directory if you do this):
  
  hlatemp=/tmp
  export hlatemp
  
  After saving the ".bashrc" shell, you can tell Linux to make the changes to the system by using the command:
  
  source .bashrc
  
  Note: this discussion only applies to users who run the BASH shell. If you are using a different shell (like the C-Shell or the Korn Shell), then the directions for setting the path and environment variables differs slightly. Please see the documentation for your particular shell if you don't know how to do this.
• At this point, HLA should be properly installed and ready to run. Try typing "HLA -?" at the command line prompt and verify that you get the HLA help message. If not, go back and figure out what you've done wrong up to this point (it doesn't hurt to start over from the beginning if you're lost).
• Now it's time to try your hand at writing an honest to goodness HLA program and verify that the whole system is working. Here's the canonical "Hello World" program written in HLA. Enter it into a text editor and save it using the filename "hw.hla":

 

program HelloWorld;

#include( "stdlib.hhf" )

begin HelloWorld;

 

stdout.put( "Hello, World of Assembly Language", nl );

 

end HelloWorld;

 

• Make sure you're in the same directory containing the "hw.hla" file and type the following command at the prompt: "hla -v hw". The "-v" option tells HLA to produce VERBOSE output during compilation. This is helpful for determining what went wrong if the system fails somewhere along the line. This command should produce output like the following:

 

HLA (High Level Assembler) Parser

Copyright 2001, by Randall Hyde, all rights reserved.

Version Version 1.32 build 4895 (prototype)

-t active

File: t.hla

 

Compiling "t.hla" to "t.asm"

HLA (High Level Assembler)

Copyright 1999, by Randall Hyde, all rights reserved.

Version Version 1.32 build 4895 (prototype)

ELF output

Using GAS assembler

GAS output

-test active

 

Files:

1: t.hla

 

Compiling 't.hla' to 't.asm'

using command line [hlaparse -v -sg -test "t.hla"]

 

Assembling "t.asm" via [as -o t.o "t.asm"]

Linking via [ld -o "t" "t.o" "/usr/hla/hlalib/hlalib.a"]

 

Versions of HLA may appear for other Operating Systems (beyond Windows and Linux) as well. Check out Webster to see if any progress has been made in this direction. Note a very unique thing about HLA: Carefully written (console) applications will compile and run on all supported operating systems without change. This is unheard of for assembly language! So if you are using multiple operating systems supported by HLA, you'll probably want to download files for all supported OSes.

For more information, please see the sections on HLA Internal Operation and Customizing HLA.

Using HLA with the RadASM Integrated Development Environment

Please see the separate RadASM/HLA user's manual for details concerning the use of the RadASM integrated development environment. The hlasetup.exe program automatically installs RadASM on your system (and properly sets up the associated INI files). When upgrading to newer versions of RadASM, be sure to save the radasm\hla folder and the hla.ini files before copying the new version of RadASM over the old version. Again, for more information concerning RadASM, see the separate RadASM user's manual.

Using HLA with the HIDE Integrated Development Environment

Sevag has written a nice HLA-specified integrated development environment for HLA called HIDE (HLA IDE). This one is a bit easier to install, set up, and use than RadASM (at the cost of being a little less flexible). HIDE is great for beginners who want to get up and running with a minimal amount of fuss. You can find HIDE at the HIDE home page:

http://www.geocities.com/kahlinor/HIDE.html

 

HLA Internal Operation

To effectively use HLA, it helps to understand how HLA translates HLA source files into executable machine code. This information is particularly useful if you install HLA incorrectly and you cannot successfully compile a simple demo program. Beyond that, this information can also help you take advantage of more advanced HLA and OS features.

As noted earlier in this document, HLA is not a single application; the HLA system is a collection of programs that work together to translate your HLA source files into executable files. This is not unusual, most compilers and assemblers provide only part of the conversion from source to executable (e.g., you still have to run a linker with most compilers and assemblers to produce an executable).

The HLA system offers a rich set of different configurations that allow you to mix and match components to efficiently process your assembly language applications. First of all, HLA is relatively portable. The compiler itself is written with Flex, Bison, C/C++, along with some platform-independent assembly language code (written in HLA, of course). This makes it fairly easy to move the compiler from one operating system to another. Currently, HLA is supported under Windows and Linux. A couple of different FreeBSD ports have been done (though official support for FreeBSD must wait until the HLA standard library gets ported to FreeBSD). Plans include porting HLA to MacOS (Intel Macintoshes) at some point in the future. Even within a single operating system, HLA offers multiple configurations that you can employ, based on your needs and desires. This section will describes some of the possible configurations you might create.

The compilation of a typical HLA source file using a command line such as "hla hw" goes through three or four major phases:

• The HLA.EXE (Windows) or hla (other OSes) program processes command-line parameters and acts as a "traffic cop" directing the execution of the remaining components of the HLA system.
• The HLAPARSE.EXE (Windows) or hlaparse (other OSes) program is responsible for translating the HLA source file into either an object file or into the syntax of some other assembler. Usually, the HLAPARSE program is run automatically by some other program such as HLA.EXE/hla or HIDE (the HLA Integrated Development Environment). You would not normally run HLAPARSE directly from a command-line (though it is certainly possible to do this if you are so inclined).
• If you elect to have HLA produce an assembly language output file rather than an object module, then the next step towards producing an executable file is to run the associated assembler on the source output that HLA produced. This step isn't strictly necessary because HLA can produce an object file directly without using some external assembler, but there are some (rather esoteric) reasons why you might want to go through some other assembler rather than having HLA directly produce the object file. Generally, the HLA.EXE (hla) program will automatically run the assembler for you.
• The last step is to run a linker to combine the object module the previous steps created with the HLA Standard Library and any other necessary object modules for the project. The output of the linkage step is an executable file (assuming, of course, there were no errors in the compilation of your program). Generally, the HLA.EXE (hla) program will automatically run the linker for you.

There is a fifth, optional, step that can also take place under Windows. If you are creating an application that makes use of compiled resources, as the fourth step (before the linking stage) the HLA.EXE (Windows only) program can run a resource compiler to translate those resources into an object module (.res) that the linker can link into your final executable.

As it turns out, the HLA system can employ a wide variety of linkers, librarians, assemblers, and other tools based on the underlying operating system. Here is the list of tools that HLA has been qualified with:

Under Windows:

• Microsoft's MASM assembler
• Microsoft's linker
• Microsoft's resource compiler
• Microsofts LIB library manager
• The Flat Assembler (FASM)
• Pelles C POLINK linker
• Pelles C POLIB library manager
• Pelles C PORC resource compiler
• Borland's Turbo Assembler

Note: you can use Borland's TLINK and TLIB utilities with HLA, but you will have to manually run these applications; the HLA system will not automatically execute them.

 

Under Linux:

• The Free Software Foundation's (FSF) Gas assembler (as)
• FSF's linker (ld)
• The Flat Assembler (FASM)

 

A couple of obvious questions that might come up: "Why provide all these options? Why not simply pick a single configuration and go with that?" Well, as it turns out, there are advantages and disadvantages to each configuration and allowing multiple configurations affords you the most flexibility when writing code.

Using the HLA Command-Line Compiler

This section assumes that you are running the standard version of HLA. If you want to run "freeHLA" rather than the standard version under Windows, then simply substitute the command "fhla" for "hla" throughout this documentation.

Once you've installed HLA and verified that it is operational, you can run the HLA compiler. The HLA compiler consists of two executables: hla(.exe)3, which is a shell that processes command line arguments, compiles ".hla" files to ".asm" files, assembles the ".asm" files by calling an assembler, and links the resulting files together using a linker program; the second executable is hlaparse(.exe) which compiles a single ".hla" file to an assembly language file. Generally, you would only run hla(.exe). The hla(.exe) program automatically runs the hlaparse(.exe) and assembler/linker programs. The hla(.exe) command uses the following syntax:

 

hla optional_command_line_parameters Filename_list

 

The filenames list consists of one or more unambiguous filenames having the extension: ".hla", ".asm" or ".obj"/".o"4. HLA will first run the hlaparse(.exe) program on all files with the HLA extension (producing files with the same basename and an ASM extension). Then HLA runs the assembler on all files with the ".asm" extension (including the files produced by hlaparse). Finally, HLA runs the linker to combine all the object files together (including the ".obj"/".o" files the assembler produces). The ultimate result, assuming there were no errors along the way, is an executable file (with an EXE extension under Windows, with no extension under Linux).

 

HLA supports the following command line parameters:

 

options:

-@ Do not generate linker response file.

-@@ Always generate a linker response file.

-axxxxx Pass xxxxx as command line parameter to assembler.

-dxx Define VAL symbol xx to have type BOOLEAN and value TRUE.

-dxx=yy Define VAL symbol xx to have type STRING and value "yy".

-e:name Executable output filename.

-i:path Specifies path to HLA include file directory.

-lib:path Specifies path to the HLALIB.LIB file.

-license Displays copyright and license info for the HLA system.

-lxxxxx Pass xxxxx as command line parameter to linker.

(note: use -lLIBPATH=<path> to specify path to HLALIB.LIB)

-m Create a map file during link

-p:path Specifies path to hold temporary working files.

-r:name <name> is a text file containing cmd line options.

-obj:path Specifies path to place object files.

-o:omf Produce OMF files (default - tasm).

-o:coff Produce win32 COFF files.

-o:elf Produce Linux ELF files.

-s Compile to .ASM files only.

-sf Compile to FASM files only.

-sg Compile to Gas files only.

-sm Compile to MASM files only (default for -s).

-st Compile to TASM files only.

-c Compile and assemble to object files only.

-co Compile and assemble to object file using internal version of FASM

-cf Compile and assemble to object file using FASM

-cm Compile and assemble to object file using MASM

-ct Compile and assemble to object file using TASM

-cg Compile and assemble to object file using GAS

-xo Compile/assemble/link to executable using internal version of FASM

-xf Compile/assemble/link to executable using FASM

-xm Compile/assemble/link to executable using MASM

-xt Compile/assemble/link to executable using TASM

-xg Compile/assemble/link to executable using GAS

-sym Dump symbol table after compile.

-test Send diagnostic info to stdout rather than stderr (This

option is intended for HLA test/debug purposes).

-v Verbose compile.

-level=h High-level assembly language

-level=m Medium-level assembly language

-level=l Low-level assembly language

-level=v Machine-level assembly language (very low level).

-w Compile as windows app (default is console app).

-? Display this help message.

 

Note that HLA ignores case when processing command line parameters (unlike typical Linux programs). Hence, "-s" is equivalent to "-S" (for example) when specifying a command line parameter.

-@
-@@

Under Windows, HLA will produce a "linker response file" that it supplies to the Microsoft LINK.EXE (or POLINK.EXE) program during the link phase. This linker response file contains necessary segment declarations and other vital linker information. By default, HLA uses any existing ".LINK" file whenever you run the compiler; it will create a new "xxx.link" file only if one does not already exist. The "-@" option tells HLA not to create a new ".LINK" file, even if one does not already exist. The "-@@" option tells HLA to always create a ".link" file, even if one already exists.

If you specify multiple ".HLA" filenames on the command line, HLA only generates a single ".LINK" file using the name of the first ".HLA" file it encounters. Linux's ld program does not require this linker response file, so the Linux version of HLA does not produce this file.

 

-r:filename

The "-r:filename" option lets you specify a response file containing a sequence of HLA command-line parameters. The file specified after this option must contain a seqeuence of HLA command-line parameters, one per line, which HLA executes exactly as though they were specified on the command line. E.g.,

 

sampleFile.resp:

 

-cf

sampleFile.hla

 

The following command treats each of the above lines as separate HLA command-line parameters:

 

hla -r:sampleFile.resp

 

 

-aXXXXX

The -aXXXXX option lets you pass assembler-specific command line options to the assembler during the assembler phase. This option is ignored if you use one of the -s options. One common form of this command often used with the MASM assembler is "-aFi -aFz" that tells MASM to generate debugging information in the object file (for use with the OllyDbg debugger program).

 

 

-c

The -c option tells HLA to run the hlaparse compiler and the (default) assembler, producing ".obj"/".o" files. HLA will process all filenames on the command line that have ".hla" or ".asm" extension, but it will ignore any filenames with ".obj" extensions. If you compile an HLA unit without compiling an HLA program at the same time, you will need to use this option or the linker will complain about not finding the main program. You may specify the ".obj"/".o" file format using the COFF or OMF command line options (MASM only, TASM always produces OMF files, Gas always produces ELF files, and FASM always produces COFF files).

One common use of this option is to compile HLA units to OBJ files. Since HLA units do not contain a main program, you cannot compile an HLA unit directly to an executable. To compile an HLA unit separately (i.e., without compiling an HLA main program during the same HLA.EXE invocation) you must specify the "-c" option or the compilation will generate an error when it attempts to link the program.

A second reason for using the "-c" option is because you want to explicitly run the linker yourself and supply linker command line options that are different than those that HLA automatically provides.

 

-co

Compiles an HLA source file directly to an object file using the internal version of FASM (overriding the default assembler). Otherwise, the actions are identical to -c.

-cf

Compiles an HLA source file to a FASM source file and then compile this source file to an object file using an external version of FASM (overriding the default assembler). Otherwise, the actions are identical to -c.

 

-cm

Compiles an HLA source file to a MASM source file and then compile this source file to an object file using MASM (overriding the default assembler). Otherwise, the actions are identical to -c. Note that this option is available only under Windows.

-ct

Compiles an HLA source file to a TASM source file and then compile this source file to an object file using TASM (overriding the default assembler). Otherwise, the actions are identical to -c. Note that this option is available only under Windows.

-cg

Compiles an HLA source file to a Gas source file and then compile this source file to an object file using Gas (overriding the default assembler). Otherwise, the actions are identical to -c. Note that this option is available only under non-Windows operating systems.

 

-xo

Compiles an HLA source file directly to an object file using the internal version of FASM (overriding the default assembler). Links the result to produce an exectuable file. Under Windows, this option uses POLINK as the default linker unless overriden. Under Linux, this command option uses the GNU LD linker.

-xf

Compiles an HLA source file to a FASM source file and then compile this source file to an object file using an external version of FASM (overriding the default assembler). Then links the result to produce an executable file. Under Windows, this option uses POLINK as the default linker unless overriden. Under Linux, this command option uses the GNU LD linker.

-xm

Compiles an HLA source file to a MASM source file and then compile this source file to an object file using MASM (overriding the default assembler). Links the resulting object file to produce an executable (Microsoft's linker is the default linker). Note that this option is available only under Windows. Under Linux, this command-line option is identical to "-sf".

-xt

Compiles an HLA source file to a TASM source file and then compile this source file to an object file using TASM (overriding the default assembler). Then links the result to produce an executable (Microsoft's linker is the default linker it uses). Under Linux, this option only produces a source file; it will not run TASM or the linker to produce an executable (that is, this command is equal to "-st" under Linux).

-cg

Compiles an HLA source file to a Gas source file and then compile this source file to an object file using Gas (overriding the default assembler). Runs GNU's LD linker to produce an executable (under Linux). If you use this option under Windows, it will not assemble and link the result (that is, this option is equal to "-sg" under Windows).

 

-d:XXXXX{=YYYYY}

The -dXXXXX option tells HLA to define the symbol XXXXX as a boolean VAL constant and initialize it with the value TRUE. Generally you use such symbols to control the emission of code during assembly using statements like "#if( @defined( XXXXX )) ..."

The -dXXXX=YYYY option tells HLA to define the symbol XXXX as a string VAL constant and give it the initial value "YYYY".

 

 

-e:name

By default, HLA creates an executable filename using the extension ".exe" (Windows) or without an extension (Linux) and the basename of the first filename on the command line. You can use the -e name option to specify a different executable file name.

 

 

-lXXXXX

The -lXXXXX option passes the text XXXXX on to the linker as a command line option. One common command to pass to the Microsoft linker is "-lDEBUG -lDEBUGTYPE:COFF" that tells the linker to generate debugging information in the object file (for use with the OllyDbg debugger program).

 

 

-m

The -m option tells the Microsoft linker or POLINK to produce a map file during the link phase. This is equivalent to the "-lmap" option. The Linux version of HLA ignores this option.

 

 

-o:omf

The -o:omf option tells the underlying assembler (MASM or TASM) to produce an Object Module Format (OMF) OBJ file. This option is generally applicable only to MASM since TASM always produces OMF files. This option is not legal when using the Gas or FASM assemblers.

 

-o:coff

The -o:coff option instructs the assembler to generate a COFF OBJ file. This option is the default for MASM/FASM and may not be available for other assemblers. This option is only available under Windows.

 

-s

The -s option tells the HLA/FHLA program to run only the hlaparse compiler to produce an assembly language source file; HLA will not run an assembler or linker. As a result, HLA ignores any ".asm" or ".obj" filenames you supply on the command line. This option is useful if you wish to view the output of an HLA compilation without producing any actual object code. If you specify this option with a version of HLA that would normally produce an object file output, HLA will emit a FASM-compatible source file instead.

 

-st

The -st option tells HLA to produce TASM-compatible assembly and stop after compilation. Note that the source file you produce will contain Windows-specific code, even if you produce the TASM-compatible source file under Linux.

 

-sm

The -sm option tells HLA to produce MASM-compatible assembly and stop after compilation. Note that the source file you produce will contain Windows-specific code, even if you produce the MASM-compatible source file under Linux.

 

-sg

The -sg option tells HLA to produce Gas-compatible assembly and stop after compilation. Note that the source file will contain Linux-specific code even if you produce the source file under Linux.

 

-sf

The -sf option tells HLA to produce FASM-compatible assembly and stop after compilation. Unlike the other source output forms, FASM output is specific to the underlying OS on which you ran HLA. That is, using the -sf option under Windows produces Windows-specific output, using the -sf option under Linux produces a Linux-compatible file.

 

-sym

The -sym option dumps the symbol table after compiling each file with an HLA extension. This option is primarily intended for testing and debugging the HLA compiler; however, this information can be useful to the HLA programmer on occasion.

 

 

-test

The -test option is intended for hlaparse testing and debugging purposes only. It causes the compiler to send all error messages to the standard output device rather than the standard error device. This allows the test code to redirect all errors to a text file for comparison against other files.

 

-v

The -v option (verbose) causes HLA to print additional information during compile to show the progress of the compilation. Due to a bug in MASM, if you do not specify the -v option the compilation isn't completely quiet. MASM will still output data to the standard error device even in quiet (non-verbose) mode.

 

 

-w

The -w option informs HLA that you are compiling a standard Windows (GUI) application rather than a console application. By default, HLA assumes that you are compiling a executable that will run from the command window. If you want to write a full Windows application, you will need to supply this option to tell HLA not to link the code for console operation. Obviously, this option doesn't apply to Linux systems.

The "-w" option tells HLA to invoke the linker using the command line option

-subsystem:windows

rather than the default

-subsystem:console

 

This provides a convenient mechanism for those who wish to create win32 GUI applications. Most likely, however, if you wish to create GUI applications, you will run the linker explicitly yourself (as this document will explain), so you'll probably not use the "-w" option very frequently. It's great for some short GUI demos, but larger GUI programs will probably not use this option. This option is only active if HLA compiles the program to an executable. If you compile the program to an OBJ or ASM file, HLA ignores this option.

If you want to develop Win32 GUI apps, take a look at Randy Hyde's book "Windows Programming in Assembly". This book provides the linker commands and makefiles for generation such applications (as well as describing how you actually write such code).

 

 

-p:path

During compilation, HLA produces several temporary files (that it doesn't delete, because they may be of interest to the HLA user). These files have a habit of cluttering up the current work directory. If you prefer, you can tell HLA to place these files in a temporary directory so they don't clutter up your working directory. One way to accomplish this is by using the "-p:dirpath" command line option. For example, the option "-p:c:\hla\tmp" tells HLA to put all temporary files (for the current assembly) into the "c:\hla\tmp" subdirectory (which must exist). Note that you can set also set the temporary directory using the hla "hlatemp" environment variable. The "-p:dirpath" option will override the environment variable (if it exists). See the description of the hlatemp environment variable for more details. Warning: as of FASM v1.51, the use of this option with FHLA is not recommend because of the way FASM handles include files. FASM's author is aware of this issue and will probably do something about it at some point. Until then, you should always CD into the directory where your HLA projects lie and leave all the temporary files in that same folder when running FHLA.EXE. This restriction will be relaxed in a future version of FHLA/FASM.

 

 

-obj:path

During compilation, HLA normally writes all object files to the current working directory. Some programmers have requested a way to specify a different directory for the .OBJ (.o under Linux) files that HLA produces. This is now accomplished using the "-obj:dirpath" command line option. The dirpath item has to be the path to a valid directory. HLA places all object files produced by the compiler and/or resource editor in this directory. Note that, unlike the -p option, there is no environment variable that lets you permanently set this path. You must specify the path on a compilation by compilation basis (use a makefile if you get tired of typing the path in on each compilation).

 

 

-level=h

-level=m

-level=l

-level=v

The -level options enable or disable certain HLA language features. These command-line options are intended for use in programming courses where the instructor needs to batch compile dozens or even hundreds of student projects at one time and the instructor would like a convenient way to ensure that the students aren't using high-level control constructs that are inappropriate for that point in the course (e.g., towards the end of a course, most instructors don't allow the use of various high-level control constructs; some instructors may never allow them). The "-level" command-line options will "turn off" various statements in the HLA language so that the HLA compiler will report an error if the student attempts to use them in a source file.

The default, "-level=h" (high) enables the entire HLA language.

The "-level=m" (medium level) disables high-level language control constructs, such as "if", "while", and "for" but still allows the use of high-level-like procedure calls in the HLA language. Medium-level assembly language also allows the use of exceptions using HLA's try..except..endtry and raise statements.

The "-level=l" (low-level assembly) disables all high-level control constructs other than the exception-handling statements and disables high-level-like procedure calls in HLA. This option also disables automatic stack frame generation and clean up in HLA procedures (that is, the programmer will be responsible for writing that code themselves).

The "-level=v" (very low-level assembly) option disables all high-level control constructs including exception handling. Only machine instructions (and user written macros) are legal in the source file. No high-level control constructs or high-level procedure calls are allowed.

 

 

-?

The -? option cause HLA to dump the list of command line options and immediately quit without further work.

Note that the command line options this document describes are for HLA v1.87 and later only. Earlier versions of HLA used a different command line set. See the documentation for the specific version you're using if you have questions.

 

-license

This command displays license information for the entire HLA system. Although the HLA source code written by Randall Hyde is all public domain, certain components of the HLA system, including the back-end assembler, the linker, and the resource editor, come from other sources. The "-license" command-line parameter lists license information about these other products.

The HLAPARSE Command Line

The HLAPARSE program was written with the expectation that it would always be invoked by some other program (e..g, HLA/HLA.EXE). If you really know what you are doing, you can invoke HLAPARSE from the command line manually. However, other than testing or debugging the HLA system, there really is no need to manually invoke HLAPARSE. All the functionality available from the HLAPARSE command line is also available from the HLA command line. Further, HLA does sanity checks on the command line parameters, fills in optional values, fetches environment variables, and so on. HLAPARSE expects a program like HLA to provide a correct set of command-line parameters; it does not do all the checks on the parameters and if they are incorrect, it may silently fail. As such, it's not intended to be used by end-uses as a matter of course.

There is one case where you might be interested in invoking HLAPARSE via some other means than via the HLA program - when writing a replacement for HLA. A good example of this is the HIDE (HLA Integrated Development Environment) system. If you're interested in writing a program that invokes HLAPARSE directly, then you'll want to learn about the HLAPARSE command-line parameter set. However, this manual is not the place to discuss such things, as they are considered internal to the HLA system. You can view the possible command-line parameters by typing "HLAPARSE -?" at the command line. However, to truly understand their semantics, you'll want to open up the hlaparse.bsn source file and study the code at the very end of the file (warning, it's about 100,000 lines long, so you'll need a good editor to open and look at this code). I (Randy Hyde) will be more than happy to answer any questions via email or via some forum concerning these parameters; they are not, however, something I want to document and give people the impression that they are available for everyday use.

Customizing HLA

Through the use of environment variables and program names, you can create a customized version of HLA that suits your particular needs. The following subsections describe different ways you can optimize HLA for your personal use.

HLA Language Elements

Starting with this section we being discussing the HLA source language. HLA source files must contain only seven-bit ASCII characters. These are text files with each source line record containing a carriage return/line feed (Windows) or a just a line feed (Linux) termination sequence (HLA is actually happy with either sequence, so text files are portable between OSes without change). White space consists of spaces, tabs, and newline sequences. Generally, HLA does not appreciate other control characters in the file and may generate an error if they appear in the source file.

Native (Primitive) Data Types in HLA

HLA provides the following basic primitive types:

 

boolean One byte; zero represents false, one represents true.

Enum One byte; user defined IDs whose value ranges from 0 to 255.

Uns8 Unsigned values in the range 0..255.

Uns16 Unsigned integer values in the range 0..65535.

Uns32 Unsigned integer values in the range 0..4,204,967,295.

Uns64 Unsigned 64-bit integer.

Uns128 Unsigned 128-bit integer.

Byte Generic eight-bit value.

Word Generic 16-bit value.

DWord Generic 32-bit value.

QWord Generic 64-bit value.

TByte Generic 80-bit value.

LWord Generic 128-bit value.

Int8 Signed integer values in the range -128..+127.

Int16 Signed integer values in the range -32768..+32767.

Int32 Signed integer values in the range -2,147,483,648..+2,147,483,647.

Int64 Signed 64-bit integer values.

Int128 Signed 128-bit integer values.

Char Character values.

WChar Unicode character values.

Real32 32-bit floating point values.

Real64 64-bit floating point values.

Real80 80-bit floating point values.

Real128 128-bit floating point values (for SSE/2 instructions).

String Dynamic length string constants. (Run-time implementation: four-byte pointer.)

ZString Zero-terminated dynamic length strings (run-time implementation: four-byte pointer).

Unicode Unicode strings.

CSet A set of up to 128 different ASCII characters (16-byte bitmap).

Text Similar to string, but text constants expand in-place (like #define in C/C++).

Thunk A set of machine instructions to execute.

 

Often, it is convenient to discuss the types above in various groups. This document will often use the following terms:

 

Ordinal: boolean, enum, uns8, uns16, uns32, byte, word, dword, int8, int16, int32, char.

Unsigned: uns8, uns16, uns32, byte, word, dword.

Signed: int8, int16, int32, byte, word, dword.

Number: uns8, uns16, uns32, int8, int16, int32, byte, word, dword

Numeric: uns8, uns16, uns32, int8, int16, int32, byte, word, dword, real32, real64, real80

Enumerated Data Types

HLA provides the ability to associate a list of identifiers with a user-defined type. Such types are known as enumerated data types (because HLA enumerates, or numbers, each of the identifiers in the list to give them a unique value). The syntax for an enumerated type declaration (in an HLA type section, see the description a little later) takes the following form:

typename : enum{ list_of_identifiers };

Here is a typical example:

type

color_t :enum{ red, green, blue, magenta, yellow, cyan, black, white };

 

Internally, HLA treats enumerated types as though they were unsigned integer values (though enum types are not directly compatible with the unsigned types). HLA associates the value zero with the first identifier in the enum list and then attaches sequentially increasing values to the following identifiers in the list. For example, HLA will associate the following values with the color_t symbolic constants:

red 0

green 1

blue 2

magenta 3

yellow 4

cyan 5

black 6

white 7

 

Because each enumerated constant in a given enum list is unique, you may compare these values, use them in computations, etc. Also note that, because of the way HLA assigns internal values to these constant names, you may compare objects in an enumerated list for less than and greater than in addition to equal or not equal.

Note that HLA uses zero as the internal representation for the first symbol of every enum list. HLA only guarantees that the values it associates with enum types is unique for a single type; it does not make this guarantee across different enumerated types (in fact, you're guaranteed that different enum types do not use unique values for their symbol sets). In the following example, HLA uses the value zero for both the internal representation of const0 and c0 . Likewise, HLA uses the value one for both const1 and c1 . And so on...

type

enumType1 :enum{ const0, const1, const2 };

enumType2 :enum( c0, c1, c2 };

 

Note that the enumerated constants you specify are not "private" to that particular type. That is, the constant names you supply in an enumerated data type list must be unique within the current scope (see the definition of identifier scope elsewhere in this document). Therefore, the following is not legal:

type

enumType1 :enum{ et1, et2, et3, et4 };

enumType2 :enum{ et2, et2a, et2b, et2c }; //et2 is a duplicate symbol!

 

The problem here is that both type lists attempt to define the same symbol: et2 . HLA reports an error when you attempt this.

One way to view the enumerated constant list is to think of it as a list of constants in an HLA const section (see the description of declaration sections a little later in this document), e.g.,

const

red : color_t := 0;

green : color_t := 1;

blue : color_t := 2;

magenta : color_t := 3;

yellow : color_t := 4;

cyan : color_t := 5;

black : color_t := 6;

white : color_t := 7;

 

By default, HLA uses eight-bit values to represent enumerated data types. This means that you can represent up to 256 different symbols using an enumerated data type. This should prove sufficient for most applications. HLA provides a special "compile-time variable" that lets you change the size of an enumerated type from one to two or four bytes. In theory, all you've got to do is assign the value two or four to this variable and HLA will automatically resize the storage for enumerated types to handle longer lists of objects. In practice, however, this feature has never been tested so it's questionable if it works well. If you need enumerated lists with more than 256 items, you might consider using HLA const definitions rather than an enum list, just to be on the safe side. Fortunately, the need for such an enum list is exceedingly remote.

HLA Type Compatibility

HLA is unusual among assembly language insofar as it does some serious type checking on its operands. While the type checking isn't quite as "strong" as some high level languages, HLA clearly does a lot more type checking than other assemblers, even those that purport to do type checking on operands (e.g., MASM). The use of strong type checking can help you locate logical errors in your code that would otherwise go unnoticed (except via a laborious and time consuming testing/debug session).

The downside to strong type checking is that experienced assembly programmers may become somewhat annoyed with HLA's reports that they are doing something wrong when, in fact, the programmer knows exactly what they are doing. There are two solutions to this problem: use type coercion (described a little bit later) or use the "untyped" types that reduce type checking to simply ensuring that the sizes of the operands match. However, before discussing how to override HLA's type checking system, it's probably a good idea to first describe how HLA uses data types.

Fundamentally, HLA divides up the data types into classes based on the size of their underlying representation. Unless you explicitly override a type with a type coercion operation, attempting to mix object sizes in a memory or register operand will produce an error (in constant expressions, HLA is a bit more forgiving; it will automatically promote between certain types and adjust the type of the result accordingly). With most of HLA's data types, it's pretty obvious what the size of the underlying representation is, because most HLA type names incorporate the size (in bits) in the type's name. For example, the uns16 data type is a 16-bit (two-byte) type. Nevertheless, this rule isn't true for all data types, so it's a good idea to begin this discussion by looking at the underlying sizes of each of the HLA types.

8 bits: boolean, byte, char, enum, int8, uns8

16 bits: int16, uns16, wchar, word

32 bits: dword, int32, pointer types, real32, string, zstring, unicode, uns32

64 bits: int64, qword, real64, uns64

80 bits: real80, tbyte

128 bits: cset, int128, lword, uns128

The byte, word, dword, qword, tbyte, and lword types are somewhat special. These are known as untyped data types. They are directly compatible with any scalar, ordinal, data type that is the same size as the type in question. For example, a byte object is directly compatible with any object of type boolean, byte, char, enum, int8, or uns8 . No special coercion is necessary when assigning a byte value to an object that has one of these other types; likewise, no special coercion operation is necessary when assigning a value of one of these other types to a byte object.

Note that cset, real32, real64, and real80 objects are not ordinal types. Therefore, you cannot directly mix these types with lword, dword, qword, or tbyte objects without an explicit type coercion operation. Also keep in mind that composite data types (see the next section) are not directly compatible with bytes, words, dwords, qwords, tbytes, and lwords, even if the composite data type has the same number of bytes (the only exception is the pointer data type, which is compatible with the dword type).

Composite Data Types

In addition to the primitive types above, HLA supports arrays, records (structures), unions, classes, and pointers of the primitive types (except for text objects).

Array Data Types

HLA allows you to create an array data type by specifying the number of array elements after a type name. Consider the following HLA type declaration that defines intArray to be an array of int32 objects:

 

type intArray : int32[ 16 ];

The "[ 16 ]" component tells HLA that this type has 16 four-byte integers. HLA arrays use a zero-based index, so the first element is always element zero. The index of the last element, in this example, is 15 (total of 16 elements with indicies 0..15).

HLA also supports multidimensional arrays. You can specify multidimensional arrays by providing a list of indicies inside the square brackets, e.g.,

 

type intArray4x4 : int32[ 4, 4 ];
type intArray2x2x4 : int32[ 2,2,4 ];

The mechanism for accessing array elements differs depending upon whether you are accessing compile-time array constants or run-time array variables. A complete discussion of this will appear in later sections.

Union Data Types

HLA implements the discriminant union type using the UNION..ENDUNION reserved words. The following HLA type declaration demonstrates a union declaration:

 

type allInts: union
i8: int8;
i16: int16;
i32: int32;
endunion;

All fields in a union have the same starting address in memory. The size of a union object is the size of the largest field in the union. The fields of a union may have any type that is legal in a variable declaration section (see the discussion of the VAR section for more details).

Given a union object, say "i" of type "allInts", you access the fields of the union using the familiar dot-notation. The following 80x86 mov instructions demonstrate how to access each of the fields of the "i" variable:

 

mov( i.i8, al );
mov( i.i16, ax );
mov( i.i32, eax );

Unions also support a special field type known as an anonymous record (see the next section for a description of records). The syntax for an anonymous record in a union is the following:

type

 

unionWrecord: union

u1Field: byte;

u2Field: word;

u3Field: dword;

record

u4Field: byte[2];

u5Field: word[3];

endrecord;

u6Field: byte;

endunion;

 

Fields appearing within the anonymous record do not necessarily start at offset zero in the data structure. In the example above, u4Field starts at offset zero while u5Field immediately follows it two bytes later. The fields in the union outside the anonymous record all start at offset zero. If the size of the anonymous record is larger than any other field in the union, then the record's size determines the size of the union. This is true for the example above, so the union's size is 16 bytes since the anonymous record consumes 16 bytes.

Record Data Types6

HLA's records allow programmers to create data types whose fields can be different types. The following HLA type declaration defines a simple record with four fields:

 

type Planet: record

x: int32;
y: int32;
z: int32;
density: real64;

endrecord;

Objects of type Planet will consume 20 bytes of storage at run-time.

The fields of a record may be of any legal HLA data type including other composite data types. Like unions, anything that is legal in a VAR section is a legal field of a record. Also like unions, you use the dot-notation to access fields of a record object.

In addition to the VAR types, you may also declare anonymous unions within a record. An anonymous union is at union declaration without a fieldname associated with the union, e.g.,

 

type DemoAU: record
x: real32;
union
u1:int32;
r1:real32;
endunion;
y:real32;
endrecord;

In this example, x, u1, r1, and y are all fields of DemoAU. To access the fields of a variable D of type DemoAU, you would use the following names: D.x, D.u1, D.r1, and D.y. Note that D.u1 and D.r1 share the same memory locations at run-time, while D.x and D.y have unique addresses associated with them.

 

Record types may inherit fields from other record types. Consider the following two HLA type declarations:

 

type
Pt2D: record

x: int32;
y: int32;

endrecord;

Pt3D: record inherits( Pt2D )

z: int32;

endrecord;

In this example, Pt3D inherits all the fields from the Pt2D type. The "inherits" keyword tells HLA to copy all the fields from the specified record (Pt2D in this example) to the beginning of the current record declaration (Pt3D in this example). Therefore, the declaration of Pt3D above is equivalent to:

 

Pt3D: record

x: int32;
y: int32;
z: int32;

endrecord;

 

In some special situations you may want to override a field from a previous field declaration. For example, consider the following record declarations:

BaseRecord:

record

a: uns32;

b: uns32;

endrecord;

 

DerivedRecord:

record inherits( BaseRecord )

b: boolean; // New definition for b!

c: char;

endrecord;

 

Normally, HLA will report a "duplicate" symbol error when attempting to compile the declaration for "DerivedRecord" since the "b" field is already defined via the "inherits( BaseRecord )" option. However, in certain cases it's quite possible that the programmer wishes to make the original field inaccessible in the derived class by using a different name. That is, perhaps the programmer intends to actually create the following record:

DerivedRecord:

record

a: uns32; // Derived from BaseRecord

b: uns32; // Derived from BaseRecord, but inaccessible here.

b: boolean; // New definition for b!

c: char;

endrecord;

 

HLA allows a programmer explicitly override the definition of a particular field by using the OVERRIDES keyword before the field they wish to override. So while the previous declarations for DerivedRecord produce errors, the following is acceptable to HLA:

BaseRecord:

record

a: uns32;

b: uns32;

endrecord;

 

DerivedRecord:

record inherits( BaseRecord )

overrides b: boolean; // New definition for b!

c: char;

endrecord;

 

 

Normally, HLA aligns each field on the next available byte offset in a record. If you wish to align fields within a record on some other boundary, you may use the ALIGN directive to achieve this. Consider the following record declaration as an example:

type

AlignedRecord:

record

b:boolean; // Offset 0

c:char; // Offset 1

align(4);

d:dword; // Offset 4

e:byte; // Offset 8

w:word; // Offset 9

f:byte; // Offset 11

endrecord;

 

Note that variable "d" is aligned at a four-byte offset while "w" is not aligned. We can correct this problem by sticking another ALIGN directive in this record:

type

AlignedRecord2:

record

b:boolean; // Offset 0

c:char; // Offset 1

align(4);

d:dword; // Offset 4

e:byte; // Offset 8

align(2);

w:word; // Offset 10

f:byte; // Offset 12

endrecord;

 

Be aware of the fact that the ALIGN directive in a RECORD only aligns fields in memory if the record object itself is aligned on an appropriate boundary. For example, if an object of type AlignedRecord2 appears in memory at an odd address, then the "d" and "w" fields will also be misaligned (that is, they will appear at odd addresses in memory). Therefore, you must ensure appropriate alignment of any record variable whose fields you're assuming are aligned.

Note that the AlignedRecord2 type consumes 13 bytes. This means that if you create an array of AlignedRecord2 objects, every other element will be aligned on an odd address and three out of four elements will not be double-word aligned (so the "d" field will not be aligned on a four-byte boundary in memory). If you are expecting fields in a record to be aligned on a certain byte boundary, then the size of the record must be an even multiple of that alignment factor if you have arrays of the record. This means that you must pad the record with extra bytes at the end to ensure proper alignment. For the AlignedRecord2 example, we need to pad the record with three bytes so that the size is an even multiple of four bytes. This is easily achieved by using an ALIGN directive as the last declaration in the record:

type

AlignedRecord2:

record

b:boolean; // Offset 0

c:char; // Offset 1

align(4);

d:dword; // Offset 4

e:byte; // Offset 8

align(2);

w:word; // Offset 10

f:byte; // Offset 12

align(4) // Ensures we're padded to a multiple of four bytes.

endrecord;

 

Note that you should only use values that are integral powers of two in the ALIGN directive.

If you want to ensure that all fields are appropriately aligned on some boundary within the record, but you don't want to have to manually insert ALIGN directives throughout the record, HLA provides a second alignment option to solve your problem. Consider the following syntax:

type

alignedRecord3 : record[4]

<< Set of fields >>

endrecord;

 

The "[4]" immediately following the RECORD reserved word tells HLA to start all fields in the record at offsets that are multiples of four, regardless of the object's size (and the size of the objects preceeding the field). HLA allows any integer expression that produces a value in the range 1..4096 inside these parenthesis. If you specify the value one (which is the default), then all fields are packed (aligned on a byte boundary). For values greater than one, HLA will align each field of the record on the specified boundary. For arrays, HLA will align the field on a boundary that is a multiple of the array element's size. The maximum boundary HLA will round any field to is a multiple of 4096 bytes.

Note that if you set the record alignment using this syntactical form, any ALIGN directive you supply in the record may not produce the desired results. When HLA sees an ALIGN directive in a record that is using field alignment, HLA will first align the current offset to the value specified by ALIGN and then align the next field's offset to the global record align value.

Nested record declarations may specify a different alignment value than the enclosing record, e.g.,

type

alignedRecord4 : record[4]

a:byte;

b:byte;

c:record[8]

d:byte;

e:byte;

endrecord;

f:byte;

g:byte;

endrecord;

 

In this example, HLA aligns fields a, b, f, and g on dword boundaries, it aligns d and e (within c ) on eight-byte boundaries. Note that the alignment of the fields in the nested record is true only within that nested record. That is, if c turns out to be aligned on some boundary other than an eight-byte boundary, then d and e will not actually be on eight-byte boundaries; they will, however be on eight-byte boundaries relative to the start of c .

In addition to letting you specify a fixed alignment value, HLA also lets you specify a minimum and maximum alignment value for a record. The syntax for this is the following:

type

recordname : record[maximum : minimum]

<< fields >>

endrecord;

 

Whenever you specify a maximum and minimum value as above, HLA will align all fields on a boundary that is at least the minimum alignment value. However, if the object's size is greater than the minimum value but less than or equal to the maximum value, then HLA will align that particular field on a boundary that is a multiple of the object's size. If the object's size is greater than the maximum size, then HLA will align the object on a boundary that is a multiple of the maximum size. As an example, consider the following record:

type

r: record[ 4:1 ];

a:byte; // offset 0

b:word; // offset 2

c:byte; // offset 4

d:dword;[2] // offset 8

e:byte; // offset 16

f:byte; // offset 17

g:qword; // offset 20

endrecord;

 

Note that HLA aligns g on a dword boundary (not qword, which would be offset 24) since the maximum alignment size is four. Note that since the minimum size is one, HLA allows the f field to be aligned on an odd boundary (since it's a byte).

If an array, record, or union field appears within a record, then HLA uses the size of an array element or the largest field of the record or union to determine the alignment size. That is, HLA will align the field without the outermost record on a boundary that is compatible with the size of the largest element of the nested array, union, or record.

HLA sophisticated record alignment facilities let you specify record field alignments that match that used by most major high level language compilers. This lets you easily access data types used in those HLLs without resorting to inserting lots of ALIGN directives inside the record.

Note that there is a big difference in the semantics between the global record alignment option (above) and the similar syntax in the STATIC, READONLY , and STORAGE declaration sections. (which is why their syntax is different) Consider the following:

static(4)

v1: byte;

v2: dword;

 

Unlike the record alignment option, this example only aligns the first field of the STATIC section, not all the variables in that section (i.e., v2 will not be aligned on a dword boundary in the example above). Keep this difference in mind when using this alignment option.

When declaring record variables in a VAR, STATIC, READONLY, STORAGE , or SEGMENT declaration section, HLA associates the offset zero with the first field of a record. Each additional field in the record is assigned an offset corresponding to the sum of the sizes of all the prior fields. So in the example immediately above, "x" would have the offset zero, "y" would have the offset four, and "z" would have the offset eight.

If you would like to specify a different starting offset, you can use the following syntax for a record declaration:

 

Pt3D: record := 4;

x: int32;
y: int32;
z: int32;

endrecord;

The constant expression specified after the assignment operator (":=") specifies the starting offset of the first field in the record. In this example x, y, and z will have the offsets 4, 8, and 12, respectively.

Warning: setting the starting offset in this manner does not add padding bytes to the record. This record is still a 12-byte object. If you declare variables using a record declared in this fashion, you may run into problems because the field offsets do not match the actual offsets in memory. This option is intended primarily for mapping records to pre-existing data structures in memory. Only really advanced assembly language programmers should use this option.

Pointer Types

HLA allows you to declare a pointer to some other type using syntax like the following:

 

pointer to base_type

 

The following example demonstrates how to create a pointer to a 32-bit integer within the type declaration section:

 

type pi32: pointer to int32;

 

HLA pointers are always 32-bit (near32) pointers.

HLA also allows you to define pointers to existing procedures using syntax like the following:

procedure someProc( parameter_list );

<< procedure options, followed by @external, @forward, or procedure body>>

.

.

.

type

p : pointer to procedure someProc;

 

The p procedure pointer "inherits" all the parameters and other procedure options associated with the original procedure. This is really just shorthand for the following:

procedure someProc( parameter_list );

<< procedure options, followed by @external, @forward, or procedure body>>

.

.

.

type

p : procedure ( Same_Parameters_as_someProc ); <<same options as someProc>>

 

The former version, however, is easier to maintain since you don't have to keep the parameter lists and procedure options in sync.

Note that HLA provides the reserved word null (or NULL, reserved words are case insensitive) to represent the nil pointer. HLA replaces NULL with the value zero. The NULL pointer is compatible with any pointer type (including strings, which are pointers).

Thunks

A "thunk" is an eight-byte variable that contains a pointer to a piece of code to execute and an execution environment pointer (i.e., a pointer to an activation record). The code associated with a thunk is, essentially, a small procedure that (generally) uses the activation record of the surround code rather than creating its own activation record. HLA uses thunks to implement the iterator "yield" statement as well as pass by name and pass by lazy evaluation parameters. In addition to these two uses of thunks, HLA allows you to declare your own thunk objects and use them for any purpose you desire. To declare a thunk variable is easy, just use a declaration like the following in a VAR or STATIC section:

 

thunkVar: thunk;

 

This declaration reserves eight bytes of storage. The first dword holds the address of the code to execute, the second dword holds a pointer to the activation record to load into EBP when the thunk executes.

Of course, like almost any pointer variable, declaring a thunk variable is the easy part; the hard part is making sure the thunk variable is initialized before attempting to call the thunk. While you could manually load the address of some code and the frame pointer value into a thunk variable, HLA provides a better syntax for initializing thunks with small code fragments: the "thunk" statement. The "thunk" statement uses the following syntax:

 

thunk thunkVar := #{ sequence_of_statements }#;

 

Consider the following example:

 

program ThunkDemo;

#include( "stdio.hhf" );

 

procedure proc1;

var

i: int32;

p1Thunk: thunk;

procedure proc2( t:thunk );

var

i:int32;

begin proc2;

mov( 25, i );

t();

stdout.put( "Inside proc2, i=", i, nl );

end proc2;

begin proc1;

 

thunk p1Thunk := #{ mov( 0, i ); }#;

mov( 1, i );

proc2( p1Thunk );

stdout.put( "i=", i, nl );

 

end proc1;

begin ThunkDemo;

 

proc1();

end ThunkDemo;

 

In this example, proc1 has two local variables, i and p1Thunk . The THUNK statement initializes the p1Thunk variable with the address of some code that moves a zero into the i variable. The THUNK statement also initializes p1Thunk with a pointer to the current activation record (that is, a pointer to proc1 's activation record). Then proc1 calls proc2 passing p1Thunk as a parameter.

The proc2 routine has its own local variable named i . Of course, this is a different variable than the i in proc1 . Proc2 begins by setting its variable i to the value 25. Then proc2 invokes the thunk (passed to it as a parameter). This thunk sets the variable i to zero. However, since the thunk uses the current activation record when the set statement was executed, this statement sets proc1 's i variable to zero rather than proc2 's i variable. This program produces the following output:

 

Inside proc2, i=25

i=0

 

Although you probably won't use thunks that often, they are quite nice for deferred execution. This is especially useful in AI (Artificial Intelligence) programs.

Class Types

Classes and object-oriented programming are the subject of a later section of this document. See Class Data Types for more details.

Regular Expression Types

The HLA compile-time language supports a special data type known as a "compiled regular expression". Please see the section on regular expression macros for more details on this data type.

Numeric Constants

HLA lets you specify several different types of numeric constants.

Decimal Constants

The first and last characters of a decimal integer constant must be decimal digits (0..9). Interior positions may contain decimal digits and underscores. The purpose of the underscore is to provide a better presentation for large decimal values (i.e., use the underscore in place of a comma in large values). Example: 1_234_265.

Note: Technically, HLA does not allow negative literal integer constants. However, you can use the unary "-" operator to negate a value, so you'd never notice this omission (e.g., -123 is legal, it consists of the unary negation operator followed by a positive decimal literal constant). Therefore, HLA always returns type unsXX for all literal decimal constants. Also note that HLA always uses a minimum size of uns32 for literal decimal constants. If you absolutely, positively, want a literal constant to be treated as some other type, use one of the compile-time type coercion functions to change the type (e.g., uns8(1), word(2), or int16(3)). Generally, the type that HLA uses for the object is irrelevant since HLA will automatically promote a value to a larger or smaller type as appropriate.

Here are the following ranges for the various HLA unsigned data types:

uns8: 0..255

uns16: 0..65,535

uns32: 0..4,294,967,295

uns64: 0..18,446,744,073,709,551,615

uns128: 0..340,282,366,920,938,463,463,374,607,431,768,211,455

Hexadecimal Constants

Hexadecimal literal constants must begin with a dollar sign ("$") followed by a hexadecimal digit and must end with a hexadecimal digit (0..9, A..F, or a..f). Interior positions may contain hexadecimal digits or underscores. Hexadecimal constants are easiest to read if each group of four digits (starting from the least significant digit) is separated from the others by an underscore. E.g., $1A_2F34_5438.

If the constant fits into 32 bits or less, HLA always returns the dword type for a hexadecimal constant. For larger values, HLA will automatically use the qword or lword type, as appropriate. If you would like the hexadecimal value to have a different type, use one of the HLA compile-time type coercion functions to change the type (e.g., byte($12) or qword($54)).

Here are the following ranges for the various HLA hexadecimal data types:

uns8: 0..$FF

uns16: 0..$FFFF

uns32: 0..$FFFF_FFFF

uns64: 0..$FFFF_FFFF_FFFF_FFFF

uns128: 0..$FFFF_FFFF_FFFF_FFFF_FFFF_FFFF_FFFF_FFFF

Binary Constants

Binary literal constants begin with a percent sign ("%") followed by at least one binary digit (0/1) and they must end with a binary digit. Interior positions may contain binary digits or underscore characters. Binary constants are easiest to read if each group of four digits (starting from the least significant digit) is separated from the others by an underscore. E.g., %10_1111_1010.

Like hexadecimal constants, HLA always associates the type dword with a "raw" binary constant; it will use the qword or lword type if the value is greater than 32 bits or 64 bits (respectively). If you want HLA to use a different type, use one of the compile-time type coercion functions to achieve this.

Obviously, binary constants may have as many binary digits as there are bits in the underlying type. This document will not attempt to write out the maximum binary literal constant!

Numeric Set Constants

HLA provides a special numeric constant form that lets you specify a numeric value by the bit positions containing ones. This corresponds to a powerset of integer values in the range 0..31. These constants take the following form:

@{ comma_separated_list_of_digits }

 

The comma_separate_list_of_digits can be empty (signifying no set bits, i.e., the value zero), a single digit, or a set of digits separated by commas. Here are some examples:

@{}

@{8}

@{1,2,8,24}

 

The corresponding numeric constant is given the type dword and is assigned the value that has ones in all the specified bit positions. For example, "@{8}" is equal to 256 since this value has a single set bit in bit position eight. Note that "@{0}" equals one, not zero (because the value one has a single set bit in position zero).

Real (Floating Point) Constants

Floating point (real) literal constants always begin with a decimal digit (never just a decimal point). A string of one or more decimal digits may be optionally followed by a decimal point and zero or more decimal digits (the fractional part). After the optional fractional part, a floating point number may be followed by "e" or "E", a sign ("+" or "-"), and a string of one or more decimal digits (the exponent part). Underscores may appear between two adjacent digits in the floating point number; their presence is intended to substitute for commas found in real-world decimal numbers.

Examples:

1.2

2.345e-2

0.5

1.2e4

2.3e+5

1_234_567.0

 

Literal real constants are always 80 bits and have the default type real80 . If you wish to specify real32 or real64 literal constants, then use the real32 or real64 compile-time coercion functions to convert the values, e.g., real32( 3.14159 ) . During compile time, it's rare that you'd want to use one of the smaller types since they are less accurate at representing floating point values (although this might be precisely why you decide to use the smaller real type, so the accuracy matches the computations you're doing at run-time).

The range of real32 constants is approximately 10±38 with 6-1/2 digits of precision; the range of real64 values is approximately 10±308 with approximately 14-1/2 digits of precision, and the range of real80 constants is approximately 10±4096 with about 18 digits of precision.

Boolean Constants

Boolean constants consist of the two predefined identifiers true and false. Note that your program may redefine these identifiers, but doing so is incredibly bad programming style. Since these are actual identifiers in the symbol table (and not reserved words), you must spell these identifiers in all lower case or HLA will complain (unlike reserved words that are case insensitive).

Internally, HLA represents true with one and false with zero. In fact, HLA's boolean operations only look at bit #0 of the boolean value (and always clear the other bits). HLA compile-time statements that expect a boolean expression do not use zero/not zero like C/C++ and a few other languages. Such expressions must have a boolean type with the values true/false; you cannot supply an integer expression and rely on zero/not zero evaluation as in C/C++ or BASIC.

Character Constants

Character literals generally consist of a single (graphic) character surrounded by apostrophes. To represent the apostrophe character, use four apostrophies, e.g., `'''.

Another way to specify a character constant is by typing the "#" symbol followed by a numeric literal constant (decimal, hexadecimal, or binary). Examples: #13, #$D, #%1101.

Unicode Character Constants

Unicode character constants are 16-bit values. HLA provides limited support for Unicode literal constants. HLA supports the UTF/7 code point (character set) which is just the standard seven-bit ASCII character set and nine high-order zero bits. To specify a 16-bit literal Unicode constant simply prefix a standard ASCII literal constant with a 'u' or 'U', e.g.,

u'A' - UTF/7 character constant for 'A'

Note that UTF/7 constants are simply the ASCII character codes zero extended to 16 bits.

HLA provides a second syntax for Unicode character constants that lets you enter values whose character codes are outside the range $20..$7E. You can specify a Unicode character constant by its numeric value using the 'u#nnnn' constant form. This form lets you specify a 16-bit value following the '#' in either binary, decimal, or hexadecimal form, e.g.,

u#1233

u#$60F

u%100100101001

String Constants

String literal constants consist of a sequence of (graphic) characters surrounded by quotes. To embed a quote within a string, insert a pair of quotes into the string, e.g., "He said ""This"" to me."

If two string literal constants are adjacent in a source file (with nothing but whitespace between them), then HLA will concatenate the two strings and present them to the parser as a single string. Furthermore, if a character constant is adjacent to a string, HLA will concatenate the character and string to form a single string object. This is useful, for example, when you need to embed control characters into a string, e.g.,

 

"This is the first line" #$d #$a "This is the second line" #$d #$a

 

HLA treats the above as a single string with a Wndows newline sequence (CR/LF) at the end of each of the two lines of text.

Unicode String Constants

HLA lets you specify Unicode string literals by prefixing a standard string constant with a 'u' or a 'U'. Such string constants use the UTF/7 character set (that is, the standard ASCII character set) but reserve 16 bits for each character in the string. Note that the high order nine bits of each character in the string will contain zero.

As this was being written, there is no support for Unicode strings in the HLA Standard Library, though support for Unicode string functions should appear shortly (note that Windows' programmers can call the Unicode string functions that are part of the Windows' API).

Character Set Constants

A character set literal constant consists of several comma delimited character set expressions within a pair of braces. The character set expressions can either be individual character values or a pair of character values separated by an ellipse (".."). If an individual character expression appears within the character set, then that character is a member of the set; if a pair of character expressions, separated by an ellipse, appears within a character set literal, then all characters between the first such expression and the second expression are members of the set. As a convenience, if a string constant appears between the braces, HLA will take the union of all the characters in that string and add those character to the character set.

Examples:

{`a','b','c'} // a, b, and c.

{`a'..'c'} // a, b, and c.

{`A'..'Z','a'..'z'} //Alphabetic characters.

{"cset"} // The character set `c', `e', `s', and `t'.

{` `,#$d,#$a,#$9} //Whitespace (space, return, linefeed, tab).

 

HLA character sets are currently limited to holding characters from the 128-character ASCII character set. In the future, HLA may support an xcset type (supporting 256 elements) or even wcset (wide character sets), but that support does not currently exist.

Structured Constants

Array Constants

Note: see Array Data Types for more details about HLA array types.

HLA lets you specify an array literal constant by enclosing a set of values within a pair of square brackets. Since array elements must be homogenous, all elements in an array literal constant must be the same type or conformable to the same type. Examples:

 

[ 1, 2, 3, 4, 9, 17 ]
[ 'a', 'A', 'b', 'B' ]
[ "hello", "world" ]

Note that each item in the list of values can actually be a constant expression, not a simple literal value.

HLA array constants are always one dimensional. This, however, is not a limitation because if you attempt to use array constants in a constant expression, the only thing that HLA checks is the total number of elements. Therefore, an array constant with eight integers can be assigned to any of the following arrays:

 

const
a8: int32[8] := [1,2,3,4,5,6,7,8];
a2x4: int32[2,4] := [1,2,3,4,5,6,7,8];
a2x2x2: int32[2,2,2] := [1,2,3,4,5,6,7,8];

Although HLA doesn't support the notation of a multi-dimensional array constant, HLA does allow you to include an array constant as one of the elements in an array constant. If an array constant appears as a list item within some other array constant, then HLA expands the interior constant in place, lengthening the list of items in the enclosing list. E.g., the following three array constants are equivalent:

 

[ [1,2,3,4], [5,6,7,8] ]
[ [ [1,2], [3,4] ], [ [5,6], [7,8] ] ]
[1,2,3,4,5,6,7,8]

Although the three array constants are identical, as far as HLA is concerned, you might want to use these three different forms to suggest the shape of the array in an actual declaration, e.g.,

 

const
a8: int32[8] := [1,2,3,4,5,6,7,8];
a2x4: int32[2,4] := [ [1,2,3,4], [5,6,7,8] ];
a2x2x2: int32[2,2,2] := [[[1,2], [3,4] ], [[5,6], [7,8]]];

Also note that symbol array constants, not just literal array constants, may appear in a literal array constant. For example, the following literal array constant creates a nine-element array holding the values one through nine at indexes zero through eight:

 

const Nine: int32[ 9 ] := [ a8, 9 ];

 

This assumes, of course, that "a8" was previously declared as above. Since HLA "flattens" all array constants, you could have substituted a2x4 or ax2x2x for a8 in the example above and obtained identical results.

As a convenience to those building array constants using the HLA compile-time language, an HLA array constant will allow an extra comma at the end of the list of array elements, e.g.,

const
a8: int32[8] := [1,2,3,4,5,6,7,8, ]; // Note extra comma after '8'

Note that this does not create an "empty" element in the array. The array (in this example) still has eight elements. Allowing the extra comma at the end of the list allows you to generate the list programmatically (using the HLA compile-time language) without requiring a special case for the last item in the list (which would normally need to be handled specially because there is no comma after the last item when using "clean" syntax). For example, consider the following definition of a8:

 

a8: int32[8] :=

[

#for( i := 1 to 8 )

i,

#endfor

];

 

This array definition is exactly equivalent to the previous one (including the extra comma at the end). Prior to the addition of this feature in HLA, you'd have to use a kludge like the following to handle the last element:

 

a8: int32[8] :=

[

#for( i := 1 to 8 )

i,

#endfor

8 // Handle last element specially

];

 

Though allowing an extra comma at the end of the list is aesthetically unappealing, kludges like the #for loop immediately above is an even worse offense.

 

You may also create an array constant using the HLA DUP operator. This operator uses the following syntax:

expression DUP [expression_to_replicate]

Where expression is an integer expression and expression_to_replicate is a some expression, possibly an array constant. HLA generates an array constant by repeating the values in the expression_to_replicate the number of times specified by the expression. (Note: this does not create an array with expression elements unless the expression_to_replicate contains only a single value; it creates an array whose element count is expression times the number of items in the expression_to_replicate ). Examples:

 

10 dup [1] -- equivalent to [1,1,1,1,1,1,1,1,1,1]

5 dup [1,2] -- equivalent to [1,2,1,2,1,2,1,2,1,2]

 

Please note that HLA does not allow class constants, so class objects may not appear in array constants. Also, HLA does not allow generic pointer constants, only certain types of pointer constants are legal. See the discussion of pointer constants for more details.

Record Constants

Note: see Record Data Types for details about HLA Records.

HLA supports record constants using a syntax very similar to array constants. You enclose a comma-separated list of values for each field in a pair of square brackets. To further differentiate array and record constants, the name of the record type and a colon must precede the opening square bracket, e.g.,

 

Planet:[ 1, 12, 34, 1.96 ]

 

HLA associates the items in the list with the fields as they appear in the original record declaration. In this example, the values 1, 12, 34, and 1.96 are associated with fields x, y, z, and density, respectively. Of course, the types of the individual constants must match (or be conformable to) the types of the individual fields.

Note that you may not create a record constant for a particular record type if that record includes data types that cannot have compile-time constants associated with them. For example, if a field of a record is a class object, you cannot create a record constant for that type since you cannot create class constants.

Union Constants

Note: see Union Data Types for more details about HLA's UNION types.

Union constants allow you to initialize static union data structures in memory as well as initialize union fields of other data structures (including anonymous union fields in records). There are some important differences between HLA compile-time union constants and HLA run-time unions (as well as between the HLA run-time union constants and unions in other languages). Therefore, it's a good idea to begin the discussion of HLA's union constants with a description of these differences.

There are a couple of different reasons people use unions in a program. The original reason was to share a sequence of memory locations between various fields whose access is mutually exclusive. When using a union in this manner, one never reads the data from a field unless they've previous written data to that field and there are no intervening writes to other fields between that previous write and the current read. The HLA comile-time language fully (and only) supports this use of union objects.

A second reason people use unions (especially in high level languages) is to provide aliases to a given memory location; particularly, aliases whose data types are different. In this mode, a programmer might write a value to one field and then read that data using a different field (in order to access that data's bit representation as a different type). HLA does not support this type of access to union constants. The reason is quite simple: internally, HLA uses a special "variant" data type to represent all possible constant types. Whenever you create a union constant, HLA lets you provide a value for a single data field. From that point forward, HLA effectively treats the union constant as a scalar object whose type is the same as the field you've initialized; access to the other fields through the union constant is no longer possible. Therefore, you cannot use HLA compile-time constants to do type coercion; nor is there any need to since HLA provides a set of type coercion operators like @byte, @word, @dword, @int8, etc. As noted above, the main purpose for providing HLA union constants is to allow you to initialize static union variables; since you can only store one value into a memory location at a time, union constants only need to be able to represent a single field of the union at one time (of course, at run-time you may access any field of the static union object you've created; but at compile-time you may only access the single field associated with a union constant).

An HLA literal union constant takes the following form:

 

typename.fieldname:[ constant_expression ]

 

The typename component above must be the name of a previously declared HLA union data type (i.e., a union type you've created in the type section). The fieldname component must be the name of a field within that union type. The constant_expression component must be a constant value (expression) whose type is the same as, or is automatically coercable to, the type of the fieldname field. Here is a complete example:

 

type

u:union

b:byte;

w:word;

d:dword;

q:qword;

endunion;

 

static

uVar :u := u.w:[$1234];

 

The declaration for uVar initializes the first two bytes of this object in memory with the value $1234. Note that uVar is actually eight bytes long; HLA automatically zeros any unused bytes when initializing a static memory object with a union constant.

Note that you may place a literal union constant in records, arrays, and other composite data structures. The following is a simple example of a record constant that has a union as one of its fields:

 

type

r :record

b:byte;

uf:u;

d:dword;

endrecord;

 

static

sr :r := r:[0, u.d:[$1234_5678], 12345];

 

In this example, HLA initializes the sr variable with the byte value zero, followed by a dword containing $1234_5678 and a dword containing zero (to pad out the remainder of the union field), followed by a dword containing 12345.

You can also create records that have anonymous unions in them and then initialize a record object with a literal constant. Consider the following type declaration with an anonymous union:

 

type

rau :record

b:byte;

union

c:char;

d:dword;

endunion;

w:word;

endrecord;

 

Since anonymous unions within a record do not have a type associated with them, you cannot use the standard literal union constant syntax to initialize the anonymous union field (that syntax requires a type name). Instead, HLA offers you two choices when creating a literal record constant with an anonymous union field. The first alternative is to use the reserved word union in place of a typename when creating a literal union constant, e.g.,

 

static

srau :rau := rau:[ 1, union.d:[$12345], $5678 ];

 

The second alternative is a shortcut notation. HLA allows you to simply specify a value that is compatible with the first field of the anonymous union and HLA will assign that value to the first field and ignore any other fields in the union, e.g.,

 

static

srau2 :rau := rau:[ 1, 'c', $5678 ];

 

This is slightly dangerous since HLA relaxes type checking a bit here, but when creating tables of record constants, this is very convenient if you generally provide values for only a single field of the anonymous union; just make sure that the commonly used field appears first and you're in business.

Although HLA allows anonymous records within a union, there was no syntactically acceptable way to differentiate anonymous record fields from other fields in the union; therefore, HLA does not allow you to create union constants if the union type contains an anonymous record. The easy workaround is to create a named record field and specify the name of the record field when creating a union constant, e.g.,

 

type

r :record

c:char;

d:dword;

endrecord;

 

u :union

b:byte;

x:r;

w:word;

endunion;

 

static

y :u := u.x:[ r:[ 'a', 5]];

 

You may declare a union constant and then assign data to the specific fields as you would a record constant. The following example provides some samples of this:

 

type

u_t :union

b:byte;

x:r;

w:word;

endunion;

 

val

u :u_t;

.

.

.

?u.b := 0;

.

.

.

?u.w := $1234;

 

The two assigments above are roughly equivalent to the following:

 

?u := u_t.b:[0];

 

and

 

?u := u_t.w:[$1234];

 

However, to use the straight assignment (the former example) you must first declare the value u as a u_t union.

To access a value of a union constant, you use the familiar "dot notation" from records and other languages, e.g.,

 

?x := u.b;

.

.

.

?y := u.w & $FF00;

 

Note, however, that you may only access the last field of the union into which you've stored some value. If you store data into one field and attempt to read the data from some other field of the union, HLA will report an error. Remember, you don't use union constants as a sneaky way to coerce one value's type to another (use the coercion functions for that purpose).

Pointer Constants

Note: see Pointer Types for more details about HLA pointer types.

HLA allows a very limited form of a pointer constant. If you place an ampersand ("&") in front of a static object's name (i.e., the name of a static variable, readonly variable, uninitialized variable, segment variable, procedure, method, or iterator), HLA will compute the run-time offset of that variable. Pointer constants may not be used in abitrary constant expressions. You may only use pointer constants in expressions used to initialize static or readonly variables or as constant expressions in 80x86 instructions. The following example demonstrates how pointer constants can be used:

 

program pointerConstDemo;

static
t:int32;
pt: pointer to int32 := &t;

begin pointerConstDemo;

mov( &t, eax );

end pointerConstDemo;

Also note that HLA allows the use of the reserved word NULL anywhere a pointer constant is legal. HLA substitutes the value zero for NULL.

Note that you may obtain the address of the current location counter as a pointer constant by applying the "&" operator to the @here keyword, e.g.,

mov( &@here, eax );

This places the address of the start of the instruction into EAX.

Regular Expression Constants

HLA uses compile-time "regex"-typed variables to hold compiled versions of regular expression. There is no literal form of a regular expression constant. The only way to generate a regular expression constant is in a VAL, CONST, or "?" declaration by assigning the "value" of a #regex macro declaration to a symbol, e.g.,

#regex someRegexMacro;

<<regex macro body>>

 

#endregex

 

const

compiledRegex :regex := someRegexMacro;

 

See the section on regular expressions for more details.

Type Checking and Type Promotion

Many dyadic (two-operand) operators expect the types of their operands to be the same. Prior to actually performing such an operation, HLA evaluates the types of the operands and attempts to make them compatible. HLA uses a type algebra to determine if two (different) types are compatible; if they are not, HLA will report a type mismatch error during assembly. If the types are compatible, HLA will make them identical via type promotion. The type algebra describes how HLA promotes one type to another in order to make the two types compatible.

Usually, you can state a type algebra easily enough by providing "algebraic" type equations. For example, in high level languages one could use a statement like "r = r + i" to suggest that the type of the resulting sum is real when the left operand is real and the right operand is integer (around the "+" operator). Unfortunately, HLA supports so many different data types and operators that any attempt to describe the type algebra in this fashion will produce so many equations that it would be difficult for the reader to absorb it all. Therefore, this document will rely on an informal English description of the type algebra to explain how HLA operates.

First of all, if two operands have the same basic type, but are different sizes, HLA promotes the smaller object to the same size as the larger object. Basic types include the following sets: {uns8, uns16, uns32, uns64, uns128}, {int8, int16, int32, int64, int128}, {byte, word, dword, qword, lword}, and {real32, real64, real80}7. So if any two operands appear from one of these sets, then both operands are promoted to the larger of the two types.

If an unsigned and a signed operand appear around an operator, HLA produces a signed result. If the unsigned operand is smaller than the signed operand, HLA assigns both operands the signed type prior to the operation. If the unsigned and signed operands are the same size (or the unsigned operand is larger), HLA will first check the H.O. bit of the unsigned operand. If it is set, then HLA promotes the unsigned operand to the next larger signed type (e.g., uns16 becomes int32 ). If the resulting signed type is larger than the other operand's type, it gets promoted as well. This scheme fails if you've got an uns128 value whose H.O. bit is set. In that case, HLA promotes both operands to int128 and will produce incorrect results (since the uns128 value just went negative when it's really positive). Therefore, you should attempt to limit unsigned values to 127 bits if you're going to be mixing signed and unsigned operations in the same expression.

Any mixture of hexadecimal types (byte, word, dword, qword, or lword) and an unsigned type produces an unsigned type; the size of the resulting unsigned type will be the larger of the two types. Likewise, any mixture of hexadecimal types and signed integer types will produce a signed integer whose size is the larger of the two types. This "strengthening" of the type (hexadecimal types are "weaker" than signed or unsigned types) may seem counter-intuitive to a die-hard assembly programmer; however, making the result type hexadecimal rather than signed/unsigned can create problems if the result has the H.O. bit set since information about whether the result is signed or unsigned would be lost at that point.

Mixing unsigned values and a real32 value will produce a real32 result or an error. HLA produces an error if the unsigned value requires more than 24 bits to represent exactly (which is the largest unsigned value you may represent within the real32 format). Note that in addition to promoting the unsigned type to real32 , HLA will also convert the unsigned value to a real32 value (promoting the type is not the same thing as converting the value; e.g., promoting uns8 to uns16 simply involves changing the type designation of the uns8 object, HLA doesn't have to deal with the actual value at all since it keeps all values in an internal 128 bit format; however, the binary representation for unsigned and real32 values is completely different, so HLA must do the value conversion as well). Note that if you really want to convert a value that requires more than 24 bits of precision to a real32 object (with truncation), just convert the unsigned operand to real64 or real80 and then convert the larger operand to real32 using the real32(expr) compile-time function. Since unsigned values are, well, unsigned and real32 objects are signed, the conversion process always produces a non-negative value.

Mixing signed and real32 values in an expression produces a real32 result. Like unsigned operands, signed operands are limited to 24 bits of precision or HLA will report an error. Technically, you should get one more bit of precision from signed operands (since the real32 format maintains its sign apart from the mantissa), but HLA still limits you to 24 bits during this conversion. If the signed integer value is negative, so will be the real32 result.

If you mix hexadecimal and real32 types, HLA treats the hexadecimal type as an unsigned value of the same size. See the discussion of unsigned and real32 values earlier for the details.

If you mix an unsigned, signed, or hexadecimal type with a real64 type, the result is an error (if HLA cannot exactly represent the value in real64 format) or a real64 result. The conversion is very similar to the real32 conversion discussed above except you get 52 bits of integer precision rather than only 24 bits.

If you mix an unsigned, signed, or hexadecimal type with a real80 type, the result is an error (if HLA cannot exactly represent the value in real80 format) or a real80 result. The conversion is very similar to the real32 conversion discussed above except you get 64 bits of integer precision rather than only 24 bits. Note that conversion of integer objects 64-bits or less will always proceed without error; 128-bit values are the only ones that will get you into trouble.

If you mix a pair of different sized real values in the same expression, HLA will promote (and convert) the smaller real value to the same size as the larger real value.

The only non-numeric promotions that take place in an expression are between characters and strings. If a character and a string both appear in an expression, HLA will promote the character to a string of length one8.

Type Coercion in HLA

HLA will report a type mismatch error if objects of incompatible types appear within an expression. Note that you may use the type-coercion compile-time functions to convert between types that HLA does not automatically support in an expression (see the discussion later in this document). You can also use the HLA type coercion operator to attach a specific type to a constant expression. The type coercion operator takes the following form:

(type typename constexpr)

The typename component must be a valid, declared type identifier (including any of the built-in types or appropriate user-defined types). The constexpr component can be any constant expression that is reasonably compatible with the specified type. Reasonably compatible means that the types are the same size or one of the primitive types. Examples:

 

(type int8 `a')

(type real32 constExpression+2)

(type boolean int8Val)

 

One important thing to remember is that type coercion is a bitwise operation. No conversion is done when coercing one type to another using this type coercion operation.

!expr

The expression must be either boolean or a number. For boolean values, not computes the standard logical not operation. Numerically, HLA inverts only the L.O. bit of boolean values and clears the remaining bits of the boolean value. Therefore, the result is always zero or one when NOTting a boolean value (even if the boolean object errantly contained other set bits prior to the "!" operation). Remember, the "!" operator only looks at the L.O. bit; if the value was originally non-zero but the L.O. bit was clear9, then "!" produces true. This is not a zero/not-zero operation.

For numbers, not computes the bitwise not operation on the bits of the number, that is, it inverts all the bits. The exact semantics of this operation depend upon the original data type of the value you're inverting. Therefore, the result of applying the "!" operator to an integer number may not always be intuitive because HLA always maintains 128-bits of precision, regardless of the underlying data type. Therefore, a full explanation of this operator's semantics must be given on a type-by-type basis.

uns8 : Bits 8..127 of an Uns8 object are always zero. Therefore, when you apply the "!" operator to an Uns8 value, the result can no longer be an Uns8 object since bits 8..127 will now contain ones. Zeroing out the H.O. bits is not wise, because you could be assigning the result of this expression to a larger data type and you may very well expect those bits to be set. Therefore, HLA converts the type of "!u8expr" to type byte (which does allow the H.O. bits to contain non-zero values). If you assign an object of type byte to a larger object (e.g., type word ), HLA will copy over the H.O. bits from the byte object to the larger object. Example:

val

u8 :uns8 := 1;

b8 := !u8; // produces $FFF..FFFE but registers as byte $FE.

w16 :word := b8; // produces $FF..FFFE but registers as word $FFFE.

 

Note: If you really want to chop the value off at eight bits, you can use the compile-time byte function to achieve this, e.g.,

val

u8 :uns8 := 1;

b8 := byte(!u8); // produces $FE.

w16 :word := b8; // produces $00FE.

 

 

uns16 : The semantics are similar to uns8 except, of course, applying "!" to an uns16 value produces a word value rather than a byte value. Again, the "!" operator will set bits 16..127 to one in the final result. If you want to ensure that the final result contains no set bits beyond bit #15, use the compile-time word function to strip the value down to 16 bits (just like the byte function in the example above).

uns32 : The semantics are similar to uns8 except, of course, applying "!" to an uns32 value produces a dword value rather than a byte value. Again, the "!" operator will set bits 32..127 to one in the final result. If you want to ensure that the final result contains no set bits beyond bit #31 use the compile-time dword function to strip the value down to 32 bits (just like the byte function in the example above).

uns64 : The semantics are similar to uns8 except, of course, applying "!" to an uns64 value produces a qword value rather than a byte value. Again, the "!" operator will set bits 64..127 to one in the final result. If you want to ensure that the final result contains no set bits beyond bit #63 use the compile-time qword function to strip the value down to 64 bits (just like the byte function in the example above).

uns128 : Applying the "!" operator to an uns128 object simply inverts all the bits. There are no funny semantics here. Resulting expression type is set to lword .

int8 : Same semantics as byte (see explanation below).

int16 : Same semantics as word (see explanation below).

int32 : Same semantics as dword (see explanation below).

int64 : Same semantics as qword (see explanation below).

int128 : Applying the "!" operator to an int128 object simply inverts all the bits. There are no funny semantics here. Resulting expression type is set to lword .

byte : Bits 8..127 of a byte ( int8 ) value must all be zeros or all ones. The "!" operator enforces this. If any of the H.O. bits are non-zero, the "!" operator sets them all to zero in the result; if all of the H.O. bits are zero, the "!" operator sets the H.O. bits to ones in the result. Of course, this operator inverts bits 0..7 in the original value and returns this inverted result. Note that the type of the new value is always byte (even if the original subexpression was int8 ).

word : Bits 16..127 of a word ( int16 ) value must all be zeros or all ones. The "!" operator enforces this. If any of the H.O. bits are non-zero, the "!" operator sets them all to zero in the result; if all of the H.O. bits are zero, the "!" operator sets the H.O. bits to ones in the result. Of course, this operator inverts bits 0..15 in the original value and returns this inverted result. Note that the type of the new value is always word (even if the original subexpression was int16 ).

dword : Bits 32..127 of a d word ( int32 ) value must all be zeros or all ones. The "!" operator enforces this. If any of the H.O. bits are non-zero, the "!" operator sets them all to zero in the result; if all of the H.O. bits are zero, the "!" operator sets the H.O. bits to ones in the result. Of course, this operator inverts bits 0..31 in the original value and returns this inverted result. Note that the type of the new value is always d word (even if the original subexpression was int32 ).

qword : Bits 64..127 of a q word ( int64 ) value must all be zeros or all ones. The "!" operator enforces this. If any of the H.O. bits are non-zero, the "!" operator sets them all to zero in the result; if all of the H.O. bits are zero, the "!" operator sets the H.O. bits to ones in the result. Of course, this operator inverts bits 0..63 in the original value and returns this inverted result. Note that the type of the new value is always q word (even if the original subexpression was int64 ).

lword : Applying the "!" operator to an lword object simply inverts all the bits. There are no funny semantics here..

No other types are legal with the "!" operator. HLA will report a type conflict error if you attempt to apply this operator to some other type.

If the operand is one of the integer types (signed, unsigned, hexadecimal), then HLA will set the type of the result to the smallest type within that class (signed, unsigned, or hexadecimal) that can hold the result (not including sign extension bits for negative numbers or zero extension bits for non-negative values).

- expr (unary negation operator)

The expression must either be a numeric value or a character set. For numeric values, "-" negates the value. For character sets, the "-" operator computes the complement of the character set (that is, it returns all the characters not found in the set).

Again, the exact semantics depend upon the type of the expression you're negating. The following paragraphs explain exactly what this operator does to its expression. For all integer values (unsXX, intXX, byte, word, dword, qword, and lword), the negation operator always does a full 128-bit negation of the supplied operand. The difference between these different data types is how HLA sets the resulting type of the expressions (as explained in the paragraphs below).

uns8 : If the original value was in the range 128..255, then the resulting type is int16 , otherwise the resulting type is int8 . Since uns8 values are always positive, the negated result is always negative, hence the result type is always a signed integer type.

uns16 : If the original value was in the range 32678..65535, then the resulting type is int32 , otherwise the resulting type is int16 . Since uns16 values are always positive, the negated result is always negative, hence the result type is always a signed integer type.

uns32 : If the original value was in the range $8000_0000..$FFFF_FFFF, then the resulting type is int64 , otherwise the resulting type is int32 . Since uns32 values are always positive, the negated result is always negative, hence the result type is always a signed integer type.

uns64 : If the original value was in the range $8000_0000_0000_0000..$FFFF_FFFF_FFFF_FFFF, then the resulting type is int128 , otherwise the resulting type is int64 . Since uns64 values are always positive, the negated result is always negative, hence the result type is always a signed integer type.

uns128 : The result type is always set to int128 . Note that there is no check for overflow. Effectively, HLA treats uns128 operands as though they were int128 operands with respect to negation. So really large positive ( uns128 ) values become smaller unsigned values after the negation. If you need to mix and match 128-bit values in an expression, you should attempt to limit your unsigned values to 127 bits.

byte, int8,

word, int16,

dword, int32,

qword, int64,

lword,

int128: Negates the expression (full 128 bits) and assigns the original expression type to the result.

real32 : Negates the real32 value and returns a real32 result.

real64 : Negates the real64 value and returns a real64 result.

real64 : Negates the real64 value and returns a real64 result.

cset : Computes the set complement (returns cset type). The set complement is all the items that were not previously in the set. Since HLA uses a bitmap representation for character sets, the complement of a character set is the same thing as inverting all the bits in the powerset.

If the operand is one of the integer types (signed, unsigned, hexadecimal), then HLA will set the type of the result to the smallest type within that class (signed, unsigned, or hexadecimal) that can hold the result (not including sign extension bits for negative numbers or zero extension bits for non-negative values).

expr1 * expr2

For numeric operands, the "*" operator produces their product. For character set operands, the "*"operator produces the intersection of the two sets. The exact result depends upon the types of the two operands to the "*" operator. To begin with, HLA attempts to make the types of the two operands identical if they are not already identical. HLA achives this via type promotion (see the discussion earlier).

If the operands are unsigned or hexadecimal operands, HLA will compute their unsigned product. If the operands are signed, HLA computes their signed product. If the operands are real, HLA computes their real product. If the operands are integer (signed or unsigned) and less than (or equal to) 64 bits, HLA computes their exact result. If the operands are greater than 64 bits and their product would require more than 128 bits, HLA quietly overflows without error. Note that HLA always performs a 128-bit multiplication, regardless of the operands' sizes; however, objects that require 64 bits or less of precision will always produce a product that is 128 bits or less. HLA automatically extends the size of the result to the next greater size if the product will not fit into an integer that is the same size as the operands. HLA will actually choose the smallest possible size for the product (e.g., if the result only requires 16 bits of precision, the resulting type will be uns16, int16 , or word ). The resulting type is always unsigned if the operands were unsigned, signed if the operands were signed, and hexadecimal if the operands were hexadecimal.

If the operands are real operands, HLA computes their product and always produces a real80 result. If you want to produce a smaller result via the '*' operator, use the real32 or real64 compile-time function to produce the smaller result, e.g., " real32( r32const * r32const2 ) ". Note that all real arithmetic inside HLA is always performed using the FPU, hence the results are always real80 . Other than trying to simulate the actual products a running program would produce, there is no real reason to coerce the product to a smaller value.

If the operands are character set operands, the '*' operator computes the intersection of the two sets. Since HLA uses a bitmap representation for character sets, this operator does a bitwise logical AND of the two 16-byte operands (this operation is roughly equivalent to applying the "&" operator to two lword operands).

If the operand is one of the integer types (signed, unsigned, hexadecimal), then HLA will set the type of the result to the smallest type within that class (signed, unsigned, or hexadecimal) that can hold the result (not including sign extension bits for negative numbers or zero extension bits for non-negative values).

expr1 div expr2

The two expressions must be integer (signed, unsigned, or hexadecimal) numbers. Supplying any other data type as an operand will produce an error. The div operator divides the first expression by the second and produces the truncated quotient result.

If the operands are unsigned, HLA will do a full 128/128 bit division and the resulting type will be unsigned (HLA sets the type to the smallest unsigned type that will completely hold the result). If the operands are signed, HLA will do a full 128/128 bit signed division and the resulting type will be the smallest intXX type that can hold the result. If the operands are hexadecimal values, HLA will do a full 128/128 bit unsigned division and set the resulting type to the smallest hex type that can hold the result.

Note that the div operator does not allow real operands. Use the "/" operator for real division.

HLA will set the type of the result to the smallest type within its class (signed, unsigned, or hexadecimal) that can hold the result (not including sign extension bits for negative numbers or zero extension bits for non-negative values).

expr1 mod expr2

The two expressions must be integer (signed, unsigned, or hexadecimal) numbers. The mod operator divides the first expression by the second and produces their remainder (this value is always positive).

If the operands are unsigned, HLA will do a full 128/128 bit division and return the remainder. The resulting type will be unsigned (HLA sets the type to the smallest unsigned type that will completely hold the result).

If the operands are signed, HLA will do a full 128/128 bit signed division and return the remainder. The resulting type will be the smallest intXX type that can hold the result.

If the operands are hexadecimal values, HLA will do a full 128/128 bit unsigned division and set the resulting type to the smallest hex type that can hold the result.

Note that the mod operator does not allow real operands. You'll have to define real modulus and write the expression yourself if you need the remainder from a real division.

HLA will set the type of the result to the smallest type within its class (signed, unsigned, or hexadecimal) that can hold the result (not including sign extension bits for negative numbers or zero extension bits for non-negative values).

expr1 / expr2

The two expressions must be numeric. The '/' operator divides the first expression by the second and produces their (real80) quotient result.

If the operands are integers (unsigned, signed, or hexadecimal) or the operands are real32 or real80 , HLA first converts them to real80 before doing the division operation. The expression result is always real80 .

expr1 << expr2

The two expressions must be integer (signed, unsigned, or hexadecimal) numbers. The second operand must be a small (32-bit or less) non-negative value in the range 0..128. The << operator shifts the first expression to the left the number of bits specified by the second expression. Note that the result may require more bits to hold than the original type of the left operand. Any bits shifted out of bit position 127 are lost.

HLA will set the type of the result to the smallest type within the left operan's class (signed, unsigned, or hexadecimal) that can hold the result (not including sign extension bits for negative numbers or zero extension bits for non-negative values). Note that the '<<' operator can yield a smaller type (specifcally, an eight bit type) if it shifts all the bits off the H.O. end of the number; generally, though, this operation produces larger result types than the left operand.

expr1 >> expr2

The two expressions must be integer (signed, unsigned, or hexadecimal) numbers. The second operand must be a small (32-bit or less) non-negative value in the range 0..128. The >> operator shifts the first expression to the right the number of bits specified by the second expression. Any bits shifted out of the L.O. bit are lost. Note that this shift is a logical shift right, not an arithmetic shift right (this is true even if the left operand is an INTxx value). Therefore, this operation always shifts a zero into bit position 127.

Shift rights may produce a smaller type that the value of the left operand. HLA will always set the type of the result value to the minimum type size that has the same base class as the left operand.

expr1 + expr2

If the two expressions are numeric, the "+" operator produces their sum.

If the two expressions are strings or characters, the "+" operator produces a new string by concatenating the right expression to the end of the left expression.

If the two operands are character sets, the "+" operator produces their union.

If the operands are integer values (signed, unsigned, or hexadecimal), then HLA adds them together. Any overflow out of bit #127 (unsigned or hexadecimal) or bit #126 (signed) is quietly lost. HLA sets the type of the result to the smallest type size that will hold the sum; the type class (signed, unsigned, hexadecimal) will be the same as the operands. Note that it is possible for the type size to grow or shrink depending on the values of the operands (e.g., adding a positive and negative number could reduce the type size, adding two positive or two negative numbers may expand the result type's size).

When adding two real values (or a real and an integer value), HLA always produces a real80 result.

Since HLA uses a bitmap to represent character sets, taking the union of two character sets is the same as doing a bitwise logical OR of all 16 bytes in the character set.

expr1 - expr2

If the two expressions are numeric, the "-" operator produces their difference.

If the two expressions are character sets, the "-" operator produces their set difference (that is, all the characters in expr1 that are not also in expr2).

If the operands are integer values (signed, unsigned, or hexadecimal), then HLA subtracts the right operand from the left operand. Any overflow out of bit #127 (unsigned or hexadecimal) or bit #126 (signed) is quietly lost. HLA sets the type of the result to the smallest type size that will hold their difference; the type class (signed, unsigned, hexadecimal) will be the same as the operands. Note that it is possible for the type size to grow or shrink depending on the values of the operands (e.g., subtracting two negative or non-negative numbers could reduce the type size, subtracting a negative value from a non-negative value may expand the result type's size).

When subtracting two real values (or a real and an integer value), HLA always produces a real80 result.

Since HLA uses a bitmap to represent character sets, taking the set of two character sets is the same as doing a bitwise logical AND of the left operand with the inverse of the right operand.

Comparisons (=, ==, <>, !=, <, <=, >, and >=)

expr1 = expr2

expr1 == expr2

expr1 <> expr2

expr1 != expr2

expr1 < expr2

expr1 <= expr2

expr1 > expr2

expr1 >= expr2

 

(note: "!=" and "<>" operators are identical. "=" and "==" operators are identical.)

The two expressions must be compatible (described earlier). These operators compare the two operands and return true or false depending upon the result of the comparison.

You may use the "=" and "<>" operators to compare two pointer constants (e.g., "&abc" or "&ptrVar"). The other operators do not allow pointer constant operands.

All the above operators allow you to compare boolean values, enumerated values (types must match), integer (signed, unsigned, hexadecimal) values, character values, string values, real values, and character set values.

When comparing boolean values, note that false < true .

One character set is less than another if it is a proper subset of the other. A character set is less than or equal to another set if it is a subset of that second set. Likewise, one character set is greater than, or greater than or equal to, another set if it is a proper superset, or a superset, respectively.

As with any programming language, you should take care when comparing two real values (especially for equality or inequality) as minor precision drifts can cause the comparison to fail.

expr1 & expr2

 

(note: "&&" and "&" mean different things to HLA. See the section on high level language control structures for details on the "&&" operator.)

The operands must both be boolean or they must both be numbers. With boolean operands the AND operator produces the logical and of the two operands (boolean result). With numeric operands, the AND operator produces the bitwise logical AND of the operands.

If the operand is one of the integer types (signed, unsigned, hexadecimal), then HLA will set the type of the result to the smallest type within that class (signed, unsigned, or hexadecimal) that can hold the result.

expr1 in expr2

The first expression must be a character value. The second expression must be a character set. The in operator returns true if the character is a member of the specified character set; it returns false otherwise.

expr1 | expr2

(note: "||" and "|" mean different things to HLA. See the section on high level language control structures for details on the "||" operator.)

The operands must both be boolean or they must both be numbers. With boolean operands the OR operator produces the logical OR of the two operands (boolean result). With numeric operands, the OR operator produces the bitwise or of the operands.

If the operand is one of the integer types (signed, unsigned, hexadecimal), then HLA will set the type of the result to the smallest type within that class (signed, unsigned, or hexadecimal) that can hold the result.

expr1 ^ expr2

The operands must both be boolean or they must both be numbers. With boolean operands the xor operator produces the logical exclusive-or of the two operands (boolean result). With number operands, the xor operator produces the bitwise exclusive-or of the operands.

If the operand is one of the integer types (signed, unsigned, hexadecimal), then HLA will set the type of the result to the smallest type within that class (signed, unsigned, or hexadecimal) that can hold the result.

( expr )

You may override the precedence of any operator(s) using parentheses in HLA constant expressions.

 

[ comma_separated_list_of_expressions ]

This produces an array expression. The type of the expression is an array type whose base element is the type of one of the expressions in the list. If there are two or more constant types in the array expression, HLA promotes the type of the array expression following the rules for mixed-mode arithmetic (see the rules earlier in this document).

 

record_type_name : [ comma_separated_list_of_field_expressions ]

This produces a record expression. The expressions appearing within the brackets must match the respective fields of the specified record type. See the discussion earlier in this chapter.

 

identifier

An identifier is a legal component of a constant expression if the identifier's classification is CONST or VAL (that is, the identifier was declared in a constant or value section of the program). The expression evaluator substitutes the current declared value and type of the symbol within the expression. Constant expressions allow the following types:

 

Boolean, enumerated types, Uns8, Uns16, Uns32, Uns64, Uns128 Byte, Word, DWord, QWord, LWord, Int8, Int16, Int32, Int64, Int128, Char, Real32, Real64, Real80, String, and Cset.

 

You may also specify arrays whose element base type is one of the above types (or a record or union subject to the following restriction). Likewise, you can specify record or union constants if all of their respective fields are one of the above primitive types or a value array, record, or union constant.

HLA allows array, record, and union constants. If you specify the name of an array, for example, HLA works with all the values of that array. Likewise, HLA can copy all the values of a record or union with a single statement.

HLA allows literal Unicode character and string constants (e.g., u'a' and u"unicode") or identifiers that are of wchar or wstring type in an expression, but no other terms are allowed in such an expression (as this is being written).

identifier1.identifier2 {...}

Selects a field from a record or union constant. Identifier1 must be a record or union object defined in a const or val section. Identifier2 (and any following dot-identifiers) must be a field of the record or union. HLA replaces this object with the value of the specified field.

Examples:

recval.fieldval

recval.subrecval.fieldval

 

Don't forget that with union constant, you may only access the last field into which you've actually stored data (see the section on union constants for more details).

identifier [ index_list ]

Identifier must be an array constant defined in either a const or val section. Index_list is a list of constant expressions separated by commas. The index list selects a specified element of the "identifier" array. HLA reports an error if you supply more indices than the array has dimensions. HLA returns an array slice if you specify fewer indices than the array has dimensions (for example, if an array is declared as "a:uns8[4,4]" and you specify "a[2]" in a constant expression, HLA returns the third row of the array (a[2,0]..a[2,3]) as the value of this term).

Examples:

arrayval[0]

aval[1,4,0]

Statement Labels

A statement label is an identifier that appears within a code section (i.e., the body of your main program or a procedure or iterator) followed by a colon. The identifier is given the type "label" and the value associated with the label is the current value of the location counter. Statement labels can be used as the target of a JMP or CALL instruction. You may also take their address with the "&" (address-of) operator.

Within an operand field, you may also use the "@here" reserved word to denote the current location counter value (by current, this means the address of the start of the current instruction). For example,

jmp @here;

creates an infinite loop, and

call @here[5];

transfers control to the subroutine three bytes beyond the call instruction.

Procedure Declarations

Procedure declarations are nearly identical to program declarations with two major differences: procedures are declared using the "procedure" reserved word and procedures may have parameters. The general syntax is:

 

procedure identifier ( optional_parameter_list ); procedure_options

declarations

begin identifier;

statements

end identifier;

 

Note that you may declare procedures inside other procedure in a fashion analogous to most block-structured languages (e.g., Pascal).

The optional parameter list consists of a list of var-type declarations taking the form:

 

optional_access_keyword identifier1 : identifier2 optional_in_reg

 

optional_access_keyword, if present, must be val, var, valres, result, name, or lazy and defines the parameter passing mechanism (pass by value, pass by reference, pass by value/result [or value/returned], pass by result, pass by name, or pass by lazy evaluation, respectively). The default is pass by value (val) if an access keyword is not present. For pass by value parameters, HLA allocates the specified number of bytes according to the size of that object in the activation record. For pass by reference, pass by value/result, and pass by result, HLA allocates four bytes to hold a pointer to the object. For pass by name and pass by lazy evaluation, HLA allocates eight bytes to hold a pointer to the associated thunk and a pointer to the thunk's execution environment (see the sections on parameters and thunks for more details).

The optional_in_reg clause, if present, corresponds to the phrase "in reg" where reg is one of the 80x86's general purpose 8-, 16-, or 32-bit registers. You must take care when passing parameters through the registers as the parameter names become aliases for registers and this can create confusion when reading the code later (especially if, within a procedure with a register parameter, you call another procedure that uses that same register as a parameter).

HLA also allows a special parameter of the form:

 

var identifer : var

 

This creates an untyped reference parameter. You may specify any memory variable as the corresponding actual parameter and HLA will compute the address of that object and pass it on to the procedure without further type checking. Within the procedure, the parameter is given the DWORD type.

The procedure_options component above is a list of keywords that specify how HLA emits code for the procedure. There are several different procedure options available: @noalignstack, @alignstack, @pascal, @stdcall, @cdecl, @align ( int_const), @use reg32 , @leave, @noleave, @enter, @noenter, and @returns ("text").

 

Option	Description
@noframe, @frame	By default, HLA emits code at the beginning of the procedure to construct a stack frame. The @noframe option disables this action ( noframe is depreciated, you should always use @noframe ). The @ frame option tells HLA to emit code for a particular procedure if stack frame generation is off by default. See the description of #frame and #noframe for details on controlling the default frame generation. HLA also uses these two special identifiers as a compile-time variable to set the default display generation for all procedures. Setting @frame to true (or @noframe to false) turns on frame generation by default; setting @frame to false (or @noframe to true) turns off frame generation.
@nodisplay, @display	By default, HLA emits code at the beginning of the procedure to construct a display within the frame. The @nodisplay option disable this action (@ nodisplay is depreciated, you should use @nodisplay ). The @ display option tells HLA to emit code to generate a display for a particular procedure if display generation is off by default. Note that HLA does not emit code to construct the display if '@ noframe ' is in effect, though it will assume that the programmer will construct this display themselves. HLA also uses these two special identifiers as a compile-time variable to set the default display generation for all procedures. Setting @display to true (or @nodisplay to false) turns on display generation by default; setting @display to false (or @nodisplay to true) turns off display generation.
@noalignstack, @alignstack	By default (assuming frame generation is active), HLA will an instruction to align ESP on a four-byte boundary after allocating local variables. Win32, Linux, and other 32-bit OSes require the stack to be dword-aligned (hence this option). If you know the stack will be dword-aligned, you can eliminate this extra instruction by specifying the @noalignstack option. Conversely, you can force the generation of this instruction by specifying the @ alignstack procedure option. HLA also uses these two special identifiers as a compile-time variable to set the default display generation for all procedures. Setting @alignstack to true (or @noalignstack to false) turns on stack alignment generation by default; setting @alignstack to false (or @noalignstack to true) turns off stack alignment code generation.
@pascal, @cdecl, @stdcall	These options give you the ability to specify the parameter passing mechanism for the procedure. By default, HLA uses the @ pascal calling sequence. This calling sequence pushes the parameters on the stack in a left-to-right order (i.e., in the order they appear in the parameter list). The @cdecl procedure option tells HLA to pass the parameters from right-to-left so that the first parameter appears at the lowest address in memory and that it is the user's responsibility to remove the parameters from the stack. The @stdcalll procedure option is a hybrid of the @ pascal and @ cdecl calling conventions. It pushes the parameters in the right-to-left order (just like @ cdecl ) but @ stdcall procedures automatically remove their parameter data from the stack (just like @ pascal ). Win32 API calls use the @ stdcall calling convention.
@align( int_constant )	The @ align ( int_const ) procedure option aligns the procedure on a 1, 2, 4, 8, or 16 byte boundary. Specify the boundary you desire as the parameter to this option. The default is @align(1) , which is unaligned; HLA also uses this special identifiers as a compile-time variable to set the default procedure alignment . Setting @align := 1 turns off procedure alignment while supplying some other value (which must be a power of two) sets the default procedure alignment to the specified number of bytes.
@use reg32	When passing parameters, HLA can sometimes generate better code if it has a 32-bit general purpose register for use as a scratchpad register. By default, HLA never modifies the value of a register behind your back; so it will often generate less than optimal code when passing certain parameters on the stack. By using the @use procedure option, you can specify one of the following 32-bit registers for use by HLA: eax, ebx, ecx, edx, esi , or edi . By providing one of these registers, HLA may be able to generate significantly better code when passing certain parameters.
@returns( "text" )	This option specifies the compile-time return value whenever a function name appears as an instruction operand. For example, suppose you are writing a function that returns its result in EAX. You should probably specify a "returns" value of "EAX" so you can compose that procedure just like any other HLA machine instruction (see the example below and the section on machine instructions for more details).
@leave, @noleave	These two options control the code generation for the standard exit sequence. If you specify the @leave option then HLA emits the x86 LEAVE instruction to clean up the activation record before the procedure returns. If you specify the @noleave option, then HLA emits the primitive instructions to achieve this, e.g., mov( ebp, esp ); pop( ebp );   The manual sequence is faster on some architectures, the LEAVE instruction is always shorter.   Note that @noleave occurs by default if you've specified @noframe or #noframe . By default, HLA assumes @noleave but you may change the default using these special identifiers as a compile-time variable to set the default LEAVE generation for all procedures. Setting @leave to true (or @noleave to false) turns on LEAVE generation by default; setting @leave to false (or @noleave to true) turns off the use of the LEAVE instruction.
@enter, @noenter	These two options control the code generation for a procedure's standard entry sequence. If you specify the @enter option then HLA emits the x86 ENTER instruction to create the activation record. If you specify the @noenter option, then HLA emits the primitive instructions to achieve this.   The manual sequence is always faster, using the ENTER instruction is usually shorter.   Note that @noenter occurs by default if you've specified @noframe or #noframe . By default, HLA assumes @noenter but you may change the default using these special identifiers as a compile-time variable to set the default ENTER generation for all procedures. Setting @enter to true (or @noenter to false) turns on ENTER generation by default; setting @enter to false (or @noenter to true) turns off the use of the ENTER instruction.

The following example demonstrates how the @ returns option works:

 

program returnsDemo;
#include( "stdio.hhf" );

procedure eax0; @returns( "eax" );
begin eax0;

mov( 0, eax );

end eax0;

begin returnsDemo;

mov( eax0(), ebx );
stdout.put( "ebx=", ebx, nl );

end returnsDemo;

To help those who insist on constructing the activation record themselves, HLA declares two local constants within each procedure: _vars_ and _parms_ . The _vars_ symbol is an integer constant that specifies the number of local variables declared in the procedure. This constant is useful when allocating storage for your local variables. The _parms_ constants specifies the number of bytes of parameters. You would normally supply this constant as the parameter to a ret() instruction to automatically clean up the procedure's parameters when it returns.

If you do not specify @ nodisplay , then HLA defines a run-time variable named _display_ that is an array of pointers to activation records. For more details on the _display_ variable, see the section on lexical scope.

You can also declare @external procedures (procedures defined in other HLA units or written in languages other than HLA) using the following syntaxes:

 

procedure externProc1 (optional parameters) ; @returns( "text" ); @external;

 

procedure externProc2 (optional parameters) ;

@returns( "text" ); @external( "external_name" );

 

As with normal procedure declarations, the parameter list and @ returns clause are optional.

The first form is generally used for HLA-written functions. HLA will use the procedure's name (externProc1 in this case) as external name.

The second form lets you refer to the procedure by one name ( externProc2 in this case) within your HLA program and by a different name ("different_name" in this example) in the MASM generated code. This second form has two main uses: (1) if you choose an external procedure name that just happens to be a MASM reserved word, the program may compile correctly but fail to assemble. Changing the external name to something else solves this problem. (2) When calling procedures written in external languages you may need to specify characters that are not legal in HLA identifiers. For example, Win32 API calls often use names like "WriteFile@24" containing illegal (in HLA) identifier symbols. The string operand to the external option lets you specify any name you choose. Of course, it is your responsibility to see to it that you use identifiers that are compatible with the linker and MASM, HLA doesn't check these names.

By default, HLA does the following:

• Creates a display for every procedure.
• Emits code to construct the stack frame for each procedure.
• Emits code to align ESP on a four-byte boundary upon procedure entry.
• HLA assumes that it cannot modify any register values when passing (non-register) parameters.
• The first instruction of the procedure is unaligned.

 

These options are the most general and "safest" for beginning assembly language programmers. However, the code HLA generates for this general case may not be as compact or as fast as is possible in a specific case. For example, few procedures will actually need a display data structure built upon procedure activation. Therefore, the code that HLA emits to build the display can reduce the efficiency of the program. Advanced programmers, of course, can use procedure options like "@nodisplay" to tell HLA to skip the generation of this code. However, if a program contains many procedures and none of them need a display, continually adding the "@nodisplay" option can get really old. Therefore, HLA allows you to treat these directives as "pseudo-compile-time-variables" to control the default code generation. E.g.,

? @display := true; // Turns on default display generation.

? @display := false; // Turns off default display generation.

? @nodisplay := true; // Turns off default display generation.

? @nodisplay := false; // Turns on default display generation.

 

? @frame := true; // Turns on default frame generation.

? @frame := false; // Turns off default frame generation.

? @noframe := true; // Turns off default frame generation.

? @noframe := false; // Turns on default frame generation.

 

? @alignstack := true; // Turns on default stk alignment code generation.

? @alignstack := false; // Turns off default stk alignment code generation.

? @noalignstack := true; // Turns off default stk alignment code generation.

? @noalignstack := false; // Turns on default stk alignment code generation.

 

? @enter := true; // Turns on default ENTER code generation.

? @enter := false; // Turns off default ENTER code generation.

? @noenter := true; // Turns off default ENTER code generation.

? @noenter := false; // Turns on default ENTER code generation.

 

? @leave := true; // Turns on default LEAVE code generation.

? @leave := false; // Turns off default LEAVE code generation.

? @noleave := true; // Turns off default LEAVE code generation.

? @noleave := false; // Turns on default LEAVE code generation.

 

?@align := 1; // Turns off procedure alignment (align on byte boundary).

?@align := int_expr; // Sets alignment, must be a power of two.

 

These directives may appear anywhere in the source file. They set the internal HLA default values and all procedure declarations following one of these assignments (up to the next, corresponding assignment) use the specified code generation option(s). Note that you can override these defaults by using the corresponding procedure options mentioned earlier.