a56 - a DSP56001 assembler - version 1.1 /* * Copyright (C) 1990-1992 Quinn C. Jensen * * Permission to use, copy, modify, distribute, and sell this software * and its documentation for any purpose is hereby granted without fee, * provided that the above copyright notice appear in all copies and * that both that copyright notice and this permission notice appear * in supporting documentation. The author makes no representations * about the suitability of this software for any purpose. It is * provided "as is" without express or implied warranty. * */ V1.1 CHANGES Many thanks to Tom Cunningham at Motorola who noted several bugs, fixing most of them. Thanks also to Stephen Edwards at CalTech for his fixes as well. Both he and Tom fixed up the XY parallel move code; I believe all the fixes are integrated. Summary of changes since v1.0: - Jcc, JScc, Tcc recognizes "CC" synonym for "HS" (Tom) - LUA syntax fixed (Tom, Stephen) - DIV syntax fixed (Tom) - XY move fixed (Tom, Stephen) - a56 now detects phase errors (changes in symbol values between pass 1 and 2) - several causes of phase errors fixed - immediate value handling improved for both integer and fixed-point values - symbols have type (INT or FLOAT) - cpp droppings are properly handled - ALIGN pseudo-op added Known bugs/deficiencies: - MOVE to a control reg generates bogus code (always use MOVEC, MOVEM, and MOVEP) The example code has been cleaned up a bit and I've included my 56K board's envelope code - tdsg.basic.a56. This code is specific to my implimentation. A note on the reverb code - I have since found since posting version 1.0 that the reverb code has a great-big almost-DC bumb in its impulse response. I found this after doing some simulations. I suspected something was wrong because I had to scale the output of the comb filters by .046 or something before going into the two all-pass stages. Overall I think it works better without the 1.7 ms all-pass. There's another reverb example on Dr. Bubb that fits Moorer's 1979 model better. I've included this in the examples but I have not yet assembled or tried it. OVERVIEW This program was written as a vehicle to learn the intricacies of the DSP56001 instruction set, and to provide a tool for Unix-based DSP code development (for those of us without a NeXT machine.) The intent was to provide compatibility with Motorola assembler's syntax. But since the author did not have Motorola's assembler or its documentation, it is no doubt far from compatible. Only a few pseudo-ops are implemented-- probably only partially. Macros are not supported, except through the use of an external macro preprocessor, such as /lib/cpp. To facilitate cpp macro expansion, multiple assembler statements on a single input line are delimited with an '@', e.g.: #define JCHEQ(c,label) move #c,x0 @cmp x0,a @jeq label #define JCHNE(c,label) move #c,x0 @cmp x0,a @jne label SUPPORTED PSEUDO-OPS The following is a list of the pseudo-ops that are recognized: = ;assign a symbol ALIGN ;set location pointer ;to next integral ;multiple of ORG ;new location pointer ORG , DC ;declare constants PAGE , , , ;ignored INCLUDE ;file inclusion END ;end In addition, a "PSECT" pseudo-op was implemented. It allows program sections to be defined and bopped in and out of, each with its own location counter and space. The Motorola assembler probably does not have this pseudo-op, but no doubt supports the concept in some way. PSECT : ;define PSECT ;switch to psect FUTURE DIRECTION The assembler probably generates bogus code here and there, and no doubt does not handle all of the syntax. I welcome all comments, fixes and enhancements. TO MAKE AND USE Type "make". gram.c and lex.yy.c are provided for those without yacc (or bison) and/or lex. The resulting program, a56, is used as follows: a56 [-b] [-l] [-o output-file] file [...] An assembler listing is sent to the standard-output and an ascii-formatted object file (a56.out) is produced. The "-b" option adds binary to the listing. "-l" causes included files to be listed. "-o" directs the output to the specified file rather than the default, a56.out. A separate program, toomf, converts a56.out into "OMF" format suitable for downloading to the 56001 via the sloader.a56 program. toomf < a56.out > file.omf AUTHOR 11/28/91 v1.1 8/6/92 Quinn C. Jensen 1374 N 40 E Orem, UT 84057 home: jensenq@zdomain.com (preferred address for a56 correspondence) work: jensenq@novell.com