CS2002

Question 1

Fetch-Execute Cycle

Answer 1

1. Load data from the IP 2. Set the ip to the next instruction 3. Control unit decodes the instruction 4. ALU executes the instruction 5. Some instr may change the pointer to something else

Question 2

ISA

Answer 2

Set of instructions understood by a CPU -

• Their encoding into bits
• Their meaning

Question 3

imulq $86, (%rax), %rcx

Answer 3

• imul: signed integer multiplication
• imulq: qword or quadword= 64 bit
• $86: Constant 86
• %rax, %rcx : registers
• anything inside brackets, is basically just derefencing

Question 4

1. Compiler
2. Interpreter
3. Assembler
4. Linker
5. Loader

Answer 4

1. Compiler: converts high level instructions to low level instructions which the CPU can understand
2. Interpreter: Program that reads high level programmes and carries it out
3. Assembler: Converts assembly language to object code, think of it as compilers for lower level languages.
4. Linker: Combines object code files to create an executable file that can run by including different library object code files and determining memory offsets for functions in different files.
5. Loader: Loads all the libraries into memory and prepares them for execution

Question 5

List the different kinds of registers

Answer 5

1. GP registers: Stores one word and are used in integer operations, typically a small number, e.g. x86 which has 16 registers
2. Floating point registers:
3. Special registers: IP a.k.a program counters

Question 6

Orthogonality of ISAs

Answer 6

When ISA doesn’t distinguish between different registers. What can be done with one register can be done with any other registers. Each instruction performs a unique task that doesn’t overlap with another instruction.

ARM is orthogonal

Question 7

Uses of Conventions

Answer 7

They help structure writing assembly code

1. Not writing values immediately to registers

Question 8

ISA Design decisions for where arguments can be loaded and stored

Answer 8

1. Register-memory: Atleast one argument can come straight from memory
   
   1. More expensive.
2. Load store: All arguments are directly from the memory, register - register
   
   1. Memory is accessed only to load data into the registers
   2. Separation of memory and arithmetic operations make design easier.

Question 9

Word Addressed

Answer 9

One address for one word and there is extra info to refer to the bytes inside the word

Question 10

Byte Addressed

Answer 10

One address per octet ( 8 bits ) and the lowest address in a word represents the address of the word.

Question 11

How does ISA allow memory access or what are the different memory addressing modes?

Answer 11

1. Immediate mode
2. Direct mode
3. Indirect mode
4. Index mode

Question 12

1. Immediate Memory Access Modes
2. Direct Mode

Answer 12

Immediate Mode: The value is specified in the instruction itself, used for constants.

Direct Mode: The value is stored at a fixed address. Used for global variables.

Question 13

1. Indirect Memory access Mode
2. Index mode

Answer 13

Indirect mode: The value is stored in a fixed address which is specified inside a register. Extra power(check)

Index Mode: The value is at a fixed address specified by register but is offset by an index.

Question 14

Register Naming

Answer 14

Question 15

Stackframes

Answer 15

The stack frame is the collection of all data on the stack associated with the function call like the arguments to the function and the return address

Question 16

The steps involved in a function call with respect to the stack frame and the different pointers

Answer 16

1. Push the arguments onto the stack
2. execute callq
3. push the value of rbp onto the stack
4. make rbp the current rsp
5. decrement rsp to make space for local variables and stuff

Question 17

What are the 2 special purpose registers?

Answer 17

1. rip: Instruction pointer
2. rflags: binary flags for comparison

Question 18

Main addressing mode in x86 ( AT&T Syntax )

Answer 18

disp(%reg1, %reg2, scale)

• disp - displlacement
• reg1 - register that holds the base address
• reg2 - register that holds the index
• scale - value which can be only 1, 2, 4, 6

This refers to the address, disp + (reg1) + scale * reg2

Question 19

What is load effective addressing?

Answer 19

Stores the effective address of the source into the destination register instead of storing the value pointed to by effective address ( this is what mov does )

Lea doesn’t read from the computed address whereas Mov does.

Question 20

Command for compiling a C file ( sam.c ) to an assembly file ( sam.s)

Answer 20

clang -S sam.c -o sam.s

Question 21

Which register is used for:

1. Passing arguments to a function
2. Returning values from a function

Answer 21

1. rdi
2. rax is used if the value fits in 64 bits, otherwise, the upper 64 bits is stored in the rdx register.

Question 22

Calling conventions (1) - How arguments are passed into functions.

Answer 22

There are 2 ways in which arguments are passed onto a function

• if there are N arguments to a function and N <= 6, and all the arguments fit into the 64 bit registers then they are stored inside these registers
  
  • rdi, rsi, rdx, rcx, r8, r9
  • the return value from function should be stored in rax
• If there are more arguments or they are longer, then these go onto the stack
• If the callee wants to use any of the arguments then they should be restored back to the original value upon return

Question 23

Call and exit in ISA

Answer 23

callq address: pushes the address of the following instruction ( this is the return address ) onto the stack and transfers the control to the function by setting the rip to the address of the function.

retq: pops the address from the stack by incrementing the stackpointer and returns control back to the calling function by setting rip to the popped address

Question 24

What are callee saved registers?

Answer 24

rbx, rbi, r12 - r15

Question 25

Calling conventions (2)

Answer 25

Stack is aligned at multiples of 16 bytes There is a 128 byte red zone beneath the stack pointer, this is for the callee to store temporary data without creating a new stackframe The last argument is pushed in first ( reverse order ) so from right to left. Usually the base pointer is used with a positive offset in order to retrieve the arguments passed into the function

Question 26

What is relative addressing and what is the advantage of using it?

Answer 26

RIP rel ad is using the current value of an rip plus an offset to locate an operand rather than using the actual address of the operand. This leads to position independent code i.e. code that runs no matter where it is in the memory.

Question 27

Why would you not want to use the heap for stack frames?

Answer 27

This is because the heap may contain gaps due to freeing up memory whereas, the stack is contiguous.

Question 28

What are rflags?

Answer 28

These are 64 bit flags that are not available as general purpose flags but they contain the result of the execution of a previous instruction. They are used for conditional jumps.

Question 29

mul vs imul div vs idiv

Answer 29

mul (signed) and imul (unsigned ) * atleast one operand should be in rax div (signed ) idiv ( unsigned ) * atleast one operand should be in rax * quotient goes to rax and remainder goes to rdx

Question 30

Shift and rotate * Logical shift * Arithemetic shift * rotate right/left * rotate with carry right/left

Answer 30

* Logical shift: shr, shl - * Arithmetic shift: sar, sal - right shift duplicates sign bit * rotate right/left: ror, rol: * rotate with carry right/left: rcr, rcl - rotate with carry, involve carry bit in rotation Artithmetic shift: preserves sign bit, can be used for multiplication and division Logical shift: Does not preserve sign bit, used for unsigned interpretation.

Question 31

Flags 1. ja, jae 2. jb, jbe 3. jz, jze 4. js, jns

Answer 31

1. Equal/not eq unsigned 2. less than or less than or equal to unsigned 3. zero or not zero 4. if signed or unsigned

Question 32

Unconditional Control flow

Answer 32

1. jmp: goto label 2. call: pushes the rip onto the stack and jumps to that address 3. ret: pops the stack and stores it in the rip

Question 33

What are returns the time stamp counter and what returns the clock cycles

Answer 33

1. The RDTSC returns the time stamp counter 2. TCS returns the count of the clock cycles

Question 34

How to compile a file foo.c: 1. Such that it is only Preprocessed 2. The preprocessed version is converted to assembly. 3. Convert assembly code to object code

Answer 34

clang -E foo.c -0 foo-preprocessed.c clang -S foo-preprocessed.c -o foo.s clang -c foo.s -o foo

Question 35

What is big endian and little endian?

Answer 35

It is the order in which bytes are stored in memory. In Big Endian, the most significant byte is stored at the ( first )lowest address and the least significant byte is stored at the highest address. In Little Endian, the least significant byte is stored at the lowest address and the most significant address is stored at the highest address.

Question 36

Difference between CISC & RISC arch:

Answer 36

RISC: - executes smaller instructions - Since they are smaller, more number of instructions are needed - Each clock cycle executes one task - The instruction since they are simpler are easier to decode CISC: - executes more complex instructions - There maybe dedicated hardware that performs some of these instructions - Complex instructions may take multiple cycles to executes. - lots of addressing modes for programmers

Question 37

How many general purpose registers are there in the modern CPU ( their size and count )

Answer 37

There are 16 64 bit general purpose registers in the CPU.

Question 38

How did the names of the general purpose registers change when they went from 8 to 16 in number and from 32 bits to 64 bits in size?

Answer 38

The new ones are numbered r8 to r15 Initially the registers were just named AX, BX, CX, DX, SI, SD, SP, BP When the 32 bit registers came, they became extended (E) EAX, EBX, ECX, EDX, ESI, ESD, ESP, EBP When the 64 bit versions came RAX, RBX, RCX, RDX, RSI, RSD, RSP, RBP

Question 39

What should you be aware of in terms of the order of operands when comparing them in AT&T syntax.

Answer 39

The inversion of operands cmp $1 %rax This value will be 1 if rax \> 1, there is an inversion of the operands. subq $1 %rax subtract 1 from rax and then store the result in rax (and not subtract rax from 1)