ARM instruction layout

{
	label
}
{
	instruction
||
	directive
||
pseudo-instruction
} {; comment}

• Even without a label, instructions, pseudo-instructions, and directives must start with whitespace (space or tab).

Case rule

• Instruction, directive, and symbolic register names may be all uppercase or all lowercase, but not mixed.

Line length

• Long lines can wrap with \.
• Nothing may follow the backslash on that line.

Label

• A symbol that stands for an address.
• The assembler computes the address during assembly.
• It measures the label relative to the origin of the section where the label is defined.
• References to labels in the same section can use PC plus or minus an offset — program-relative addressing.

Local label

• A subclass of labels.
• Local labels start with a number in the range 0–99.
• Unlike ordinary labels, you can define the same local label more than once.
• That helps when macros generate labels.
• When the assembler sees a local-label reference, it binds it to a nearby instance of that local label.

Comment

• The first semicolon on a line starts a comment, unless it sits inside a string constant.
• The end of the line ends the comment.
• A line that is only a comment is still valid.
• The assembler ignores all comments.

ARM instruction format

AREA     ARMtest, CODE, READONLY
                                ; name this code block ARMtest
        ENTRY                   ; mark the first instruction to run
start
        MOV      r0, #10        ; set parameters
        MOV      r1, #3
        ADD      r0, r0, r1     ; r0 = r0 + r1
stop
        MOV      r0, #0x18      ; angel_SWIreason_ReportException
        LDR      r1, =0x20026   ; ADP_Stopped_ApplicationExit
        SWI      0x123456       ; ARM semihosting SWI
        END                     ; Mark end of file

ELF sections and the AREA directive

• An ELF section is an independent, named, indivisible run of code or data.

• You need at least one code section to build an application.

• Assembly or compile output may include:

  • One or more code sections, usually read-only
  • One or more data sections, usually read-write, sometimes zero-initialized

• The linker places each section into the program image under its placement rules.

• Adjacent sections in a source file are not necessarily adjacent in the image.

• In ARM assembly sources, a section starts with the AREA directive.

• That directive names the section and sets its attributes.

• Attributes follow the name, comma-separated.

ENTRY directive

• Marks the first instruction to execute.
• In apps that include C, the entry point also lives in the C library startup code.
• Initialization code and exception handlers carry entry points as well.

Application start

• Execution begins at the start label.
• Decimal 10 and 3 load into r0 and r1.
• Those registers are added; the sum lands in r0.

Application terminate

• After the main code, the app returns control to the debugger and exits.
• That uses the ARM semihosting SWI with the parameters shown above.

END directive

• Tells the assembler to stop processing this source file.
• Every assembly language source module must end with END on its own line.

Calling subroutines

• Call a subroutine with branch-and-link:

BL destination

• destination is usually a label on the subroutine's first instruction.
• It can also be a program-relative or register-relative expression.

Branch instructions

B (Branch == Jump)

BL

• Puts the return address in link register r14 (lr).
• Sets the PC to the subroutine address.

BX (Branch and Exchange)

BLX (Branch with Link and Exchange)

Branch : B{<cond>} label
Branch with Link : BL{<Cond>} sub routine label

• After the subroutine runs, return with the usual move of lr into pc.
• By convention, r0–r3 carry parameters in and results back out.

MOV pc, lr

AREA    subRoutTest, CODE, READONLY
                   
        ENTRY                     
start   MOV     r0, #10           
        MOV     r1, #3
        BL      apple             ; Call subroutine
stop    MOV     r0, #0x18         ; angel_SWIreason_ReportException
        LDR     r1, =0x20026      ; ADP_Stopped_ApplicationExit
        SWI     0x123456          ; ARM semihosting SWI
apple   ADD     r0, r0, r1        ; Subroutine code
        MOV     pc, lr            ; Return from subroutine
        END                       ; Mark end of file

Thumb and ARM assembly

AREA ThumbArm, CODE, READONLY           
        ENTRY                           
        CODE32                          ; following instructions are ARM
header  ADR     r0, start + 1           ; processor starts in ARM state
        BX      r0                      ; tiny ARM header
                                        ; call Thumb main program
        CODE16                          ; following instructions are Thumb
start
        MOV     r0, #10                 
        MOV     r1, #3
        BL      apple                   
stop
        MOV     r0, #0x18               ; angel_SWIreason_ReportException
        LDR     r1, =0x20026            ; ADP_Stopped_ApplicationExit
        SWI     0xAB                    ; Thumb semihosting SWI
apple
        ADD     r0, r0, r1              ; Subroutine code
        MOV     pc, lr                  ; Return from subroutine
        END                             ; Mark end of file

CODE32 and CODE16

• Direct the assembler to emit ARM (CODE32) or Thumb (CODE16) instructions.
• They do not assemble into runtime state-change ops.
• Only the assembler's state changes.

BX

• A branch that can change processor state at runtime.
• The low bit of the target address distinguishes ARM vs Thumb instructions.