Revision history for CPU-Z80-Assembler

2.13 2010-11-21

        Bug Fixes
        * Bug detecting circular references: was giving false positives 
          in the following example code:
                        err equ 10
                        y0  equ err
                        ld a,(iy+err-y0) ; gives circular reference error, and should not

        New Features
        * Add defm, defmz, defm7, equ
        * Add z80asm_file function.

        Feature Changes
        * Replace AsmTable by Asm::Z80::Table. This has the impacts:
          - accept 'jr <po|pe|p|m>,NN', compile 'jp ...' instead
          - no longer support for variants of ixh, ixl, iyh, iyl (e.g. xh, hx, ...)
          - no longer support variants db, dw, dt, dm of defb, defw, deft, defm
          - no longer support for [] instead of ()
          - no longer support for exa

        Other
        * Merge CPU::Z80::Assembler::Preprocessor and CPU::Z80::Assembler::Lexer
          into CPU::Z80::Assembler, they were just wrapper functions on other modules.

2.12 2010-10-02

• Sequence (?|...) not recognized in regex in Perl 5.8

2.11 2010-10-01

• Make CPU::Z80::Assembler::Lexer as a wrapper to Asm::Preproc::Lexer, use Asm::Preproc::Token instead of CPU::Z80::Assembler::Token
• Solve #59195 (use hex instead of eval)

2.10 2010-09-12

• Make CPU::Z80::Assembler::Preprocessor as a wrapper to Asm::Preproc
• Error messages stopped showing actual text of line containing the error - cluters the display
• Use L<zzz|zzz> instead of L<zzz> in POD to avoid "the zzz manpage" rendering

2.09 2010-03-23

• bump version

2.08_01 2010-03-22

• CPAN cannot unpack tarball, permission denied; give full control to Win7 source files before make dist, add noacl to cygwin fstab

2.08 2010-03-21

• Failed t/JumpOpcode.t test 36 in perl 5.11.5, operation "eq": no method found when comparing two CPU::Z80::Assembler::Line; convert to string before passing on to is()

2.07 2009-11-12

• One test of t/Preprocessor-include.t failed on localized version of $!

2.06 2009-10-26

• Bump up version number

2.05_06 2009-09-22

• Move back to 5.008, CPAN of 5.006 did not compile, Regexp::Trie requires 5.008
• '$' no longer used as label prefix, to be able to support $FF syntax
• '%' (bin), '$' (hex), '#' (hex) number prefixes
• removed ORG label - does not make sense with several segments
• Rename link() to _locate() : only address location is done; make it private, called by bytes()
• Remove restriction on ORG - each new org creates a new segment, gaps between the segments are filled with 0xFF (configurable)
• Change 'JR' to 'JP', and 'DJNZ' to 'DEC B:JP NZ' if distance too far
• Publish AsmTable.pm - usefull for disassembler

2.05_05 2009-08-26

• Fix Parser.pm version
• Remove dead code
• Improve documentation
• Tests failed in Unix due to lack of \r processing
• Compatible with perl 5.006, downgraded requirement from 5.008.
• Error in macro expansion : was eating statement terminator

2.05_04 2009-08-23

• New Stream class - faster than HOP::Stream, because head element is always available to be read without computation (is_promise() in HOP::Stream) and no test is needed to distinguish a stream node from a array-ref element (is_node() in HOP::Stream). The stream has an head element, and a list of the next elements, which may contain iterators.
• List object uses the input to the assembler to read the lines again during the listing phase, instead of globing all tokens in memory
• Remove usage of HOP::Stream
• Move the macro processing into the parser - removes one level of stream processing
• Added the unofficial Z80 opcodes to the test cases
• Benchmark assembly times: 0.337 sjasmplus 2.734 z80masm-2.01 2.738 z80masm-2.05 2.898 z80masm-current 3.555 z80masm-2.05_03 11.342 z80masm-1.03

2.05_03 2009-08-11

• Performance :
  • hand-coded contructor and accessor for Line.pm and Token.pm instead of Class::Struct, because too much time was being spent on Token->new()
• Remove Class.pm : plugin replacement to Class::Class to solve RT #42710
• Remove Node.pm : cannot use inheritance with Class::Struct; define different modules (Program.pm, Segment.pm, Opcode.pm, Expr.pm) with the same interface (child contains list of children of this node).
• All test scripts now pass
• Move test scripts data files to t\data
• Move all bechmarking scripts to tools\benchmark
• New List.pm class to handle output of the assembly file listing at the code generation stage.
• New ParserGenerator.pm class to generate a try-based non-backtracking parser. Builds a complete state table for the input grammar, where each state is a map of current-token => next-state, or current-token => accept-rule, or current-token => call-sub-rule. The input grammar accepts rule => token token ... [sub-rule] ... action The sub-rule call accepts optional ('?'), 0 or more ('*') or 1 or more ('+'). The output is the new Parser.pm module with about 7K states and 20K lines of Perl code --> PERFORMANCE PENALTY.
• Remove ParserTable.pm : whole Parser.pm is now generated by tools\build_Parser.pl
• Expr.pm parser now checks for the correct syntax. The previous parser just extracted balanced parentheses.
• Lexer.pm :
  • accept db, dw, dt, dm as alias to defb, defw, deft, defm
  • remove quotes from STRING token value
• Line.pm : added back is_equal() and is_different()
• Macro.pm :
  • return the list of newline tokens parsed inside of a macro definition back to the stream, so that the assembly listing contains also these lines.
  • replace statement_end() by /[:\n]/ -> performance
• Program.pm, Segment.pm : new org() and add() methods to be called by code generator.

  The lexer is still the performance botleneck, together with the heavy usage of HOP::Stream. %Time ExclSec CumulS #Calls sec/call Csec/c Name 43.7 0.500 3.052 71417 0.0000 0.0000 HOP::Stream::tail 33.6 0.385 0.665 98780 0.0000 0.0000 HOP::Stream::head 24.6 0.282 0.466 170197 0.0000 0.0000 HOP::Stream::is_node 22.4 0.257 1.214 17902 0.0000 0.0001 CPU::Z80::Assembler::Lexer::__ANON__ 21.3 0.244 0.257 67163 0.0000 0.0000 HOP::Stream::is_promise 20.7 0.237 2.173 34954 0.0000 0.0001 CPU::Z80::Assembler::Macro::__ANON__ 20.4 0.234 0.250 9 0.0260 0.0278 CPU::Z80::Assembler::Macro::BEGIN 19.6 0.225 0.225 242140 0.0000 0.0000 UNIVERSAL::isa 18.6 0.213 3.250 71417 0.0000 0.0000 HOP::Stream::drop

2.05_02 2009-04-25

• Line.pm, Token.pm :
  • go back to Class::Struct instead of Class::Class (bad performance of the later)
  • do not use "eval(Data::Dump::dump())" for clone - bad performance
  • add overload for 'bool' and '0+' - '""' was being used, and is too slow because it uses Data::Dump::dump() internally
  • remomve is_equal() and is_different() methods (Line.pm)

2.05_01 2009-04-23

• Add Line, Token classes. No need to create LINE tokens - each token carries the line number.
• Change to single-pass design.
• Use Class::Class hierarchy for parsed objects.
• Several test scripts fail.
• Documentation needs update.
• Dead code needs to be removed.

2.05 2009-01-04

• Fixed VERSION file - wrong version parsed by CPAN.

2.04 2009-01-04

• Added MANIFEST.SKIP. Created VERSION file.

2.03 2009-01-04

• Fixed MANIFEST - missing files.

2.02 2009-01-04

• Support undocumented Z80 instructions (t/test_z80.asm)
• Auto-generate ParserTable.pm and test_z80.asm, cross check the assembly with the sjasmplus assembler.
• Fix documentation and show usage when called without arguments.
• Integer out of range is now a warning instead of a fatal error. "+d" in "(ix+d)" is optional.

2.01 2008-07-13

• 10-lexer.t and 11-z80pp.t tests failed in Unix.

2.00 2008-07-13

• Re-implemented for speed improvement.
• Split into three layers - lexical scanner, instructions parser and code generator.
• Added Z80 undocumented instructions.
• $ in front of labels is optional, accepts lower case assembly instructions, improved error messages, accepts multiple assembly instructions in a line separated by ':'

1.03 2008-06-25

• Bugfixes from Paulo Custodio: see https://rt.cpan.org/Ticket/Display.html?id=36991

1.02 2008-06-22

• Added INCLUDE
• Added support for #comments at start of line (for C pre-processor friendliness)

1.01 2008-06-14

• Macro parameters no longer clash with labels whose names start the same as a parameter. eg, a param $r no longer clashes with label $rr
• Likewise labels

1.0     2008-06-13 
        * Original release