Copyright 2014 – Noah Mendelsohn UM Macro Assembler Functions Noah Mendelsohn Tufts University Web: COMP 40: Machine Structure and Assembly Language Programming (Fall 2014)

© 2010 Noah Mendelsohn 2 Today  Creating and calling functions using UMASM  Recursion and using the stack  Invariants

© 2010 Noah Mendelsohn 3 Review x86-64 Standard Function Linkage

© 2010 Noah Mendelsohn Why have a standard AMD 64 “linkage” for calling functions? (Review)  Functions are compiled separately and linked together  We need to standardize enough that function calls can be freely mixed  We may “cheat” when caller and callee are in the same source file 4

© 2010 Noah Mendelsohn Function calls on Linux/AMD 64  Caller “pushes” return address on stack  Where practical, arguments passed in registers  Exceptions: –Structs, etc. –Too many –What can’t be passed in registers is at known offsets from stack pointer!  Return values –In register, typically %rax for integers and pointers –Exception: structures  Each function gets a stack frame –Leaf functions that make no calls may skip setting one up  Caller vs. callee saved registers 5

© 2010 Noah Mendelsohn AMD 64 The stack – general case (review) 6 Before call ???? %rsp Arguments Return address After callq %rsp Arguments %rsp If callee needs frame ???? Return address Args to next call? Callee vars sub $0x{framesize},%rsp Arguments framesize

© 2010 Noah Mendelsohn AMD 64 arguments/return values in registers 7 Arguments and return values passed in registers when types are suitable and when there aren’t too many Return values usually in %rax, %eax, etc. Callee may change these and some other registers! MMX and FP 87 registers used for floating point Read the specifications for full details! Operand Size Argument Number %rdi%rsi%rdx%rcx%r8%r9 32%edi%esi%edx%ecx%r8d%r9d 16%di%si%dx%cx%r8w%r9w 8%dil%sil%dl%cl%r8b%r9b

© 2010 Noah Mendelsohn fact:.LFB2: pushq %rbx.LCFI0: movq %rdi, %rbx movl $1, %eax testq %rdi, %rdi je.L4 leaq -1(%rdi), %rdi call fact imulq %rbx, %rax.L4: popq %rbx ret AMD 64 Factorial Revisited 8 int fact(int n) { if (n == 0) return 1; else return n * fact(n - 1); }

© 2010 Noah Mendelsohn 9 The UMASM Standard Function Linkage

© 2010 Noah Mendelsohn Why have a standard UMASM function linkage?  Functions are compiled separately and linked together  We need to standardize enough that function calls can be freely mixed  It’s tricky, so doing it the same way every time minimizes bugs and makes code clearer  We may “cheat” when caller and callee are in the same source file 10

© 2010 Noah Mendelsohn Summarizing calling conventions (review)  r0 is always 0 (points to segment 0, used for goto and for the call stack)  r1 is the return-address register and the result register  r2 is the stack pointer and must have same value after return (nonvolatile)  r3, r4 are nonvolatiles  r5 is a helpful volatile register (good to copy return address)  r6, r7 are volatile and are normally.temps (but up to each procedure)  Arguments are on the stack, with first argument at the “bottom” (where r2 points) 11 EXCEPT FOR VOLATILES AND RESULT REGISTER CALLEE MUST LEAVE REGISTERS AND STACK UNCHANGED!

© 2010 Noah Mendelsohn Startup code with stack (REVIEW)  Goals of startup code: –Set up useful registers 12 init (instructions) start.section init.temps r6, r7.section stk.space endstack:.section init.zero r0 start: r0 := 0 r2 := endstack goto main linking r1 halt

© 2010 Noah Mendelsohn Startup code with stack (REVIEW)  Goals of startup code: –Set up useful registers 13 init (instructions) start.section init.temps r6, r7.section stk.space endstack:.section init.zero r0 start: r0 := 0 r2 := endstack goto main linking r1 halt

© 2010 Noah Mendelsohn Startup code with stack (REVIEW)  Goals of startup code: –Set up useful registers –Set up a stack 14 init (instructions) start stk words endstack (r2).section init.temps r6, r7.section stk.space endstack:.section init.zero r0 start: r0 := 0 r2 := endstack goto main linking r1 halt

© 2010 Noah Mendelsohn Startup code with stack (REVIEW)  Goals of startup code: –Set up useful registers –Set up a stack –Let UMASM know about useful registers 15 init (instructions) start stk words endstack (r2).section init.temps r6, r7.section stk.space endstack:.section init.zero r0 start: r0 := 0 r2 := endstack goto main linking r1 halt

© 2010 Noah Mendelsohn Startup code with stack (REVIEW)  Goals of startup code: –Set up useful registers –Set up a stack –Let UMASM know about useful registers –Invoke or drop through to program code 16 init (instructions) start stk words endstack (r2).section init.temps r6, r7.section stk.space endstack:.section init.zero r0 start: r0 := 0 r2 := endstack goto main linking r1 halt

© 2010 Noah Mendelsohn Summarizing calling conventions  r0 is always 0 (points to segment 0, used for goto and for the call stack)  r1 is the return-address register and the result register  r2 is the stack pointer and must have same value after return (nonvolatile)  r3, r4 are nonvolatiles  r5 is a helpful volatile register (good to copy return address)  r6, r7 are volatile and are normally.temps (but up to each procedure)  Arguments are on the stack, with first argument at the “bottom” (where r2 points) 17 EXCEPT FOR VOLATILES AND RESULT REGISTER CALLEE MUST LEAVE REGISTERS AND STACK UNCHANGED!

© 2010 Noah Mendelsohn 18 Calling a Function with No Arguments

© 2010 Noah Mendelsohn Calling a function with no arguments 19 r1 := return_addr // Offset in seg 0 r7 := somefunc // Offset in seg 0 goto *r7 in program m[r0] // Call the function return_addr: // Function returns here …back in caller… call somefunc();

© 2010 Noah Mendelsohn Calling a function with no arguments 20 r1 := return_addr // Offset in seg 0 r7 := somefunc // Offset in seg 0 goto *r7 in program m[r0] // Call the function return_addr: // Function returns here …back in caller… call somefunc(); Use of r1 for return address is required

© 2010 Noah Mendelsohn Calling a function with no arguments 21 r1 := return_addr // Offset in seg 0 r7 := somefunc // Offset in seg 0 goto *r7 in program m[r0] // Call the function return_addr: // Function returns here …back in caller… call somefunc(); Use of r7 for function address is programmer choice

© 2010 Noah Mendelsohn Calling a function with no arguments 22 r1 := return_addr // Offset in seg 0 r7 := somefunc // Offset in seg 0 goto *r7 in program m[r0] // Call the function return_addr: // Function returns here …back in caller… call somefunc(); goto somefunc linking r1 // goto main, setting r1 to the return address UMASM provides this built-in macro to do the above for you!

© 2010 Noah Mendelsohn Calling a function with no arguments 23 r1 := return_addr // Offset in seg 0 r7 := somefunc // Offset in seg 0 goto *r7 in program m[r0] // Call the function return_addr: // Function returns here …back in caller… call somefunc(); goto somefunc linking r1 // goto main, setting r1 to the return address UMASM provides this built-in macro to do the above for you! Also remember: r1 is both the linking and the result register. If somefunc calls another function, then r1 must be saved.

© 2010 Noah Mendelsohn 24 Organizing Startup Code

© 2010 Noah Mendelsohn Language startup routines  The C language uses a startup/closeup routine called crt0 to prepare arguments and invoke main –Sets up argv and argc –Sets up exception handling –Gathers return/exit value from main –Special purpose versions possible: e.g. gcrt0 used when profiling  So…startup code, run before main(), is shared by all programs. Can we do the same for UMSAM? –Can we package our UMASM programs into a main() function?  …Yes! 25

© 2010 Noah Mendelsohn urt0: A reuseable UMASM startup routine 26.zero r0.temps r7.section stk.space endstack:.section init _ustart: r0 := 0 r2 := endstack goto main linking r1 halt // Finis urt0.ums

© 2010 Noah Mendelsohn urt0: A reuseable UMASM startup routine 27.zero r0.temps r7.section stk.space endstack:.section init _ustart: r0 := 0 r2 := endstack goto main linking r1 halt // Finis Establish r0 as always containing 0

© 2010 Noah Mendelsohn urt0: A reuseable UMASM startup routine 28.zero r0.temps r7.section stk.space endstack:.section init _ustart: r0 := 0 r2 := endstack goto main linking r1 halt // Finis Build stack and set up r2 as stack pointer

© 2010 Noah Mendelsohn urt0: A reuseable UMASM startup routine 29.zero r0.temps r7.section stk.space endstack:.section init _ustart: r0 := 0 r2 := endstack goto main linking r1 halt // Finis Invoke main() as a function! Trick: umasm urt0.ums mainprog.ums otherfile.ums Concatenates all source files before assembling.

© 2010 Noah Mendelsohn Startup code with stack (REVIEW)  Goals of startup code: –Set up useful registers –Set up a stack –Let UMASM know about register assignments –Invoke or drop through to user code 30.section init.temps r6, r7.section stk.space endstack:.section init.zero r0 start: r0 := 0 r2 := endstack goto main linking r1 halt init (instructions) start stk words endstack (r2) WE’VE NOW MADE A COMMON VERSION OF THIS IN urt0.ums!!

© 2010 Noah Mendelsohn 31 Writing A Function That Takes No Arguments We will use main() as an example

© 2010 Noah Mendelsohn Pushing and popping (Review)  Push macro  Pop macro 32.zero r0.temps r6, r7 push r3 on stack r2 r6 := -1 // macro instruction (not shown) r2 := r2 + r6 m[r0][r2] := r3 Expands to: pop r3 off stack r2 r3 := m[r0][r2] r6 := 1 r2 := r2 + r6 Expands to: These instructions can be use to save and restore registers and to put function arguments on the stack!

© 2010 Noah Mendelsohn Our program can go in the “main” function 33.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to # Can skip one or both of the next two if registers not needed push r3 on stack r2 // Callee saves non-volatile reg push r4 on stack r2 // Callee saves non-volatile reg...DO REAL WORK HERE... output "Hello world\n" pop r4 off stack r2 // restore caller's r4 pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5

© 2010 Noah Mendelsohn Now our program can go in the “main” function 34.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to # Can skip one or both of the next two if registers not needed push r3 on stack r2 // Callee saves non-volatile reg push r4 on stack r2 // Callee saves non-volatile reg...DO REAL WORK HERE... output "Hello world\n" pop r4 off stack r2 // restore caller's r4 pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5 Remember…we’re assuming urt0 has set up stack and r0

© 2010 Noah Mendelsohn Now our program can go in the “main” function 35.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to # Can skip one or both of the next two if registers not needed push r3 on stack r2 // Callee saves non-volatile reg push r4 on stack r2 // Callee saves non-volatile reg...DO REAL WORK HERE... output "Hello world\n" pop r4 off stack r2 // restore caller's r4 pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5 Here’s where our useful work will go.

© 2010 Noah Mendelsohn Now our program can go in the “main” function 36.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to # Can skip one or both of the next two if registers not needed push r3 on stack r2 // Callee saves non-volatile reg push r4 on stack r2 // Callee saves non-volatile reg...DO REAL WORK HERE... output "Hello world\n" pop r4 off stack r2 // restore caller's r4 pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5 Return address saved from r1..but popped into R5

© 2010 Noah Mendelsohn Now our program can go in the “main” function 37.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to # Can skip one or both of the next two if registers not needed push r3 on stack r2 // Callee saves non-volatile reg push r4 on stack r2 // Callee saves non-volatile reg...DO REAL WORK HERE... output "Hello world\n" pop r4 off stack r2 // restore caller's r4 pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5 Only save non-volatiles if necessary

© 2010 Noah Mendelsohn Now our program can go in the “main” function 38.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to # Can skip one or both of the next two if registers not needed push r3 on stack r2 // Callee saves non-volatile reg push r4 on stack r2 // Callee saves non-volatile reg...DO REAL WORK HERE... output "Hello world\n" pop r4 off stack r2 // restore caller's r4 pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5 umasm urt0.ums hello.ums...more ums here if needed... > hello Remember to combine all the pieces when you build

© 2010 Noah Mendelsohn 39 UMASM Function Arguments and the Call Stack

© 2010 Noah Mendelsohn UMASM function arguments (Review)  At entry to a function, the stack must look like this m[r0][r2 + 0] // first parameter m[r0][r2 + 1] // second parameter m[r0][r2 + 2] // third parameter  The called function may push more data on the stack which will change r2…  …when this happens the parameters will appear to be at larger offsets from the (modified) r2! 40 At function entry ???? r2 Third parm Second parm First parm

© 2010 Noah Mendelsohn 41 Recursion: Factorial

© 2010 Noah Mendelsohn Can we build UMASM code for this? 42 int fact(int n) { if (n == 0) return 1; else return n * fact(n - 1); } REMEMBER THE CONTRACT BETWEEN CALLER AND CALLEE!

© 2010 Noah Mendelsohn Summarizing calling conventions  r0 is always 0 (points to segment 0, used for goto and for the call stack)  r1 is the return-address register and the result register  r2 is the stack pointer and must have same value after return (nonvolatile)  r3, r4 are nonvolatiles  r5 is a helpful volatile register (good to copy return address)  r6, r7 are volatile and are normally.temps (but up to each procedure)  Arguments are on the stack, with first argument at the “bottom” (where r2 points) 43 EXCEPT FOR VOLATILES AND RESULT REGISTER CALLEE MUST LEAVE REGISTERS AND STACK UNCHANGED!

© 2010 Noah Mendelsohn Factorial 44.section init.zero r0.temp r6,r7 fact: push r1 on stack r2 // save return address push r3 on stack r2 // save nonvolatile register r3 := m[r0][r2 + 2] // first parameter 'n' // Handle base case: if n == 0 if (r3 == 0) goto base_case // recursive call r5 := r3 - 1 // OK to step on volatile register push r5 on stack r2 // now (n - 1) is a paraemter goto fact linking r1 // make the recursive call pop stack r2 // pop parameter r1 := r3 * r1 // n * fact(n-1) goto finish_fact base_case: r1 := 1 finish_fact: // What are invariants here? pop r3 off stack r2 // restore saved register pop r5 off stack r2 // put return address in r5 goto r5

© 2010 Noah Mendelsohn Factorial 45.section init.zero r0.temp r6,r7 fact: push r1 on stack r2 // save return address push r3 on stack r2 // save nonvolatile register r3 := m[r0][r2 + 2] // first parameter 'n' // Handle base case: if n == 0 if (r3 == 0) goto base_case // recursive call r5 := r3 - 1 // OK to step on volatile register push r5 on stack r2 // now (n - 1) is a paraemter goto fact linking r1 // make the recursive call pop stack r2 // pop parameter r1 := r3 * r1 // n * fact(n-1) goto finish_fact base_case: r1 := 1 finish_fact: // What are invariants here? pop r3 off stack r2 // restore saved register pop r5 off stack r2 // put return address in r5 goto r5 What are the invariants here?

© 2010 Noah Mendelsohn fact:.LFB2: pushq %rbx.LCFI0: movq %rdi, %rbx movl $1, %eax testq %rdi, %rdi je.L4 leaq -1(%rdi), %rdi call fact imulq %rbx, %rax.L4: popq %rbx ret Compare with the AMD 64 version 46 int fact(int n) { if (n == 0) return 1; else return n * fact(n - 1); }

© 2010 Noah Mendelsohn Main program to call factorial 47 void main(void) { for (int i = 10; i != 0; i--) { print_d(i); puts("! == "); print_d(fact(i)); putchar('\n'); } return EXIT_SUCCESS; }

© 2010 Noah Mendelsohn Main program to call factorial 48.zero r0.temps r6,r7 main: push r1 on stack r2 // Address main returns to push r3 on stack r2 // Callee saves non-volatile reg r3 := 10 // Using r3 for i - loop var goto main_test main_loop: push r3 on stack r2 // Pass i as first arg to function goto print_d linking r1 // Call print_d pop stack r2 // Function leaves arg on stack output "! == " // puts() is inlined push r3 on stack r2 // Pass i as first arg to function goto fact linking r1 // Call fact push r1 on stack r2 // Pass fact's return value as arg goto print_d linking r1 // Call print_d r2 := r2 + 2 // pop 2 parameters, 1 for each call output '\n' // putc() inlined --- it's a C macro r3 := r3 - 1 // i-- main_test: if (r3 != 0) goto main_loop // while (i != 0) pop r3 off stack r2 // restore caller's r3 pop r5 off stack r2 // get return address r1 := 0 // EXIT_SUCCESS is the result goto r5 // return