F28PL1 Programming Languages

Lecture 2: Assembly Language 1

CISC & RISC

• CISC: complex instruction set computer• original CPUs very simple• poorly suited to evolving high level languages • extended with new instructions– e.g. function calls, non-linear data structures, array bound

checking, string manipulation

• implemented in microcode– sub-machine code for configuring CPU sub-components– 1 machine code == > 1 microcode

• e.g. Intel X86 architecture

CISC & RISC

• CISC introduced integrated combinations of features– e.g. multiple address modes, conditional execution– required additional circuitry/power– multiple memory cycles per instruction– over elaborate/engineered

• many feature combinations not often used in practise

• could lead to loss of performance• better to use combinations of simple instructions

CISC & RISC

• RISC: reduced instruction set computer• return to simpler design• general purpose instructions with small number of

common features– less circuitry/power– execute in single cycle

• code size grew• performance improved• e.g. ARM, MIPS

ARM

• Advanced RISC Machines• UK company• originally Acorn– made best selling BBC Microcomputer in 1980s

• world leading niche processors– 2009: 90% of all embedded processors– mobile phones, media players, games, peripherals etc– e.g. Nokia, iPod, iPad, Blackberry, Nintendo

ARM

• 32 bit RISC processor architecture family• many variants • general reference in library:– W. Hohl, ARM Assembly Language, CRC Press, 2009– ARM7TDMI

• ARM does not make chips• licences the IP core (intellectual property) logic

design to chip manufacturers

Cortex-M3

• ARM CPU in STM32-Discovery board• Thumb2 instruction set– based on ARMV7– subset of full ARM instructions– 16 bit length instructions

• Cortex-M3 manual• http://infocenter.arm.com/help/topic/com.arm.doc.dui05

52a/DUI0552A_cortex_m3_dgug.pdf

MDK-ARM

• ARM development environment for:– C & assembler– many ARM-based systems

• includes simulator• free restricted download via:• http://www.keil.com/arm/mdkbasic.asp

• MDK-ARM/Cortex-M3 guide• http://www.st.com/internet/com/TECHNICAL_RESOURCES

/TECHNICAL_LITERATURE/USER_MANUAL/CD00283786.pdf

Cortex-M3 memory map

• linear 4GB address space

• modified Harvard architecture in CPU

Code

SRAM

Peripheral

External RAM

External Device

Vendor SpecificExternal Peripheral BusInternal Peripheral Bus

0x00000000

0x20000000

0x40000000

0x60000000

0xA0000000

0xE00400000xE0000000

0xE0100000

0xFFFFFFFF

0.5GB

1GB

1GB

0.5GB

0.5GB

0.5GB

Memory

• linear addressing • 2 descending stacks– main stack– process stack – used by OS

• all operations based on registers• must move data between registers & memory

Registers

• 16 * 32 bit registers• working registers– R0-R7 – low registers– R8-R12 – high registers

• R13 – stack pointer (SP)• R14 - link register (LR)• R15 – program counter (PC)• PSR – program status register– bits for negative (N), zero (Z), carry (C), overflow (V)

Data formats

• word – 32 bits– long

• half word– 16 bits– int

• byte – 8 bits– char

Programming approach

• in high level programming:– identify variables & types– compiler maps these to memory & data representation– use of memory v registers invisible

• in low level programming– must make explicit use of memory & registers– must choose data representation

• registers much faster than memory• try to map variables to registers• what if more variables than registers...?

Example: sum of 1st 10 integers

• we write:#define MAX 11int s;int n;n = 0;while(n<MAX){ s = s+n; n = n+1;}

• C compiler generates:MOVS r2,#0x00

MOVS r1,#0x00B TEST

LOOP ADD r2,r2,r1ADDS r1,r1,#1

TEST CMP r1,#0x0ABLT LOOP

Example: sum of 1st 10 integers

• we write:#define MAX 11int s;int n;n = 0;while(n<MAX){ s = s+n; n = n+1;}

• C compiler generates:MOVS r2,#0x00

MOVS r1,#0x00B TEST

LOOP ADD r2,r2,r1ADDS r1,r1,#1

TEST CMP r1,#0x0ABLT LOOP

Example: sum of first 10 integers

• s == r2 - register 2• n == r1 – register 1

MOVS r2,#0x00• move hexadecimal constant 0x00 to register 2 • set condition code flags MOVS r1,#0x00• move hexadecimal constant 0x00 to register 2 • set condition code flags

B TEST• branch to label TEST

Example: sum of first 10 integers

LOOP ADD r2,r2,r1• label LOOP• put in r2 sum of r2 and r1

ADDS r1,r1,#1• put in r1 sum of r1 and decimal constant #1• set condition code flags

Example: sum of first 10 integers

TEST CMP r1,#0x0A• label TEST• compare r1 and hexadecimal constant #0x0A == 10• set condition code flags

BLT LOOP• branch to label LOOP if flags indicate “less than”

Instruction format

mnemonic operands– general format– operands depend on instruction

mnemonic – meaningful short form– reminds us what instruction does

• ARM uses {...} for options• Rd or Rm == register -r0..r15

Operands

• first operand usually rd• Operand2 == flexible 2nd operand– constant – Rm

• constant == 8 hexadecimal digits• constant == decimal – size of decimal depends on instruction

• #imm16 == 0-65535 == 16 bit

MOV: move

MOV Rd, Operand2 Rd = Operand2• don’t update condition code flagsMOVS Rd, Operand2• as MOV but update condition code flagsMOV Rd, #imm16 Rd = imm16

ADD: addition

ADD Rd, Rn, Operand2 Rd = Rn+Operand2ADD Rn, Operand2 Rn = Rn+Operand2ADDC• add with carry • like ADD + carry flagADDS/ADDCS• like ADD/ADDC but set condition code flags

SUB: subtraction

SUB Rd, Rn, Operand2 Rd = Rn-Operand2SUB Rn, Operand2 Rn = Rn-Operand2SUBC• subtract with carry • like SUB -1 if carry flag not setSUBS/SUBCS• like SUB/SUBC but set condition code flags

CMP: compare

CMP Rn, Operand2• subtract Operand2 from Rn BUT ...• ... do not modify Rn – otherwise same as SUBS

• update Z, N, C, V flagsCMN RN, Operand2• add Operand2 to Rn BUT ...• ... do not modify Rn• otherwise same as ADDS

• update Z, N, C, V flags

Flags

• N – 1 if result <0; 0 otherwise• Z – 1 if result =0; 0 otherwise• C – 1 if result led to carry; 0 otherwise– i.e. X+Y > 232

– i.e. X-Y >= 0

Flags

• V – 1 if result led to overflow; 0 otherwise• (-) == negative• (+) == positive

– i.e. (-)X+(-)Y > 0– i.e. (+)X+(+)Y < 0– i.e. (-)X-(+)Y > 0– i.e. (+X)-(-Y) < 0

B: branch

B label• branch to label• i.e. reset PC to address for labelBcond label• branch on condition to label

conditionsuffix flagsEQ == equal Z=1NE == not equal Z=0CS/HS == carry set/ higher or same - unsigned C=1CC/LO == carry clear/lower - unsigned C=0MI == negative N=1PL == positive or 0 N=0VS == overflow V=1VC == no overflow V=0

conditionsuffix flagsHI == higher - unsigned C=1 & Z=0LS == lower or same - unsigned C=0 or Z=1GE == greater than or equal - signed N=VLT == less than - signed N!=VGT == greater than - signed Z=0 & N=VLE == less than or equal, signed Z=1 & N!=VAL == any value default if not

cond

Label

• identifier• can precede any line of assembler• represents the address of:– instruction– literal data

• relative to program counter (PC)• turned into PC + offset in machine code

NOP: no operation

NOP• do nothing• use for padding/layout• no cost

Layout

• ARM assembly language is layout sensitive• must precede instruction with a tab• for labelled instruction, use tab immediately after

label• can’t use equivalent number of spaces

Example: multiply by adding

• m == x*yint x;int y;int m;x = 8;y = 10;m = 0;while(y!=0){ m = m+x; y = y-1;}

• x == R1; y == R2; m == R3MOV R1,#0x08MOV R2,#0x0AMOV R3,#0x00

LOOP CMP R2,#0x00BEQ DONEADD R3,R1SUB R2,#0x01B LOOP

DONE NOP

Register name definition

name RN expression• expression 0..15• name renames the

registere.g.x RN 1y RN 2m RN 3

MOV x,#0x08MOV y,#0x0AMOV m,#0x00

LOOP CMP y,#0x00BEQ DONEADD m,xSUB y,#0x01B LOOP

DONE NOP

Constant name definition

name EQU expression• name is replace with

value from expressione.g.x RN 1y RN 2m RN 3xinit EQU 8yinit EQU 10

MOV x,#xinitMOV y,#yinitMOV m,#0x00

LOOP CMP y,#0x00BEQ DONEADD m,xSUB y,#0x01B LOOP

DONE NOP