@Database "LITTEL.guide" @Node Main "LITTEL r4" LITTEL Language Compiler v0.16 Public Release 4 @{" Introduction " link intro} @{" Requirements " link requ} @{" Installation " link install} @{" Feautures " link feautures} @{" The Instructions/Keywords " link keywords} @{" What the startupcode does " link startcode} @{" Some Rules " link syntaxrules} @{" Calling Subroutines " link callsub} @{" Using Data " link usingdata} @{" Currrent Bugs/Limitations " link buglim} @{" Copying/Using/Disclaimer " link disc} @{" About the implementation " link implement} @{" Examples " link examples} @{" In the feauture " link feauture} @{" History " link history} @{" The Author " link author} Copyright © Leif Salomonsson '99 @EndNode @Node intro "Introduction" Its a small and simple language, similar to assembler, but with some higher-level 'instructions' or keywords that make things like loops and conditional things much easyer. The simple integer-instructions also, compared to 68k-assembler, is easier to use, have common 'adressing-modes' and so on. The main idea from the beginning was to make a kind of cpu independant, but still lowlevel and fast language as a 'base-language' for building higher level languages on top of. For now it uses PhxAss/Lnk by Frank Wille to assemble executables for Amiga 68k. @EndNode @Node requ "Requirements" Amiga OS 2 + 68000 + (68010+ for PROCs with params) PhxAss 4 PhxLnk 4 Some knowledge about Asm/C Amiga Programming. Knowing what Objects and Includes are and so on.. @EndNode @Node install "Installation" Copy Litteldrawer whwrever you want. Assign Littel: to the drawer. @EndNode @Node feautures "Feautures" * The LITTEL Compiler currently supports 68000 and higher CPUs. * Stackbased and registerbased params. * PROCs in littel witch recieves params on the stack uses RTD (68010+) * Easy creation and use of standard objectfiles .o .lib * Easy creation and use of standard Shared Libraries. * Local and Global variables. * E-like typelessness. ..an E-frontend _could_ be made! * All kinds of variables and registers can be used directly as PTRs with optional offset. (ex D4[B].offset = the byte D4 points at + offset) * All instructions are INLINE. * Its small and simple. * Read on and try it. @EndNode @Node keywords "The Keywords (or instructions)" LITTEL is still in beta and anything may change! dvc = direct value, like : $FFFE/123/-2048/5/etc any = STACK+/-STACK/localVar/_globaVar/%parameter/Dx/Ax/localVar[]/_globalvar[]/%paramater[]/Dx[]/Ax[] **************************** *DATA TRANSFER INSTRUCTIONS* **************************** COPY source:any/dvc dest:any SWAP target1:any target2:any REGS2STACK [reglist] default is D1-D5/A0-A3 STACK2REGS [reglist] ************************* *ARITMETHIC INSTRUCTIONS* ************************* MUL with:any/dvc target/dest:any altdest:any DIV with:any/dvc target/dest:any altdest:any SHL steps:any/dvc target/dest:any altdest:any SHR steps:any/dvc target/dest:any altdest:any ADD source:any/dvc target/dest:any altdest:any SUB source:any/dvc target/dest:any altdest:any INC amount:dvc target:any DEC amount:dvc target:any ABS target/dest:any altdest:any NEG target/dest:any altdest:any MAX source1:any/dvc source2:any/dvc dest:any MIN source1:any/dvc source2:any/dvc dest:any LIMIT min:any/dvc max:any/dvc target:any ******************** *LOGIC INSTRUCTIONS* ******************** BGET bitnum:any/dvc source:any/dvc dest:any (0 or <>0) BCLR bitnum:any/dvc target/dest:any altdest:any BSET bitnum:any/dvc target/dest:any altdest:any BCHG bitnum:any/dvc target/dest:any altdest:any OR source:any/dvc target/dest:any altdest:any XOR source:any/dvc target/dest:any altdest:any AND source:any/dvc target/dest:any altdest:any NOT target/dest:any altdest:any ******************* *MISC INSTRUCTIONS* ******************* DOPROC name <%param %param pa...> CALL address:any/dvc offset:dvc PADR procname dest:any RETURN list of any/dvc PUTPRMS p1:any/dvc p2 p3....p19 REMPRMS nrof:any/dvc ******************* *LABELING KEYWORDS* ******************* PROC name:name [%param1 %param2 ....] GVAR _name:first letter '_' // global var STRING _name "this is a string!" // null-terminated string LBLK/WBLK/BBLK _name nrofelements:dvc // long/word/byte - block // uninitialised // l/w/b - block // initialised LONGS/WORDS/BYTES _name listofelements: ex. 10,-16,54,32,100 Inside PROC : VAR var [var2 var3...var19] ****************************** *CONDITIONAL/LOOPING KEYWORDS* ****************************** ?? = EQ/NE/GT/LT/GE/LE IF any/dvc ?? any/dvc ELSE // no letting this one out ! ENDIF SELECT Dx // only Dx can be used here! may change CASE any/dvc CASE any/dvc C... DEFAULT // no letting this one out! ENDSELECT WHILE any/dvc ?? any/dvc ENDWHILE REPEAT ENDREPEAT any/dvc ?? any/dvc ****************** *SUPPORT KEYWORDS* ****************** OPENLIBRARY anylibrary.library ver:dvc globaltoputbasein CLOSELIBRARY libbase:any **************************************** *MISC COMPILER CONTROL/SUPPORT KEYWORDS* **************************************** #equ Constantname value:dvc #xref name [name name name name..] #xdef name [name name name..] #incdir "path" #include "file" ["file" "file" "file"...] // include asm .i type files with constants #mode EXE (default) #mode OBJECT #mode LIBRARY ver rev name idstr #linkdir "path" // not ready, LITTEL searches LITTEL:lib/ for now #link file.o/file.lib [file2 file3...] link with theese files.. #insert "src.l" // not ready yet! inserts littel src at spot. #nostartup // 15b : dont link in startupcode, know what youre doin :) @EndNode @Node startcode "The startup code of LITTEL v0.14b does :" Startup Code Opens Theese Global Vars : _ExecBase _DOSBase _GraphicsBase _IntuitionBase _UtilityBase _stdin _stdout _arg @EndNode @Node syntaxrules "Some Rules" LITTEL v0.14 The only allowed commenting is commenting preceeded by '//' // this is a comment EVERY single line of code starts with an instruction or keyword! Only ONE keyword/instruction is allowed per line ! Local Vars must begin with a-z Global Vars must begin with _ Stackparameters (received) begins with % Dataregs begins with D followed by 0-7 AdrRegs begins with A followed by 0-7 Constants must begin with A-Z Direct Values : %01010, 2000, -3000, $FFFe, -$19, etc.. PROC name. name can be anyname anycase. The first PROC in source is the 'main' procedure VAR declarations in PROC must end with ENDVAR PARAM declarations in PROC must end with ENDPARAM Dont forget the last END in the source ! PTRs [] : With no character between the brackets its treated as a PTR TO LONG. Valid PTR sizes : L or W or B (LONG, WORD, BYTE) ex. myptr[B] Illegal : mywindow[w].height Legal : mywindow[W].height Ptrs can have offsets added to them : myptr[].offset offset must be direct value or constant. NO SPACEes in any names! NO SPACEes in following : mywindow[W].height this is illegal : mywindow[ W ].height mywindow[] .height etc.. Between keywords and arguments there can be as much spaces you want. Calling a PROC with parameters : DOPROC procname any/dvc any/dvc any/dvc ...... etc. Max 7 params (for now)! ex. DOPROC myproc 1 2 A0 D4 _myGlobal Add FLUSH after the last parameter if the routine you are calling doesnt cleanup the parameters on stack. The register A4 and A5 is used internally by LITTEL. NO TOUCHING! Also registers D6, D7 and A6 is used for scratch by various instructions. So dont put anything in them an expect it to be there :) See examples for more information. @EndNode @Node callsub "Calling Subroutines" A subroutine in LITTEL looks like : PROC name [VAR bla] [CODESTART] ENDPROC If using callbackhooks from other code then Littel u must get the A4 register if you want to use any globals or librarybases. Do it with GETA4 first thing in PROC after (any) variables or params declarations. There are two ways of calling in LITTEL. Using DOPROC or CALL 1. DOPROC procname like BSR in assembler, it branches to a labeled subroutine. You can put params after the procname. All PROCs created in LITTEL automatically flushes the params on stack, (if any) by using RTD. So there is a special DOPROC-keyword, 'FLUSH' that you must put after the params when calling external #imported functions if they do not flush the params by them selfes. Parameters are put on the stack in _reverse_ order. (C-style ?) No registers are saved by default. If external functions could trash A4 or A5 then you must save them first! By using the instruction 'PADR procname dest' you can get the adr of a labeled subroutine. Example of use : callbackhooks. 2. CALL address offset This one you will likely use for calling libraries. address can be anything, register or ptr or variablel etc. offset can be a direct value or constant. If not calling libraries but just an address you can put just 0 as offset. ;The internal registers A4/A5 will get saved before CALLing. NOT ANYMORE! (v0.16) Registers D0-D5/A0-A3 can ofcource also be used for parameterpassing at any time! @EndNode @Node usingdata "Using Data" Variables All variables in LITTEL are of the 32bit LONG Type (signed). Definition Local to PROC VAR name [name name...] VAR name V.... Global outside of PROC GVAR _name [_name _name...] Registers D0 D1 D2 D3 D4 D5 A0 A1 A2 A3 (you shouldnt touch D6/D7/A4-A7) Variables as PTRs PTR TO LONG myvar[] D0[] _global[] A0[] %param1[] myvar[] D0[L] _global[L] A0[] %param1[L] ..etc PTR TO WORD, BYTE anyvariable[W] anyvar[B] D3[W] ..etc.. PTRs with Offset myptr[L/W/B].Offset mywindow[W].w_Heigth myptr[B].1200 myptr[L].-32000 etc.. Stack STACK+ -STACK Example COPY STACK+ D0 COPY D0 -STACK Should be of rare use. Blocks Currently in LITTEL all blocks of data are Global. Therefore all block names must start with '_'. BlockTypes Initialized STRING _name 'string!' LONGS _name val,val,val,val,val.... WORDS BYTES UnInitialized LBLK _name nrofelements WBLK BBLK Example STRING _my "mooo" COPY ptr[B] D0 // get first char ('m') from my to D0 @EndNode @Node buglim "Current Bugs / Limitations" Very little errorchecking! Probably will not fix this.. (Should be the task of Language-Frontends to produce ok code) SELECT crashes compiler sometimes ABS producing bad code If u dont give lc any arguments it crashes Also be sure to type lc15 namewithout.l or it crashes! something else ? @EndNode @Node examples "Examples" // Hello World in LITTEL :) #include "dosfuncoffsets" // change to your config.. STRING _hello "hello world!\n" PROC main COPY _hello D1 // address of our hellostring in D1 COPY 13 D2 // len of string to D2 CALL _DOSBase WriteChars // print it ENDPROC END // dont forget this one ****** OR ************************* // #link amiga.lib // for example... #xref _puts STRING _hello "hello world!" PROC main DOPROC _puts _hello FLUSH // print it ENDPROC END ***** lotsa ways of doin things :) ***** Please look into the Examples drawer. @EndNode @Node history "History" v0.11b *********************** introduced Global vars added FUNC, DOFUNC added some security checks.. 1:st release to Aminet *********************** v0.12b *********************** 991119 new naming stuff : local vars : first two letters lowercase, a-z constants : first letter uppercase A-Z global vars : first letter '_' FUNC/PROC/STRING, BLK, DC names : anycase, anything new ADR syntax : ADR STRING/BLK/DC/FUNC/PROC name xx string_name: proc_name: func_name blk_name: dc_name: done! :) -> even better : STRING, PROC, FUNC, LBLK, WBLK, BBLK, LONGS, WORDS, BYTES globals working added SWAP instruction now u can have the same name of local vars in diffrent PROCs. found bug that didnt allow '=' anywhere in src. extremely experimental ELSE ! hell.. not working..discarding it. Added OPT LARGE in sources nnow ELSE is working! New format for IF..: IF ELSE -> must be there! ENDIF added DEFAULT in SELECT block.. DEFAULT MUST be there! changing CALL name offset -> CALL adr offset v0.13b ******************************** date ?? getting rid of FUNC just PROC if a VAR is declared i a PROC then we add the link/unlk instructions.. that simple! but.. we need ENDVAR PROC nisse VAR ulla VAR albert ENDVAR done! optimised PROC/ENDVAR/ENDPROC litle... made copy2.e simpler but not as optimised.. but that can be done later.. for now it works fine atleast.. bugfixed some in copy 2.. thinking of adding support for parameters to PROCs using rtd #paramsize isteda of rts, IF params are used. HOPEFULLY.. @param support is now added! :) 2:nd release to aminet ********************************* v0.14b *********************************'' date ?? lc14 : removing parser, using my own class instead. extractwords.e adding new 'parametermodes' added ABS, NEG, MAX, MIN, LIMIT added support for external PROCs/subroutines with no own stackcleanup! like powerD.lib for example.. use FLUSH as last param. changed names of STORE_REGS, RESTORE_REGS to REGS2STK, STK2REGS well its still beta! :) things to do : possibly implement "ABCD" -> value just remebered.. think the ABS wasnt working right.. extractwords.e now handles '//' by it self :) commented out #export and #mode. no use for them yet anyway. added returnvalues after ENDPROC. values in D0/D1/D2....etc added RETURN instruction.. put optional values after it. created dynamic stringstack, lc14 now only writes to _one_ file. (speedup) this will probably later meen that SELECT can handle _any_ type instead of Dx. SELECT now handles any type Complete rewrite of the COPY instruction (witch other instructions uses) much smarter and efficient. v0.15b ************************************* Dec 8 99 DEFAULT in SELECT block no longer must be there. Added CONT, an alternativ to CASE. CONTinue with this... Added #nostartup keyword. guess what it does... pretty dangerous to use this switch right now! (the smalldata isnt initialized) doesnt use the stringstack for tempdata anymore but its now alot faster then before, on the bad side : more memory is required and memory could get overwritten if my very nasty memory-use-prediction should fail :( Strange bug in something.. :) the compiler hangs just before exiting. code generation is ok... Added GETA4. Handy when getting called from other code (callbackhooks). Fixed the strange bug. Bug in EC it seems.. Now im not ENDing some objects at programend.. One dont have to anyway.. but it shouldnt crash. Changed 'ADR' into ADR and PADR, for global datablocks and PROCedures respectively. Fixed the #mode keyword. Now you can create objects .o ! :) Use '#mode OBJECT'. This automatically sets #nostartup also. #export is back! guess why.. Changed #export, #import -> #xdef, #xref :) Optimised COPY instruction to the maximum! Optimised Stackrestoration when calling external procedures using FLUSH. MOVE.L (A7)+, D7 times nrofparams -> LEA nrofparams*4(A7), A7 :) SELECT still buggy, compiler crashes... :( 3:d release to Aminet ********************************* Ohyeah, just found out that FLUSH doesnt work, forgot to write the code to tempstrings.. :) it didnt hang in the test.l example so who could know ?:) *Fixed* Also forgot to mention in the .readme that LITTEL now requires PhxLnk.. :( Added New Targets : -STACK and STACK+ Added 'FLSHSTK nrofparams' instrcution. flush out nroflongs from stack. Changed em back again :) SAVEREGS, LOADREGS -> REGS2STACK, STACK2REGS #mode LIBRARY doesnt seem that far away now.. :) Just did my first library by hand, with 99% help from aminet/dev/src?/LibrariesGuide.lha v0.16a ********************************************* Dec 15 99 Removing ADR Changing internal structures (global initialized blocks in code-section instead of data-section adding vars for the blocks, vars in data-section..) Changing CALL, OPENLIBRARY, CLOSELIBRARY on to more changes.. : Scrapping PARAM..ENDPARAM and ENDVAR! instead params go like this : PROC myproc %param1 ¶m2 %param3.. etc if this is gonna work we'll have to welcome the old PROCCODE back! :) maybe change its name.. CODESTART.. must be in _every_ PROC..! (that uses either locals or params!) see examples/parvar.l PROC blabla [%params...] [declarations] CODESTART Changed Names of the startup-globals : see 'What the startupcode does'. Removed automatic includes for dos and exec! #incdir working!! Changed internal outputbuffers.. MUCH lesser memory used 4k is allocated regardless of src-size, compared to the old style that allocated ~ src-size * 4 bytes. no risk of overwriting memory. Removed some not-dangerous enforcerhits. (Gotta look good) Created librarymodule.e LITTEL now compiles Shared Libraries! not tested the functionality yet and there is no support for any global stuff, yet. Libraries tested and working! (still no globals) I fiddled with it so long and then it was an error in the docs I used.. :/ The doc txt said 'return libbase in D0 in the initfunction..' Then in the doc example, wich I followed, NIL was returned.. Now I am trying to pass stackparams to the functions. Why doesnt that work ? It should! I almost forgot.. one have to do a '#link library.lib' in the library src now. The library.lib contains internal functions of the library. that way the compiler get smaller and faster + easy changing of the libraries behaviour by just using another 'libraries.lib'. rajrajraj.. *singing* new keywords for libraries: #mode LIBRARY ver rev name idstr #fdef FunctionName [additional info] ** #endfdef basename Added functionality to #xdef, #xref, #include, VAR, GVAR they now accepts multiple arguments.. Changed name and functionality of #const to #equ ex. #equ MyConstant 10 Increased the maximum nr of stackparams to send with DOPROC to function : 15! @EndNode @Node feauture "What the feauture might bring" * Higher Level Language Preprocessors as plugins. This could be your job! :) contact me if youre interested. * LITTEL as a shared library. Different language frontends calls the librarys functions to create code.. * Probably caseinsensitive keywords. * Maybe PPC support, its just about money. As PPC would fits this language perfecly it would rock! :) * Exception handling...not sure how to implement it right now. * Local blocks of data (not just LONG-vars) * #mode LIBRARY *DONE* :) * Real Optimising for 020+ * xxx[index:Size]. * suggestions ? * Betatesters I hope.. @EndNode @Node author "The Author" suggestions, bugreports or whatever leif_salomonsson@swipnet.se @EndNode @Node implement "The implementation" LITTEL from v0 to this release, v0.15b, is/was written completely in E. It uses a very simple one-pass technique. Just wandering from the top line to the last line, parsing the instructions and their arguments using my own parser routine. some versions ago, I used afc/parser for that. The asm-strings from conversionroutines is put into 4 precalcedsized stringbuffers. At the end, the buffers are written to the asmfile in a special order. The file gets assemled by PhxAss and Linked by PhxLnk. All code uses small-data. The executable is split into 2 hunks : code and data-bss. @EndNode @Node disc "blablabla" its freeware anyway but i dont want people hacking it and releasing their own Littels.... talk to me instead. AND.. I will not take ANY responsibility for ANYTHING that migth go wrong when/after using this compiler. its all AT YOUR OWN RISK! The archive may be distributed (untouched) in any way as long as no money is earned doing it. @EndNode