AS10k1 Assembler version A0.99 ------------------------------ This is an assembler for the emu10k1 DSP chip present in the creative SB live, PCI 512, and emu APS sound cards. It is used to make audio effects such as a flanger, chorus or reverb. Author: Daniel Bertrand <firstname.lastname@example.org> This version of the assembler was modified for the ALSA driver by Jaroslav Kysela <email@example.com>. Usage: as10k1 <asm file> [bin output file] ------ Making binary DSP programs: example, type: ./as10k1 chorus.asm (it creates chorus.bin) -- Loading Binary DSP files <..TODO..> Description of files included ------------------------------ Chorus.asm -- chorus effect Flanger.asm -- flanger effect Delay.asm -- generates Echos, this is _not_ a reverb eq5.asm -- A 5 band Equalizer (needs a bit more work) fir.asm -- A low pass filter (A demo of a fir filter implementation) sine.asm -- A sinewave generator (can be useful for debuging) vibrato.asm -- A vibrato effect (or is it a tremolo?) vol_ctrl.asm -- provides support for the hardware volume control emu_constants.asm -- not an effect, just contains handy constants and macros =============================================================================== Programming Usage: ================= Assembly Syntax --------------- Assembly lines generally have four fields seperated by spaces or tabs: Name_Field Opcode_field Operand_Field Comment_Field ---------- ------------ ------------- ------------- [symbol] [mnemonic] [operands] [text] With this assembler, each line can have a maximum of 256 characters and each symbol can be a maximum of 32 characters. Symbols ARE case sensitive, opcodes ARE NOT. OPCODES -------- All instructions require 4 operands, they have the format <opcode> R,A,X,Y (note some documentation out there the call the R operand as Z and the A operand as W). Here are 16 opcodes. 0x0 (MACS) : R = A + (X * Y >> 31) ; saturation 0x1 (MACS1) : R = A + (-X * Y >> 31) ; saturation 0x2 (MACW) : R = A + (X * Y >> 31) ; wraparound 0x3 (MACW1) : R = A + (-X * Y >> 31) ; wraparound 0x4 (MACINTS) : R = A + X * Y ; saturation 0x5 (MACINTW) : R = A + X * Y ; wraparound (31-bit) 0x6 (ACC3) : R = A + X + Y ; saturation 0x7 (MACMV) : R = A, acc += X * Y >> 31 0x8 (ANDXOR) : R = (A & X) ^ Y 0x9 (TSTNEG) : R = (A >= Y) ? X : ~X 0xa (LIMIT) : R = (A >= Y) ? X : Y 0xb (LIMIT1): R = (A < Y) ? X : Y 0xc (LOG) : ... 0xd (EXP) : ... 0xe (INTERP) : R = A + (X * (Y - A) >> 31) ; saturation 0xf (SKIP) : R,CCR,CC_TEST,COUNT Special note on the accumulator: mac* instruction with ACCUM as A operand => uses Most significant 32 bits. macint* instruction with ACCUM as A operand => uses Least significant 32 bits. For more details on the emu10k1 see the dsp.txt file distributed with the linux driver. Operands -------- Operands can be specified as either a symbol or a value. hex values are prefixed by $, octal by @, and binary by %. e.g.: 123 decimal value $123 hex value @123 octal value %01101 binary value The operands for emu10k1 instructions are always addresses of registers, there are no instruction which take immediate values. Operands currently support basic arithmetic, It does not support bedmas (or is it bodmas) so don't try to use (). Infact don't put spaces either (for now, until I fix this). Summary of assembler directives ------------------------------- NAME "string" ;give a name to the patch <symbol> IO ;defines an Input/output pair <symbol> CONTROL <symbol> ;defines a controlable GPR <symbol> DYNamic <number of storage spaces> ;defines a temporary GPR <symbol> STAtic <initial value> ;defines a constant GPR /w initial value <symbol> EQU <Value equated> ;assembly time constant <symbol> CONstant <value> ;defines a read-only GPR <symbol> DELAY <value> ;defines a Delay line <symbol> TABLE <value> ;defines a lookup table <symbol> TREAD <tram id>,<value> ;defines a tram read <symbol> TWRITE <tram id>,<value> ;defines a tram write INCLUDE <"file name"> ;includes an external file FOR <variable>=<start>:<finish> ;Assembly-time 'for' statement ENDFOR ;ends a for loop <symbol> MACRO arg1,arg2,arg3.... ;used for define a macro ENDM ;end a macro definition END ;ends the code Detailed description of directives: ---------------------------------- ( <> brackets indicate required fields,  brackets indicate optional fields) DYNamic directive (replaces DS): Defines a storage space from the gpr pool on the emu10k1. The assembler maintains a pointer to the gpr registers (starting at $100). The symbol is assigned the value of the address of the gpr pointer. The pointer is increment by the number following the dynamic directive. syntax: <symbol> dynamic <number of storage spaces or <symbol> dyn <number of storage spaces> -- STAtic directive (replaces DC): Similar to dynamic, but places an initial value in the memory location. The values specified are slightly different from operands for instructions. The values are 32 bit signed intergers so that a maximum magnitude of 2^31 can be stored. values can be in signed decimal, unsigned octal, binary and hex, and in fractional decimal (values between -1 to 1) for filter coefficients. A fractional decimal is specified using the '#' prefix and can include an exponent. These values should be used with the fractional "mac" instructions. NEW! fractional numbers are now handle automatically, a value between 1 and -1 will be converted into fractional form. The old # form still works though. (BUG:confusion occurs at 1 and -1 however, should 1 be represented as $1 or $7ffffff?, currently defaults to $1, so #1 still has some importance) examples: .03412 123E-3 #-0.1236 syntax: <symbol> static <initial value> or <symbol> sta <initial value> -- CONTROL Control registers are similar to DC, but they also include a min and max value. The control register is used by a mixer app to change values in a GPR (a volume control, for example). syntax: <symbol> CONTROL <initial value>,<min>,<max> -- IO Defines an input and an output register. <symbol> IO It defines two register, but they both use the symbol. The assembler handles it automagically depending on whether you're performing a read (X, Y or Z operand) or a write (R operand) to the GPR. - If you insist on having two different symbols for read/write (for readability or whatever), use an EQU, i.e.: IN IO OUT EQU IN - To force a read from the output (for whatever reason) use <symbol>.o (i.e. OUT.o) Writing to an input is not allowed. -- CONSTANT defines a read-only constant GPR When the assembler encounters a CONSTANT define, it'll try three things. First it'll check to see if the defined constant is a hardware constant, if so substitutes that instead. Next the assembler check to see if another constant has alrady been declared with the same value, if so it'll substitute it. Else it'll declare a new GPR for holding the value of the constant. syntax: <symbol> constant <value> or <symbol> con <value> -- DELAY LINES Delay lines are defined via three directives: -- DELAY Directive Define Delay, used for allocating an amount of TRAM for a delay line. <symbol> DELAY <value> The symbol is used to identify this delay line.The value is the amount of TRAM allocated, it may be specified as a decimal,hex, octal, binary or time value. The time value is prefixed with '&' and represents seconds of time. e.g. foo DELAY &100e-3 ;;a 100msec delay line bar DELAY 1000 ;;a 1000 sample delay line -- TABLE directive Define lookup Table same as DELAY but for lookup tables. -- TREAD Directive Define read: used for defining a TRAM read point <symbol1> TREAD <symbol2>,<value> The value represents the read point within the delay line. symbol2 defines which delay line this read belongs to. Symbol1 is a pointer to TRAM data register associated with this TRAM read operation. The assembler will create <symbol1>.a which points to the TRAM address register. example: fooread TREAD 100e-3,foo macs fooread.a,one,two,three ; writes a new tram read address macs temp,fooread,one,two ; reads the data from the delay line -- WRITE Direcive Define write: same as TREAD but used for writing data to a delay line. <symbol1> TWRITE <symbol2>,<value> -- EQU directive: Equates a symbol to a be constant which is substituted at assembly time: syntax: <symbol> EQU <Value equated> -- END directive The END directive should be placed at the end of the assembly source file. If the END directive is not found, a warning will be generated. All text located after the END directive is ignored. Syntax: [symbol] END -- INCLUDE Directive The include directive is used to include external asm files into the current asm file. Syntax: INCLUDE <"file name"> The file name Must be enclosed in "" or '' . examples: include 'qwerty.asm' include "foobar.asm" -- MACRO directive Used for defining a macro Defining Macro: <symbol> macro arg1,arg2,arg3.... .... <opcode> arg4,arg1,arg2... ;;for example .... .... endm were the <symbol> used is the nmeumonic representing the macro. arg1,arg2,arg3... can be any symbols (auto-defining and local to a macro) as long as the symbol is not already in use outside the macro (i.e. as a DC, DS, etc.). There's no limit to how many arguments can be used. Using Macro: <macro nmeumonic> arg1,arg2,arg3.... where arg1,arg2,arg3,... are values or symbols. -- Assembly-time For loop usage: For <symbol>=<start>:<stop> ... ... macs <symbol>,.... ... endfor <start> and <stop> must be integers -- Handling Skips the as10k1 assembler handles skips in a special way explained best by an example: skip CRR,CRR,CC_test,.foo ... ... ... .foo ... the "." tell the assembler that the symbol is for skipping purposes, it will automatically define a GPR when parsing the skip instruction, and when the second .foo is encountered it will insert the number of instructions to skip. (the skip instruction needs a GPR by design, so don't blame me for the half-assness of it). Features NOT YET Supported ========================== any ideas?