★ HARDWARE Manual ★ SCU User's Manual

▲ Back ｜ Forward ▼

SCU DSP Arambra Instruction Manual

'94 / 07/08

1. Assembler function

2. Execution of assembler

3. Program description format

4. Instruction list

5. Sample program

1. Outline of assembler function

This assembler is for generating the instruction code for DSP, and is intended to be operated on MS-DOS and UNIX. The output code is Motorola's S format, and the code is output directly, so no link is required.
The DSP assembler requires a lot of knowledge about the hardware, so please familiarize yourself with the DSP hardware before using the assembler.

2. Execution of assembler

dspam [option] <source file name>

The following can be specified for the option. (File creation is only at the end of normal)
-l [filename]: List output specification
-a [file name]: Specify data format output for SH assembler
-c [file name]: C language data format output specification
-m: Specified when using the MODEL M development environment
The source file name does not have a default extension.
The error is displayed only for the first one detected, so repeat the correction and assembly until the error disappears.

3. Program description format

[Label] [△ Operation [△ Operand]]… [Comment]
Example: LABEL: MOV MC0, X; Comment

(1) Label

It is defined by the programmer and can be used as the name of the jump destination of the JMP instruction.
When writing a label, start writing from the first column or add a ":" at the end, such as "LABEL:".
The maximum length of the label is 32 characters, the characters that can be used are uppercase letters, lowercase letters, numbers, and underscores (_), and only numbers can be placed at the beginning. In addition, uppercase and lowercase letters are equated.

(2) Operation

Describe the execution instruction of DSP.
When you start writing from an operation, put a space character in front of it.
Up to 6 operations may be arranged for only arithmetic instructions.

(3) Operand

Describes what the operation is to be executed.
Use a blank character to separate it from the operand.

(4) Comment

You can write annotations to make the program easier to understand.
Comments are from the position marked with ";" to the end of the line.

* Notes on description

Basically, the operation ~ operand is described so that it fits on one line, but if it does not fit on one line, you can add "\" before the line feed code and connect it to the next line. However, if you want to connect operations after a line with a comment, add "\" before ";". At this time, do not exceed the maximum length of one line, 255 bytes.
Operations and operands can be in uppercase or lowercase.
Use hexadecimal / $ xx / decimal / xxx / binary /% xxxxxxxx for the numerical notation.
The address where the output code is placed can be specified by the ORG pseudo-instruction.
The program area in the DSP has only 256 instructions, but in order to smoothly divide and optimize the processing, Warning is issued and the code for 2048 instructions can be output. This code can only be supported by the simulator, so when using it in an actual DSP, it is necessary to shorten it within 256 instructions and assemble it. Also, when using a label with an address that exceeds 256 instructions, please note that 8-bit values cannot be used for substitution.

* About reserved words

The following names used for operands cannot be used for labels.
{ALH, ALL, ALU, M0, M1, M2, M3, MC0, MC1, MC2, MC3, MUL}

* About numerical calculation

The following operators can be used when setting values for labels and using numbers for operands. (However, please note that you cannot enter blank characters when using it in the operand. Example: ○ JMP $ + 2, × JMP $ + 2)

Arithmetic relation

+… Addition

--… Subtraction

*… Multiplication

/… Division

%… Exclusion

~… Bit negation

&… Bit product

|… Bit sum

^… Exclusive bit sum

<<… Left shift

>>… Right shift

Priority

1. +-~ (unary operator)

2. * /%

3. +-

4. <<>>

Five. &

6. | ^

4. Instruction list

Arithmetic instructions
NOP AND OR XOR ADD SUB AD2 SR RR SL RL RL8 CLR MOV
Load immediate instruction
MVI
DMA command
DMA DMAH
JUMP instruction
JMP
LOOP BOTTOM instruction
BTM LPS
END command
END ENDI

Pseudo-instruction

EQU (=) ……………… Used to define labels.
ORG ……………… Specify the first address to place the instruction.
ENDS ……………… If you put it at the end of the program, it will be ignored after that.
IF <numerical value / label>… If the calculation result of the numerical value or label is other than 0, then ELSE or ENDIF will be assembled.
IFDEF <label> ……… If the label is defined in front, ELSE or ENDIF will be assembled thereafter.
The nesting level of IF and IFDEF is up to 16. )

5. Sample program

(1) When copying the contents of the internal RAM 0 of the DSP to the internal RAM 1.

; ------- sample (1) start -------

COPY_SIZE = 12; Copy size RAM0_ADR = $ 00; Copy source address RAM1_ADR = $ 00; Copy destination address MOV RAM0_ADR, CT0; Set the copy source address of RAM0 MOV RAM1_ADR, CT1; Set the copy destination address of RAM1 MOV COPY_SIZE-1, LOP ; Set transfer size -1 to LOP register LPS; Execution of 1 instruction loop MOV MC0, MC1; Transfer from RAM0 to RAM1 ENDI

; ------- sample (1) end -------

(2) When calculating 2x3 + 4x5. (RAM0 x RAM1 + RAM0 x RAM1 = RAM2) (sample 2b is an optimized version of 2a.)

; ------- sample (2a) start -------

RAM0_ADR = $ 00; 2, 4 Storage address start RAM1_ADR = $ 00; 3, 5 Storage address start RAM2_ADR = $ 00; Result storage address MOV RAM0_ADR, CT0; Set RAM0 address MOV RAM1_ADR, CT1; Set RAM1 address MVI # 2 , MC0; RAM0 set to "2" MVI # 3, MC1; RAM1 set to "3" MVI # 4, MC0; RAM0 set to "4" MVI # 5, MC1; RAM1 set to "5" MOV RAM0_ADR, CT0; set RAM0 address MOV RAM1_ADR, CT1; set RAM1 address MOV RAM2_ADR, CT2; set RAM2 address MOV MC0,X; transfer data from RAM0 to RX MOV MC1, Y; from RAM1 to RY Data transfer to MOV MUL, P; RX and RY integration results stored in PH, PL MOV MC0, X; Data transfer from RAM0 to RX MOV MC1, Y; Data transfer from RAM1 to RY CLR A; ACH, ACL AD2 MOV ALU, A; PH, PL and ACH, ACL addition result is stored in ACH, ACL MOV MUL, P; RX and RY integration result is stored in PH, PL AD2 MOV ALL, MC2 ; PH, PL and ACH, ACL addition result stored in RAM2 ENDI

; ------- sample (2a) end -------
; ------- sample (2b) start -------

RAM0_ADR = $ 00; 2, 4 Storage address start RAM1_ADR = $ 00; 3, 5 Storage address start RAM2_ADR = $ 00; Result storage address MOV RAM0_ADR, CT0
	                                                 MOV RAM1_ADR, CT1
	MVI # 2, MC0
	MVI # 3, MC1
	MVI # 4, MC0
	MVI # 5, MC1
	                                                 MOV RAM0_ADR, CT0
	                                                 MOV RAM1_ADR, CT1
	     MOV MC0, X MOV MC1, Y MOV RAM2_ADR, CT2
	     MOV MC0, X MOV MUL, P MOV MC1, Y CLR A
	AD2 MOV MUL, P MOV ALU, A
	AD2 MOV ALL, MC2
	ENDI

; ------- sample (2b) end -------

(3) When calculating the movement process for the matrix. (RAM0 x RAM1 = RAM2)

 / M00 M01 M02 M03 \ / 100x \ / M00 M01 M02 M03 \
｜ M10 M11 M12 M13 ｜｜ 010y ｜ → ｜ M10 M11 M12 M13 ｜
 \ M20 M21 M22 M23 / | 001z | \ M20 M21 M22 M23 /
 \ 0001 /

; ------- sample (3) start -------

DATA_TOP = $ 10000 >> 2; External memory address is in 4-byte units MAT_SIZE = $ 0C; Array size RAM0_ADR = $ 00; Start address RAM1_ADR = $ 00 that stores the amount of movement of X, Y, Z; Array address RAM2_ADR = $ 00; Address of the original array; (Transfer the array with the movement amount set to RAM0 from the external memory)
;
	MVI DATA_TOP, RA0
	                                                 MOV RAM0_ADR, CT0
	DMA D0, MC0, # $ 02
;
(Copy of array to operate from RAM2 to RAM1)
	                                                 MOV RAM2_ADR, CT2
	                                                 MOV RAM1_ADR, CT1
	                                                 MOV MAT_SIZE-1, LOP
	LPS
	                                                 MOV MC2, MC1
WAITING:
	JMP T0, WAITING
;
(Calculation execution of array)
	                                                 MOV RAM0_ADR, CT0
	                                                 MOV RAM1_ADR, CT1
	     MOV MC0, X MOV MC1, Y 
	     MOV MC0, X MOV MUL, P MOV MC1, Y CLR A
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y MOV ALU, A MOV RAM0_ADR, CT0
	AD2 MOV MUL, P MOV MC1, Y MOV ALU, A MOV # 1, RX
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y MOV ALU, A MOV RAM2_ADR + 3, CT2
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y CLR A MOV ALL, MC2
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y MOV ALU, A MOV RAM0_ADR, CT0
	AD2 MOV MUL, P MOV MC1, Y MOV ALU, A MOV # 1, RX
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y MOV ALU, A MOV RAM2_ADR + 7, CT2
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y CLR A MOV ALL, MC2
	AD2 MOV MC0, X MOV MUL, P MOV MC1, Y MOV ALU, A MOV RAM0_ADR, CT0
	AD2 MOV MUL, P MOV MC1, Y MOV ALU, A MOV # 1, RX
	AD2 MOV MUL, P MOV ALU, A MOV RAM2_ADR + 11, CT2
	AD2 MOV ALL, MC2
	ENDI

; ------- sample (3) end -------

that's all

▲ Back ｜ Forward ▼

★ HARDWARE Manual ★ SCU User's Manual

SCU DSP Arambra Instruction Manual

table of contents

1. Outline of assembler function

2. Execution of assembler

3. Program description format

(1) Label

(2) Operation

(3) Operand

(4) Comment

Arithmetic relation

Priority

4. Instruction list

Pseudo-instruction

5. Sample program

(1) When copying the contents of the internal RAM 0 of the DSP to the internal RAM 1.

(2) When calculating 2x3 + 4x5. (RAM0 x RAM1 + RAM0 x RAM1 = RAM2) (sample 2b is an optimized version of 2a.)

(3) When calculating the movement process for the matrix. (RAM0 x RAM1 = RAM2)