Mercurial > hg > index.cgi
diff docs/manual/x659.html.orig @ 455:cad5937314cb
Add operandsizewarning pragma
Add operandsizewarning pragma that will raise warnings for certain operands
if the operand size could be smaller. (Long branch used instead of short
branch, for instance.)
| author | William Astle <lost@l-w.ca> |
|---|---|
| date | Fri, 16 Feb 2018 22:53:46 -0700 |
| parents | |
| children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/docs/manual/x659.html.orig Fri Feb 16 22:53:46 2018 -0700 @@ -0,0 +1,553 @@ +<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN""http://www.w3.org/TR/html4/loose.dtd"> +<HTML +><HEAD +><TITLE +>Assembler Modes and Pragmas</TITLE +><META +NAME="GENERATOR" +CONTENT="Modular DocBook HTML Stylesheet Version 1.79"><LINK +REL="HOME" +TITLE="LW Tool Chain" +HREF="index.html"><LINK +REL="UP" +TITLE="LWASM" +HREF="c62.html"><LINK +REL="PREVIOUS" +TITLE="Object Files and Sections" +HREF="x595.html"><LINK +REL="NEXT" +TITLE="Convenience Instructions" +HREF="x800.html"></HEAD +><BODY +CLASS="SECTION" +BGCOLOR="#FFFFFF" +TEXT="#000000" +LINK="#0000FF" +VLINK="#840084" +ALINK="#0000FF" +><DIV +CLASS="NAVHEADER" +><TABLE +SUMMARY="Header navigation table" +WIDTH="100%" +BORDER="0" +CELLPADDING="0" +CELLSPACING="0" +><TR +><TH +COLSPAN="3" +ALIGN="center" +>LW Tool Chain</TH +></TR +><TR +><TD +WIDTH="10%" +ALIGN="left" +VALIGN="bottom" +><A +HREF="x595.html" +ACCESSKEY="P" +>Prev</A +></TD +><TD +WIDTH="80%" +ALIGN="center" +VALIGN="bottom" +>Chapter 3. LWASM</TD +><TD +WIDTH="10%" +ALIGN="right" +VALIGN="bottom" +><A +HREF="x800.html" +ACCESSKEY="N" +>Next</A +></TD +></TR +></TABLE +><HR +ALIGN="LEFT" +WIDTH="100%"></DIV +><DIV +CLASS="SECTION" +><H1 +CLASS="SECTION" +><A +NAME="AEN659" +>3.10. Assembler Modes and Pragmas</A +></H1 +><P +>There are a number of options that affect the way assembly is performed. +Some of these options can only be specified on the command line because +they determine something absolute about the assembly process. These include +such things as the output target. Other things may be switchable during +the assembly process. These are known as pragmas and are, by definition, +not portable between assemblers.</P +><P +>LWASM supports a number of pragmas that affect code generation or +otherwise affect the behaviour of the assembler. These may be specified by +way of a command line option or by assembler directives. The directives +are as follows.</P +><P +></P +><DIV +CLASS="VARIABLELIST" +><DL +><DT +>PRAGMA <CODE +CLASS="PARAMETER" +>pragma[,...]</CODE +></DT +><DD +><P +>Specifies that the assembler should bring into force all <CODE +CLASS="PARAMETER" +>pragma</CODE +>s +specified. Any unrecognized pragma will cause an assembly error. The new +pragmas will take effect immediately. This directive should be used when +the program will assemble incorrectly if the pragma is ignored or not supported.</P +></DD +><DT +>*PRAGMA <CODE +CLASS="PARAMETER" +>pragma[,...]</CODE +></DT +><DD +><P +>This is identical to the PRAGMA directive except no error will occur with +unrecognized or unsupported pragmas. This directive, by virtue of starting +with a comment character, will also be ignored by assemblers that do not +support this directive. Use this variation if the pragma is not required +for correct functioning of the code.</P +></DD +><DT +>*PRAGMAPUSH <CODE +CLASS="PARAMETER" +>pragma[,...]</CODE +></DT +><DD +><P +>This directive saves the current state of the specified pragma(s) for later retrieval. See discussion below for more information.</P +><P +>This directive will not throw any errors for any reason.</P +></DD +><DT +>*PRAGMAPOP <CODE +CLASS="PARAMETER" +>pragma[,...]</CODE +></DT +><DD +><P +>This directive restores the previously saved state of the specified pragma(s). See discussion below for more information.</P +><P +>This directive will not throw any errors for any reason.</P +></DD +></DL +></DIV +><P +>Each pragma supported has a positive version and a negative version. +The positive version enables the pragma while the negative version disables +it. The negatitve version is simply the positive version with "no" prefixed +to it. For instance, "pragma" vs. "nopragma". When only one version is +listed below, its opposite can be obtained by prepending "no" if it is not +present or removing "no" from the beginning if it is present.</P +><P +>Pragmas are not case sensitive.</P +><P +></P +><DIV +CLASS="VARIABLELIST" +><DL +><DT +>6800compat</DT +><DD +><P +>When in force, this pragma enables recognition of various +compatibility instructions useful when assembling 6800 code. These +compatibility instructions are assembled into equivalent 6809 instructions. +This mode also includes several analogous instructions which are not +strictly 6800 instructions but allow the similar style to be applied to 6809 +specific features.</P +><P +>Technically, a compliant 6809 assembler must recognize these +instructions by default since Motorola advertised the 6809 as being source +compatible with the 6800. However, most source code does not require this +compatibility and LWASM itself did not support these instructions prior to +version 4.11 so this mode is disabled by default.</P +></DD +><DT +>6809</DT +><DD +><P +>This pragma allows you to mark a section of code as 6809-only. In ths mode, +the assembler will throw an error if any 6309 instructions are used.</P +></DD +><DT +>6309</DT +><DD +><P +>This pragma enables the use of 6309 instructions and disables any 6809 specific +instructions. It also changes the cycle count listing output (if selected) +to display 6309 timings.</P +></DD +><DT +>6809conv, 6309conv</DT +><DD +><P +>These pragmas enable convenience instructions extending the 6809 and 6309 +instruction sets respectively. For more information, see +<A +HREF="x800.html" +>Section 3.11</A +>.</P +></DD +><DT +>index0tonone</DT +><DD +><P +>When in force, this pragma enables an optimization affecting indexed addressing +modes. When the offset expression in an indexed mode evaluates to zero but is +not explicity written as 0, this will replace the operand with the equivalent +no offset mode, thus creating slightly faster code. Because of the advantages +of this optimization, it is enabled by default.</P +></DD +><DT +>cescapes</DT +><DD +><P +>This pragma will cause strings in the FCC, FCS, and FCN pseudo operations to +have C-style escape sequences interpreted. The one departure from the official +spec is that unrecognized escape sequences will return either the character +immediately following the backslash or some undefined value. Do not rely +on the behaviour of undefined escape sequences.</P +></DD +><DT +>importundefexport</DT +><DD +><P +>This pragma is only valid for targets that support external references. When +in force, it will cause the EXPORT directive to act as IMPORT if the symbol +to be exported is not defined. This is provided for compatibility with the +output of gcc6809 and should not be used in hand written code. Because of +the confusion this pragma can cause, it is disabled by default.</P +></DD +><DT +>undefextern</DT +><DD +><P +>This pragma is only valid for targets that support external references. When in +force, if the assembler sees an undefined symbol on the second pass, it will +automatically define it as an external symbol. This automatic definition will +apply for the remainder of the assembly process, even if the pragma is +subsequently turned off. Because this behaviour would be potentially surprising, +this pragma defaults to off.</P +><P +>The primary use for this pragma is for projects that share a large number of +symbols between source files. In such cases, it is impractical to enumerate +all the external references in every source file. This allows the assembler +and linker to do the heavy lifting while not preventing a particular source +module from defining a local symbol of the same name as an external symbol +if it does not need the external symbol. (This pragma will not cause an +automatic external definition if there is already a locally defined symbol.)</P +><P +>This pragma will often be specified on the command line for large projects. +However, depending on the specific dynamics of the project, it may be sufficient +for one or two files to use this pragma internally.</P +></DD +><DT +>export</DT +><DD +><P +>This pragma causes all symbols to be added to the export list +automatically. This is useful when a large number of symbols need to be +exported but you do not wish to include an EXPORT directive for all of them. +This is often useful on the command line but might be useful even inline +with the PRAGMA directive if a large number of symbols in a row are to be +exported.</P +></DD +><DT +>dollarlocal</DT +><DD +><P +>When set, a "$" in a symbol makes it local. When not set, "$" does not +cause a symbol to be local. It is set by default except when using the OS9 +target.</P +></DD +><DT +>dollarnotlocal</DT +><DD +><P +> This is the same as the "dollarlocal" pragma except its sense is +reversed. That is, "dollarlocal" and "nodollarnotlocal" are equivalent and +"nodollarlocal" and "dollarnotlocal" are equivalent. </P +></DD +><DT +>pcaspcr</DT +><DD +><P +> Normally, LWASM makes a distinction between PC and PCR in program +counter relative addressing. In particular, the use of PC means an absolute +offset from PC while PCR causes the assembler to calculate the offset to the +specified operand and use that as the offset from PC. By setting this +pragma, you can have PC treated the same as PCR. </P +></DD +><DT +>shadow</DT +><DD +><P +>When this pragma is in effect, it becomes possible to define a macro +that matches an internal operation code. Thus, it makes it possible to +redefine either CPU instructions or pseudo operations. Because this feature +is of dubious utility, it is disabled by default.</P +></DD +><DT +>nolist</DT +><DD +><P +>Lines where this pragma is in effect will not appear in the assembly +listing. Also, any symbols defined under this pragma will not show up in +the symbol list. This is most useful in include files to avoid spamming the +assembly listing with dozens, hundreds, or thousands of irrelevant +symbols.</P +></DD +><DT +>autobranchlength</DT +><DD +><P +>One of the perennial annoyances for 6809 programmers is that the +mneumonics for the short and long branch instructions are different (bxx vs. +lbxx), which is at odds with the rest of the instruction set. This pragma +is a solution to those annoying byte overflow errors that short branch +instructions tend to aquire.</P +><P +>When this pragma is in effect, which is not the default, whenever any +relative branch instruction is used, its size will be automatically +determined based on the actual distance to the destination. In other words, +one can write code with long or short branches everywhere and the assembler +will choose a size for the branch.</P +><P +>Also, while this pragma is in effect, the > and < symbols can be used +to force the branch size, analogous to their use for other instructions with +< forcing 8 bit offsets and > forcing 16 bit offets.</P +><P +>Because this pragma leads to source that is incompatible with other +assemblers, it is strongly recommended that it be invoked using the PRAGMA +directive within the source code rather than on the command line or via the +*PRAGMA directive. This way, an error will be raised if someone tries to +assemble the code under a different assembler.</P +></DD +><DT +>nosymbolcase, symbolnocase</DT +><DD +><P +>Any symbol defined while this pragma is in force will be treated as +case insensitive, regardless whether the pragma is in force when the symbol +is referenced.</P +><P +>It is important to note that this pragma will not work as expected in +all cases when using the object file assembly target. It is intended for +use only when the assembler will be producing the final binary.</P +></DD +><DT +>condundefzero</DT +><DD +><P +>This pragma will cause the assembler to change the way it handles +symbols in conditional expressions. Ordinarily, any symbol that is not +defined prior to the conditional will throw an undefined symbol error. With +this pragma in effect, symbols that are not yet defined at the point the +conditional is encountered will be treated as zero.</P +><P +>This is not the default because it encourages poor code design. One +should use the "IFDEF" or "IFNDEF" conditionals to test for the presence of +a symbol.</P +><P +>It is important to note that if a symbol is defined but it does not +yet evaluate to a constant value at the point where the conditional appears, +the assembler will still complain about a non constant condition.</P +></DD +><DT +>forwardrefmax</DT +><DD +><P +>This pragma will disable forward reference optimization completely. +Ordinarily, LWASM will attempt to select the shortest possible addressing +mode for forward references. However, in many source files, especially +those not using the PCR relative addressing modes, this optimization is +pointless since the assembler will almost certainly settle on a 16 bit +offset or address. If all variables in the direct page are defined before +the main body of the code, the benefit of forward reference optimization +almost certainly vanishes completely. However, the cost of doing that +optimization remains and can result in a very long assembly time.</P +><P +>Enabling this pragma will cause all forward references to use the +maximum offset or address size, much the same has EDTASM and other pure +two pass assemblers do. The side effect is that all line lengths and +symbol values are fully resolved after the initial parsing pass and the +amount of work to resolve everything becomes almost nil.</P +><P +>While this pragma can be applied selectively to sections of source +code (use *PRAGMA if doing so and compatibility with other assemblers +is desired), it is likely more useful when provided as a command line +pragma.</P +><P +>It should be noted that the presence or absence of this pragma +will not change the correctness of the generated code unless cycle counts +or byte counts are critical (which they usually are not). It also will +not override the operand size override prefixes (< and >). It only +applies when the assembler is left to guess what the operand size is.</P +></DD +><DT +>operandsizewarning</DT +><DD +><P +>Enabling this pragma will cause LWASM to show a warning when it +detects that a smaller addressing mode could be used for an instruction. +This is particularly useful for finding places where long branches are used +where short branches could be used instead. It will also show the warnings +for indexing offsets (regardless of whether the operand size is +forced).</P +><P +>As of LWASM 4.16, no other checks are performed.</P +></DD +><DT +>qrts</DT +><DD +><P +> Enables the use of the ?RTS branch target. ?RTS is implemented to maintain +compatibility with the MACRO-80c assembler. It works by searching backward +in the code for an RTS instruction. If none is found, it inverts the branch +logic and inserts an RTS following the branch instruction. Below you can +see how a BMI (2B xx) has been assembled as a BPL *+1 (2A 01) to skip over an +inserted RTS (39).</P +><PRE +CLASS="PROGRAMLISTING" +>1D1E 7D1D1D TST WHICH1 +1D21 2A0139 BMI ?RTS +1D24 BD1D65 JSR INV</PRE +></DD +><DT +>m80ext</DT +><DD +><P +> This pragma (along with pragma qrts) enables some uncommon behaviors to +accomodate The Micro Works MACRO-80c assembler from 1982. This assembler +was used by a number of notable TRS-80 Color Computer applications and the +goal of this pragma is to allow them to build identical binaries from +unmodified, vintage source code.</P +><P +> In m80ext mode, the handling of the "END" pseudo-op changes when used inside +an include file. Instead of terminating all assembly, it merely stops +processing of the current include file (this behavior matches the original +Motorola 6809 assembler). In addition, loading an ASCII value with a single +quote (e.g., LDA #'N) is extended to 16-bit registers (e.g., LDD #'NO). +LWASM normally supports this via double quote and that is the proper use in +modern code. Finally, the FCC pseudo-op is extended to handle FCB-like +behavior after the closing delimiter:</P +><PRE +CLASS="PROGRAMLISTING" +> FCC "Greetings from 1982",13,0</PRE +></DD +><DT +>testmode</DT +><DD +><P +> This pragma is intended for internal testing purposes. In testmode, the +assembler searches for a specially-formatted comment starting with a +semicolon followed by a period. Immediately afterward are a list of hex +bytes that the assembler is expected to generate. Likewise, if the +assembler is expected to throw an error or warning on a given line, you can +check by specifying "E:" followed by the error number. In this case the +error is ignored and the assembler continues ignoring the line in question. </P +><PRE +CLASS="PROGRAMLISTING" +>1D1E 7D1D1D TST WHICH1 ;.7d1d1d +1D21 2A0139 BMI ?RTS ;.2a0139 +1D24 1D24 FDB * ;.1d24 +1D26 xyz INV ;.E:32 (Error 32 is "Bad opcode")</PRE +></DD +></DL +></DIV +><P +>As a convenience, each input file has a pragma state stack. This +allows, through the use of *PRAGMAPUSH and *PRAGMAPOP, a file to change a +pragma state and then restore it to the precise state it had previously. +If, at the end of an input file, all pragma states have not been popped, +they will be removed from the stack. Thus, it is critical to employ +*PRAGMAPOP correctly. Because each input file has its own pragma stack, +using *PRAGMAPUSH in one file and *PRAGMAPOP in another file will not +work.</P +><P +>Pragma stacks are more useful in include files, in particular in +conjunction with the nolist pragma. One can push the state of the nolist +pragma, engage the nolist pragma, and then pop the state of the nolist +pragma at the end of the include file. This will cause the entire include +file to operate under the nolist pragma. However, if the file is included +while nolist is already engaged, it will not undo that state.</P +></DIV +><DIV +CLASS="NAVFOOTER" +><HR +ALIGN="LEFT" +WIDTH="100%"><TABLE +SUMMARY="Footer navigation table" +WIDTH="100%" +BORDER="0" +CELLPADDING="0" +CELLSPACING="0" +><TR +><TD +WIDTH="33%" +ALIGN="left" +VALIGN="top" +><A +HREF="x595.html" +ACCESSKEY="P" +>Prev</A +></TD +><TD +WIDTH="34%" +ALIGN="center" +VALIGN="top" +><A +HREF="index.html" +ACCESSKEY="H" +>Home</A +></TD +><TD +WIDTH="33%" +ALIGN="right" +VALIGN="top" +><A +HREF="x800.html" +ACCESSKEY="N" +>Next</A +></TD +></TR +><TR +><TD +WIDTH="33%" +ALIGN="left" +VALIGN="top" +>Object Files and Sections</TD +><TD +WIDTH="34%" +ALIGN="center" +VALIGN="top" +><A +HREF="c62.html" +ACCESSKEY="U" +>Up</A +></TD +><TD +WIDTH="33%" +ALIGN="right" +VALIGN="top" +>Convenience Instructions</TD +></TR +></TABLE +></DIV +></BODY +></HTML +> \ No newline at end of file