1
0
mirror of https://github.com/fadden/6502bench.git synced 2024-11-03 23:06:09 +00:00
6502bench/docs/sgmanual/codegen.html
Andy McFadden 22c47e1d0b Update reference manual formatting
Switched from XHTML to HTML5.  Added formatting for menu items and
keyboard shortcuts.

Made various minor edits to the text.
2021-10-18 17:56:08 -07:00

436 lines
21 KiB
HTML

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8"/>
<meta name="viewport" content="width=device-width, initial-scale=1" />
<link rel="stylesheet" href="main.css"/>
<title>Code Generation &amp; Assembly - 6502bench SourceGen</title>
</head>
<body>
<div id="content">
<h1>SourceGen: Code Generation &amp; Assembly</h1>
<p><a href="index.html">Back to index</a></p>
<p>SourceGen can generate an assembly source file that, when fed into
the target assembler, will recreate the original data file exactly.
Every assembler is different, so support must be added to SourceGen
for each.</p>
<p>The generation / assembly dialog can be opened with
<samp>File &gt; Assemble</samp>.</p>
<p>If you want to show code to others, perhaps by adding a page to
your web site, you can "export" the formatted code as text or HTML.
This is explained in more detail <a href="#export-source">below</a>.
<h2 id="generate">Generating Source Code</h2>
<p>Cross assemblers tend to generate additional files, either compiler
intermediaries ("<samp>file.o</samp>") or metadata
("<samp>_FileInformation.txt</samp>"). Some
generators may produce multiple source files, perhaps a link script or
symbol definition header to go with the assembly source. To avoid
spreading files across the filesystem, SourceGen does all of its work
in the same directory where the project lives. Before you can generate
code, you have to have assigned your project a directory. This is why
you can't assemble a project until you've saved it for the first time.</p>
<p>The <samp>Generate and Assemble</samp> dialog has a drop-down list near
the top that lets you pick which assembler to target. The name of the
assembler will be shown with the detected version number. If the assembler
executable isn't configured, "<samp>[latest version]</samp>" will be
shown instead of a version number.</p>
<p>The <samp>Settings</samp> button will take you directly to the
assembler configuration tab in the application settings dialog.</p>
<p>Hit the <samp>Generate</samp> button to generate the source code into
a file on disk.
The file will use the project name, with the <code>.dis65</code> extension
replaced by <code>_&lt;assembler&gt;.S</code>.</p>
<p>The first 64KiB of each generated file will be shown in the preview
window. If multiple files were generated, you can use the "preview file"
drop-down to select between them. Line numbers are
prepended to each line to make it easier to track down errors.</p>
<h3 id="localizer">Label Localizer</h3>
<p>The label localizer is an optional feature that automatically converts
some labels to an assembler-specific less-than-global label format. Local
labels may be reusable (e.g. using "<code>@LOOP</code>" for multiple
consecutive loops is easier to understand than giving each one a unique
label) or reduce the size of a generated link table. There are usually
restrictions on local labels, e.g. references to them may not be allowed
to cross a global label definition, which the localizer factors in
automatically.</p>
<h3 id="reserved-labels">Reserved Label Names</h3>
<p>Some label names aren't allowed. For example, 64tass reserves the
use of labels that begin with two underscores. Most assemblers will
also prevent you from using opcode mnemonics as labels (which means
you can't assemble the infinite loop <code>jmp jmp jmp</code>).</p>
<p>If a label doesn't appear to be legal, the generated code will use
a suitable replacement (e.g. <code>jmp_1 jmp jmp_1</code>).</p>
<h3 id="platform-features">Platform-Specific Features</h3>
<p>SourceGen needs to be able to assemble binaries for any system
with any assembler, so it generally avoids platform-specific features.
One exception to that is C64 PRG files.</p>
<p>PRG files start with a 16-bit value that tells the OS where the
rest of the file should be loaded. The value is not usually part of
the source code, but instead is generated by the assembler, based on
the address of the first byte output. If SourceGen detects that
a file is PRG, the source generators for some assemblers will suppress
the first 2 bytes, and instead pass appropriate meta-data (such as
an additional command-line option) to the assembler.</p>
<p>A file is treated as a PRG if:</p>
<ul>
<li>it is between 3 and 65536 bytes long (inclusive)</li>
<li>the format at offset +000000 is a 16-bit numeric data item
(not executable code, not two 8-byte values, not the first part
of a 24-bit value, etc.)</li>
<li>there is an address region start directive at +000002
<li>the 16-bit value at +000000 is equal to the address of the
byte at +000002</li>
<li>there is no label at offset +000000 (explicit or auto-generated)</li>
</ul>
<p>The definition is sufficiently narrow to avoid most false-positives.
If a file is being treated as PRG and you'd rather it weren't, you
can add a label or reformat the bytes. This feature is currently only
enabled for 64tass.</p>
<h2 id="assemble">Cross-Assembling Generated Code</h2>
<p>After generating sources, if you have a cross-assembler executable
configured, you can run it by clicking the <samp>Run Assembler</samp> button.
The command-line output will be displayed, with stdout and stderr separated.
(I'd prefer them to be interleaved, but that's not what the system
provides.)</p>
<p>The output will show the assembler's exit code, which will be zero
on success (note: sometimes they lie). If it appeared to succeed,
SourceGen will then compare the assembler's output to the original file,
and report any differences.</p>
<p>Failures here may be due to bugs in the cross-assembler or in
SourceGen. However, SourceGen can generally work around assembler bugs,
so any failure is an opportunity for improvement.</p>
<h2 id="supported">Supported Assemblers</h2>
<p>SourceGen currently supports the following cross-assemblers:</p>
<ul>
<li><a href="#64tass">64tass</a></li>
<li><a href="#acme">ACME</a></li>
<li><a href="#cc65">cc65</a></li>
<li><a href="#merlin32">Merlin 32</a></li>
</ul>
<h3 id="version">Version-Specific Code Generation</h3>
<p>Code generation must be tailored to the specific version of the
assembler. This is most easily understood with an example.</p>
<p>If the code has a statement like <code>MVN #$01,#$02</code>, the
assembler is expected to output <code>54 02 01</code>, with the arguments
reversed. cc65 v2.17 got it backward; the behavior was fixed in v2.18. The
bug means we can't generate the same <code>MVN</code>/<code>MVP</code>
instructions for both versions of the assembler.</p>
<p>Having version-dependent source code is a bad idea. If we generated
reversed operands (<code>MVN #$02,#$01</code>), we'd get the correct
output with v2.17, but the wrong output for v2.18. Unambiguous code can
be generated for all versions of the assembler by just outputting raw hex
bytes, but that's ugly and annoying, so we don't want to be stuck doing
that forever. We want to detect which version of the assembler is in
use, and output actual <code>MVN</code>/<code>MVP</code> instructions
when producing code for versions of the assembler that don't have the bug.</p>
<p>When you configure a cross-assembler, SourceGen runs the executable with
version query arguments, and extracts the version information from the output
stream. This is used by the generator to ensure that the output will work
correctly with the installed assembler.
If the assembler is present on the system, SourceGen will produce code
optimized for the latest supported version of the assembler.</p>
<h3 id="quirks">Assembler-Specific Bugs &amp; Quirks</h3>
<p>This is a list of bugs and quirky behavior in cross-assemblers that
SourceGen works around when generating code.</p>
<p>Every assembler seems to have a different way of dealing with expressions.
Most of them will let you group expressions with parenthesis, but that
doesn't always help. For example, <code>PEA label &gt;&gt; 8 + 1</code> is
perfectly valid, but writing <code>PEA (label &gt;&gt; 8) + 1</code> will cause
most assemblers to assume you're trying to use an alternate (and non-existent)
form of <code>PEA</code> with indirect addressing, causing the assembler
to halt with an error message. The code generator needs
to understand expression syntax and operator precedence to generate correct
code, but also needs to know how to handle the corner cases.</p>
<h4>Undocumented Opcodes</h4>
<p>The data sheet for the 6502 does not define all 256 possible opcodes.
Analysis and experimentation have found that many of these "undocumented"
operations actually do useful things. The people who did the research
didn't always use the same mnemonic names for them, which led to a bit
of confusion in assemblers.</p>
<p>The most authoritative source is
<a href="https://csdb.dk/release/download.php?id=229739"><i>NMOS 6510 Unintended Opcodes</i> (PDF)</a>.
The document defines a primary mnemonic and lists common aliases for
each operation. SourceGen will output the primary mnemonic unless the
target assembler doesn't handle it.</p>
<h3 id="64tass">64tass</h3>
<p>Tested versions: v1.53.1515, v1.54.1900, v1.55.2176, v1.56.2625
<a href="https://sourceforge.net/projects/tass64/">[web site]</a></p>
<p>Bugs:</p>
<ul>
<li>[Fixed in v1.55.2176]
Undocumented opcode <code>SHA (ZP),Y</code> ($93) is not supported;
the assembler appears to be expecting <code>SHA ABS,X</code> instead.</li>
<li>[Fixed in v1.55.2176] WDM is not supported.</li>
</ul>
<p>Quirks:</p>
<ul>
<li>The underscore character ('_') is allowed as a character in labels,
but when used as the first character in a label it indicates the
label is local. If you create labels with leading underscores that
are not local, the labels must be altered to start with some other
character, and made unique.</li>
<li>Labels starting with two underscores are "reserved". Trying to
use them causes an error.</li>
<li>By default, 64tass sets the first two bytes of the output file to
the load address. The <code>--nostart</code> flag is used to
suppress this.</li>
<li>By default, 64tass is case-insensitive, but SourceGen treats labels
as case-sensitive. The <code>--case-sensitive</code> flag must be passed
to the assembler.</li>
<li>If you set the <code>--case-sensitive</code> flag, <b>all</b> opcodes
and operands must be lower-case. Most of the SourceGen options that
cause things to appear in upper case must be disabled.</li>
<li>For 65816, selecting the bank byte is done with the grave accent
character ('`') rather than the caret ('^'). (There's a note in the
docs to the effect that they plan to move to carets.)</li>
<li>Instructions whose argument is formed by combining with the
65816 Program Bank Register (16-bit <code>JMP</code>/<code>JSR</code>)
must be specified as 24-bit values for code that lives outside bank 0.
This is true for both symbols and raw hex (e.g. <code>JSR $1234</code>
is invalid outside bank 0). Attempting to JSR to a label in bank
0 from outside bank 0 causes an error, even though it is technically
a 16-bit operand.</li>
<li>The arguments to COP and BRK require immediate-mode syntax
(<code>COP #$03</code> rather than <code>COP $03</code>).
<li>For historical reasons, the default behavior of the assembler is to
assume that the source file is PETSCII, and the desired encoding for
strings is also PETSCII. No character conversion is done, so anybody
assembling ASCII files will get ASCII strings (which works out pretty
well if you're assembling code for a non-Commodore target). However,
the documentation says you're required to pass the "--ascii" flag when
the input is ASCII/UTF-8, so to build files that want ASCII operands
an explicit character encoding definition must be provided.</li>
</ul>
<h3 id="acme">ACME</h3>
<p>Tested versions: v0.96.4, v0.97
<a href="https://sourceforge.net/projects/acme-crossass/">[web site]</a></p>
<p>Bugs:</p>
<ul>
<li>The "pseudo PC" is only 16 bits, so any 65816 code targeted to run
outside bank zero cannot be assembled. SourceGen currently deals with
this by outputting the entire file as a hex dump.</li>
<li>Undocumented opcode $AB (<code>LAX #imm</code>) generates an error.</li>
<li>BRK is not allowed to have an operand.</li>
</ul>
<p>Quirks:</p>
<ul>
<li>The assembler shares some traits with one-pass assemblers. In
particular, if you forward-reference a zero-page label, the reference
generates a 16-bit absolute address instead of an 8-bit zero-page
address. Unlike other one-pass assemblers, the width is "sticky",
and backward references appearing later in the file also use absolute
addressing even though the proper width is known at that point. This is
worked around by using explicit "force zero page" annotations on
all references to zero-page labels.</li>
<li>Undocumented opcode <code>ALR</code> ($4b) uses mnemonic
<code>ASR</code> instead.</li>
<li>Does not allow the accumulator to be specified explicitly as an
operand, e.g. you can't write <code>LSR A</code>.</li>
<li>[Fixed in v0.97.]
Syntax for <code>MVN</code>/<code>MVP</code> doesn't allow '#'
before 8-bit operands.</li>
<li>Officially, the preferred file extension for ACME source code is ".a",
but this is already used on UNIX systems for static libraries (which
means shell filename completion tends to ignore them). Since ".S" is
pretty universally recognized as assembly source, code generated by
SourceGen for ACME also uses ".S".</li>
<li>Version 0.97 started interpreting '\' in strings as an escape
character, to allow C-style escapes like "\n". This requires escaping
all occurrences of '\' in data strings as "\\". Compiling an older
source file with a newer version of ACME may fail unless you pass
a backward-compatibility command-line argument.</li>
</ul>
<h3 id="cc65">cc65</h3>
<p>Tested versions: v2.17, v2.18
<a href="https://cc65.github.io/">[web site]</a></p>
<p>Bugs:</p>
<ul>
<li>PC relative branches don't wrap around at bank boundaries.</li>
<li>BRK can only be given an argument in 65816 mode.</li>
<li>[Fixed in v2.18] The arguments to <code>MVN</code>/<code>MVP</code> are reversed.</li>
<li>[Fixed in v2.18] <code>BRK &lt;arg&gt;</code> is assembled to opcode
$05 rather than $00.</li>
<li>[Fixed in v2.18] <code>WDM</code> is not supported.</li>
</ul>
<p>Quirks:</p>
<ul>
<li>Operator precedence is unusual. Consider <code>label &gt;&gt; 8 - 16</code>.
cc65 puts shift higher than subtraction, whereas languages like C
and assemblers like 64tass do it the other way around. So cc65
regards the expression as <code>(label &gt;&gt; 8) - 16</code>, while the
more common interpretation would be <code>label &gt;&gt; (8 - 16)</code>.
(This is actually somewhat convenient, since none of the expressions
SourceGen currently generates require parenthesis.)</li>
<li>Undocumented opcode <code>SBX</code> ($cb) uses the mnemonic
<code>AXS</code>. All other opcodes match up with the
"unintended opcodes" document.</li>
<li>ca65 is implemented as a single-pass assembler, so label widths
can't always be known in time. For example, if you use some zero-page
labels, but they're defined via <code>.ORG $0000</code> after the point
where the labels are used, the assembler will already have generated them
as absolute values. Width disambiguation must be applied to operands
that wouldn't be ambiguous to a multi-pass assembler.</li>
<li>Assignment of constants and variables (<code>=</code> and
<code>.set</code>) ends local label scope, so the label localizer
has to take variable assignment into account.</li>
<li>The assembler is geared toward generating relocatable code with
multiple segments (it is, after all, an assembler for a C compiler).
A linker configuration script is expected to be provided for anything
complex. SourceGen generates a custom config file for each project.</li>
</ul>
<h3 id="merlin32">Merlin 32</h3>
<p>Tested Versions: v1.0
<a href="https://www.brutaldeluxe.fr/products/crossdevtools/merlin/">[web site]</a>
<a href="https://github.com/apple2accumulator/merlin32/issues">[bug tracker]</a>
</p>
<p>Bugs:</p>
<ul>
<li>PC relative branches don't wrap around at bank boundaries.</li>
<li>For some failures, an exit code of zero is returned.</li>
<li>Immediate operands with a comma (e.g. <code>LDA #','</code>)
or curly braces (e.g. <code>LDA #'{'</code>) cause an error.</li>
<li>Some DP indexed store instructions cause errors if the label isn't
unambiguously DP (e.g. <code>STX $00,X</code> vs.
<code>STX $0000,X</code>). This isn't a problem with project/platform
symbols, which are output as two-digit hex values when possible, but
causes failures when direct page locations are included in the project
and given labels.</li>
<li>The check for 64KiB overflow appears to happen before instructions
that might be absolute or direct page are resolved and reduced in size.
This makes it unlikely that a full 64KiB bank of code can be
assembled.</li>
</ul>
<p>Quirks:</p>
<ul>
<li>Operator precedence is unusual. Expressions are generally processed
from left to right. The byte-selection operators have a lower
precedence than all of the others, and so are always processed last.</li>
<li>The byte selection operators ('&lt;', '&gt;', '^') are actually
word-selection operators, yielding 16-bit values when wide registers
are enabled on the 65816.</li>
<li>Values loaded into registers are implicitly mod 256 or 65536. There
is no need to explicitly mask an expression.</li>
<li>The assembler tracks register widths when it sees SEP/REP instructions,
but doesn't attempt to track the emulation flag. So if you issue a
<code>REP #$20</code>
while in emulation mode, the assembler will incorrectly assume long
registers. Ideally it would be possible to configure that off, but
there's no way to do that, so instead we occasionally generate
additional width directives.</li>
<li>Non-unique local labels should cause an error, but don't.</li>
<li>No undocumented opcodes are supported, nor are the Rockwell
65C02 instructions.</li>
</ul>
<h2 id="export-source">Exporting Source Code</h2>
<p>The "export" function takes what you see in the code list in the app
and converts it to text or HTML. The options you've set in the app
settings, such as capitalization, text delimiters, pseudo-opcode names,
operand expression style, and display of cycle counts are all taken into
account. The file generated is not expected to work with an actual
assembler.</p>
<p>The text output is similar to what you'd get by copying lines to the
clipboard and pasting them into a text file, except that you have greater
control over which columns are included. The HTML version is augmented
with links and (optionally) images.</p>
<p>Use File &gt; Export to open the export dialog. You have several
options:</p>
<ul>
<li><b>Include only selected lines</b>. This allows you to choose between
exporting all or part of a file. If no lines are selected, the entire
file will exported. This setting does <b>not</b> affect link generation
for HTML output, so you may have some dead internal links if you don't
export the entire file.</li>
<li><b>Include notes</b>. Notes are normally excluded from generated
sources. Check this to include them.</li>
<li><b>Show &lt;Column&gt;</b>. The leftmost five columns are optional,
and will not appear in the output unless the appropriate option is
checked.</li>
<li><b>Column widths</b>. These determine the minimum widths of the
rightmost four columns. These are not hard limits: if the contents
of the column are too wide, the next column will start farther over.
The widths are not used at all for CSV output.</li>
<li><b>Text vs. CSV</b>. For text generation, you can choose between
plain text and Comma-Separated Value format. The latter is useful
for importing source code into another application, such as a
spreadsheet.</li>
<li><b>Generate image files</b>. When exporting to HTML, selecting this
will cause GIF images to be generated for visualizations.</li>
<li><b>Overwrite CSS file</b>. Some aspects of the HTML output's format
are defined by a file called "SGStyle.css", which may be shared between
multiple HTML files and customized. The file is copied out
of the RuntimeData directory without modification. It will be
created if it doesn't exist, but will not be overwritten unless this
box is checked. The setting is <b>not</b> sticky, and will revert
to unchecked. (Think of this as a proactive alternative to "are you
sure you wish to overwrite SGStyle.css?")</li>
</ul>
<p>Once you've picked your options, click either "Generate HTML" or
"Generate Text", then select an output file name from the standard file
dialog. Any additional files generated, such as graphics for HTML pages,
will be written to the same directory.</p>
<p>All output uses UTF-8 encoding. Filenames of HTML files will have '#'
replaced with '_' to make linking easier.</p>
</div>
<div id="footer">
<p><a href="index.html">Back to index</a></p>
</div>
</body>
<!-- Copyright 2018 faddenSoft -->
</html>