HW1 Project Will be assigned tonight Proposal due next Wed - - PDF document

Chenyang Lu CSE 467S 1

HW1

• Will be assigned tonight
• Due on 2/24 (Wed), 5pm
• Submit to hand-in bins outside Lopata 408.
• Homework will NOT be accepted after due

date/time.

• Collaboration is NOT allowed.

Chenyang Lu CSE 467S 2

Project

• Proposal due next Wed (2/16)
• Do you have a team?
• Do you have an idea?
• Have you started reading related

materials?

• Welcome to discuss with me!

Chenyang Lu CSE 467S 3

More project ideas

• Program applications on Agilla
• Existing examples:
• Fire tracking (see today’s demo)
• Intruder tracking
• Network exploration
• Implement power management modules
• n TinyDB

Chenyang Lu CSE 467S 4

Now, we head to MobiLab!

Chenyang Lu CSE 467S 5

Software Optimization and Analysis

Chenyang Lu CSE 467S 6

Transformation of Software

compile assembly assemble assembly assembly link load assembly assembly

• bject

HLL HLL HLL Relative Address Absolute Address Physical Address executable

• ptimize

Analyze

Chenyang Lu CSE 467S 7

Program design and analysis

• Optimizing for execution time.
• Optimizing for energy/power.
• Optimizing for program size.
• They may conflict with each other!

Chenyang Lu CSE 467S 8

Optimization for Performance

Chenyang Lu CSE 467S 9

Why do we need to know?

• Understand various optimization levels (-O1, -

O2, etc.)

• Look at mixed compiler/assembler output.
• Modifying compiler output requires care:
• correctness;
• loss of hand-tweaked code.

Chenyang Lu CSE 467S 10

Expression simplification

• Constant folding:
• 8+1 = 9
• Algebraic:
• a*b + a*c = a*(b+c)
• Strength reduction:
• a*2 = a<<1

Chenyang Lu CSE 467S 11

Dead code elimination

• Dead code:

#define DEBUG 0 if (DEBUG) dbg(p1);

• Can be eliminated by

analysis of control flow, constant folding.

dbg(p1); 1

Chenyang Lu CSE 467S 12

Function calls

Chenyang Lu CSE 467S 13

ARM

• Branch and link instruction:

BL foo == MOV r14, r15 B foo

• r15 contains the current PC
• Copies current PC to r14.
• To return from subroutine:

MOV r15,r14

Chenyang Lu CSE 467S 14

Function calls

• Understand the overhead of function calls
• Caller and callee must follow a consistent

protocol (order) to access the stack

• Different compilers and programmers may follow

different orders

• Access the stack (ARM)
• Push: STR r0, [r13, #4]!
• Pop: SUB r13, #4

Chenyang Lu CSE 467S 15

Example Protocol for ARM

• R13 always points to the top of stack
• Caller: call a function
• Push parameters to stack
• BL (R15 (PC) R14; jump)
• Callee: receive a call
• Read parameters from stack
• Overwrite top of stack with return address (R14)
• Callee: return
• Load PC with return address (on top of stack)
• Caller: receive a return
• Pop callee’s return address from stack
• SHARC: PC stack no need to push/pop return addresses

Chenyang Lu CSE 467S 16

Nested function calls

int main() { f1(int x); } void f1(int a) { f2(a); } ; f1 is called by main() f1 LDR r0, [r13] ; load para. into r0 from stack STR r14, [r13] ; store f1’s return addr. ; f1 calls f2() STR r0, [r13, #4]! ; push para. for f2 to stack BL f2 ; branch and link to f2 ; f1 receives return from f2() SUB r13, #4 ; pop f2’s para. off stack ; f1 returns to main() LDR r15, [r13] ; restore register and return Correct p. 82 of textbook!

Chenyang Lu CSE 467S 17

Function inlining

int foo(a,b,c) { return a + b - c;} z = foo(w,x,y);

• z = w + x - y;

Improve performance

• Eliminates procedure linkage overhead:

May increase code size

• Not always…
• Affect instruction cache behavior

Chenyang Lu CSE 467S 18

Reading

• Sec 5.5, 5.6, 5.7