Mechanisms in Procedures. CS429: Computer Organization and Architecture. x86-64 Stack. x86-64 Stack Pushing

Size: px

Start display at page:

Download "Mechanisms in Procedures. CS429: Computer Organization and Architecture. x86-64 Stack. x86-64 Stack Pushing"

Elwin Wilkerson
5 years ago
Views:

1 CS429: Computer Organization and Architecture Dr. Bill Young Department of Computer Sciences University of Texas at Austin Last updated: February 28, 2018 at 06:32 Mechanisms in Procedures Passing Control To beginning of procedure code Back to urn point Passing Data Procedure arguments Return value Memory Management Allocate during procedure execution Deallocate upon urn Mechanisms all implemented with machine instructions x86-64 implementation of a procedure uses only those mechanisms required. P() y = Q(x) ; printf (y) ; int Q( int i ) int t = 3 i ; int v [10]; urn v[ t ]; CS429 Slideset 9: 1 CS429 Slideset 9: 2 x86-64 x86-64 Pushing "bottom" "bottom" Region of memory managed with stack discipline Grows toward lower addresses Increasing Addresses. Increasing Addresses Register contains lowest stack address address of top element "top" Pushing pushq Src Decrement by 8 Write operand at address given by pointer 8 "top" Grows Down CS429 Slideset 9: 3 CS429 Slideset 9: 4

x86-64 Popping Code Examples "bottom" void multstore (long x, long y, long dest ) long t = mult2(x, y) ; dest = t ; Popping popq Dest Read operand at address given by Increment by 8 pointer +8 "top"

2 x86-64 Popping Code Examples "bottom" void multstore (long x, long y, long dest ) long t = mult2(x, y) ; dest = t ; Popping popq Dest Read operand at address given by Increment by 8 pointer +8 "top" Increasing Addresses Grows Down <multstore >: : push # Save : mov %rdx, # Save dest : q <mult2> # mult2(x, y) 40054D: mov %rax, () # Save at dest : pop # Restore : q # Return Write to Dest CS429 Slideset 9: 5 CS429 Slideset 9: 6 Code Examples (2) Procedure Control Flow long mult2 (long a, long b) long s = a b; urn s ; <mult2 >: : mov %rdi, %rax # a : imul %rsi, %rax # a b : q # Return Use stack to support procedure and urn Procedure : label Push urn address on stack Jump to label Return address: Address of next instruction right after Procedure urn: Pop address from stack Jump to address CS429 Slideset 9: 7 CS429 Slideset 9: 8

3 Control Flow Example #1 Control Flow Example #2 Poised to execute in multstore. After in multstore. Now in mult <multstore >: : q <mult2> 40054D: mov %rax, (). 0x130. 0x128. 0x <multstore >: : q <mult2> 40054D: mov %rax, (). 0x130. 0x128. 0x120 0x118 0x40054D <mult2 >: : mov %rdi, %rax : q %rip 0x120 0x <mult2 >: : mov %rdi, %rax : q %rip 0x118 0x CS429 Slideset 9: 9 CS429 Slideset 9: 10 Control Flow Example #3 Control Flow Example #4 Execute through mult2 to q. After q in mult2. Back in multstore <multstore >: : q <mult2> 40054D: mov %rax, (). 0x130. 0x128. 0x120 0x118 0x40054D <multstore >: : q <mult2> 40054D: mov %rax, (). 0x130. 0x128. 0x <mult2 >: : mov %rdi, %rax : q %rip 0x118 0x <mult2 >: : mov %rdi, %rax : q %rip 0x120 0x40054D CS429 Slideset 9: 11 CS429 Slideset 9: 12

4 Procedure Data Flow Data Flow Examples Registers: First 6 arguments %rdi %rsi %rdx %rcx %r8 %r9 Mnemonic to remember the order: Diane s silk dress cost $89. Args past 6 are pushed on the stack in reverse order. Return value %rax Arg n Arg 8 Arg 7 Only allocate stack space when needed. void multstore (long x, long y, long dest ) long t = mult2(x, y) ; dest = t ; <multstore >: # x in %rdi, y in %rsi, dest in %rdx : mov %rdx, # Save dest : q <mult2> # mult2(x, y) # t in %rax 40054C: mov %rax, () # Save at dest CS429 Slideset 9: 13 CS429 Slideset 9: 14 Data Flow Examples (continued) -Based Languages long mult2 (long a, long b) long s = a b; urn s ; <mult2 >: # a in %rdi, b in %rsi : mov %rdi, %rax # a : imul %rsi, %rax # a b # s in %rax : q # Return Languages that support recursion e.g., C, Pascal, Java Code must be reentrant : multiple simultaneous instantiations of single procedure Need some place to store state of each instantation: arguments, local variables, urn pointer discipline State for given procedure needed for a limited time: from to urn Callee urns before er does allocated in Frames State for a single procedure instantiation CS429 Slideset 9: 15 CS429 Slideset 9: 16

5 Call Chain Example Frames Code Structure () () ; () () ; () ; () () ; CS429 Slideset 9: 17 Procedure is recursive. In IA32, almost every procedure generated a stack frame. In x86-64, most s do not generate an explicit frame. Contents Return information Local storage (if needed) Temporary space (if needed) Management Space allocated when enter procedure Set-up code Includes push by instruction Deallocated when urning Finish code Includes pop by instruction CS429 Slideset 9: 18 Previous Frame Frame for proc top Frame pointer: %rbp (optional) pointer: Example (1) Example (2) () () ; () () ; () () ; CS429 Slideset 9: 19 CS429 Slideset 9: 20

6 Example (3) Example (4) () () () () ; () () () () ; CS429 Slideset 9: 21 CS429 Slideset 9: 22 Example (5) Example (6) () () () () ; () () () () ; CS429 Slideset 9: 23 CS429 Slideset 9: 24

7 Example (7) Example (8) () () () () ; () () ; () () ; CS429 Slideset 9: 25 CS429 Slideset 9: 26 Example (9) Example (10) () () () () ; () () ; () () ; CS429 Slideset 9: 27 CS429 Slideset 9: 28

8 Example (11) x86-64/linux Frame () () ; Current Frame ( Top to Bottom) Argument build: parameters for function about to Local variables, if can t keep in registers Saved register context Old frame pointer (optional) Caller frame Arguments 7+ Return addr Frame pointer Old %rbp %rbp (optional) pointer Saved Register + Local vars Argument Build (optional) Caller Frame Return address (pushed by instruction) Arguments for this CS429 Slideset 9: 29 CS429 Slideset 9: 30 Building a Frame Example: incr IA32 routines almost always constructed an explicit stack frame; x86-64 routines usually don t. If you do build a stack frame: proc : pushq %rbp movq, %rbp # save er s frame base # set ee s frame base # body of routine proc This example doesn t use explicit frames! long incr (long p, long val ) long x = p; long y = x + val ; p = y; urn x; movq %rbp, # discard ee s frame popq %rbp # reset er s frame base Reserve %rbp for this purpose if you re doing this. What does this do? incr : movq (%rdi ), %rax %rax, %rsi movq %rsi, (%rdi ) %rdi Argument p %rsi Argument val, y %rax x, urn value CS429 Slideset 9: 31 CS429 Slideset 9: 32

9 Example: Calling incr #1 Example: Calling incr #2 long incr () long v2 = incr(&v1, 3000) ; urn v1 + v2; incr : subq $16, movq $15213, 8( ) movl leaq $3000, %esi 8( ), %rdi incr 8( ), %rax $16, Initial Structure Rtn addr Resulting Structure Rtn addr Unused long incr () long v2 = incr(&v1, 3000) ; urn v1 + v2; incr : subq $16, movq $15213, 8( ) movl leaq $3000, %esi 8( ), %rdi incr 8( ), %rax $16, Structure Rtn addr Unused %rdi &v1 %rsi 3000 CS429 Slideset 9: 33 CS429 Slideset 9: 34 Example: Calling incr #3 Example: Calling incr #4 long incr () long v2 = incr(&v1, 3000) ; urn v1 + v2; incr : subq $16, movq $15213, 8( ) movl $3000, %esi leaq 8( ), %rdi incr 8( ), %rax $16, Structure Rtn addr Unused %rdi &v1 %rsi 3000 long incr () long v2 = incr(&v1, 3000) ; urn v1 + v2; incr : subq $16, movq $15213, 8( ) movl $3000, %esi leaq 8( ), %rdi incr 8( ), %rax $16, Structure Rtn addr Unused %rdi &v1 %rsi 3000 Updated Structure Rtn addr CS429 Slideset 9: 35 CS429 Slideset 9: 36

10 Example: Calling incr #5 long incr () long v2 = incr(&v1, 3000) ; urn v1 + v2; incr : subq $16, movq $15213, 8( ) movl $3000, %esi leaq 8( ), %rdi incr 8( ), %rax $16, Updated Structure Rtn addr %rax Return value Final Structure Register Saving Conventions When procedure s : is the er is the ee Can register be used for temporary storage? : movq $15213, %rdx %rdx, %rax : subq $18213, %rdx Contents of register %rdx are overwritten by This could be trouble; something should be done! Need to coordinate between er and ee. CS429 Slideset 9: 37 CS429 Slideset 9: 38 Register Saving Conventions When procedure s : is the er is the ee Can register be used for temporary storage? Conventions Caller Saved : Caller saves temporary values in its frame before the Callee Saved : Callee saves temporary values in its frame before using Callee restores them before urning to er x86-64 Linux Caller-saved Registers %rax Return value Also er-saved Can be modified by procedure %rdi,%rsi,%rdx,%rcx,%r8,%r9 Arguments Also er-saved Can be modified by procedure %r10, %r11 Caller-saved Can be modified by procedure Return value %rax Argument 1 %rdi Argument 2 %rsi Argument 3 %rdx Argument 4 %rcx Argument 5 %r8 Argument 6 %r9 Caller-saved %r10 temporaries %r11 CS429 Slideset 9: 39 CS429 Slideset 9: 40

11 x86-64 Linux Callee-saved Registers, %r12, %r13, %r14 %rbp Callee-saved Callee must save and restore Callee-saved Callee must save and restore May be used as frame pointer Can mix and match Special form of ee-saved Restored to original value upon exit from procedure Callee-saved %r12 Temporaries %r13 %r14 Special %rbp Special Callee-Saved Example #1 long incr2 ( long x ) long v2 = incr ( &v1, 3000 ) ; urn x + v2; incr2 : pushq subq $16, movq %rdi, movq $15213, 8( ) movl $3000, %esi leaq 8( ), %rdi incr, %rax $16, popq Initial Structure Rtn address Resulting Structure Rtn address Saved Unused CS429 Slideset 9: 41 CS429 Slideset 9: 42 Callee-Saved Example #2 Recursive Function long incr2 ( long x ) long v2 = incr ( &v1, 3000 ) ; urn x + v2; incr2 : pushq subq $16, movq %rdi, movq $15213, 8( ) movl $3000, %esi leaq 8( ), %rdi incr, %rax $16, popq Resulting Structure Rtn address Saved Unused Pre-urn Structure Rtn address urn 0; urn (x & 1) pushq movq %rdi, andl $1, %ebx shrq $1, %rdi pcount r, %rax popq CS429 Slideset 9: 43 CS429 Slideset 9: 44

12 Recursive Function Terminal Case Recursive Function Register Save urn 0; urn (x & 1) %rdi Argument x %rax Return value pushq movq andl shrq popq %rdi, $1, %ebx $1, %rdi pcount r, %rax urn 0; urn (x & 1) Rtn address Saved pushq movq andl shrq popq %rdi, $1, %ebx $1, %rdi pcount r, %rax %rdi Argument x %rax Return value CS429 Slideset 9: 45 CS429 Slideset 9: 46 Recursive Function Call Setup Recursive Function Call urn 0; urn (x & 1) Type %rdi x >> 1 Rec. argument %rax x & 1 Caller-saved pushq movq andl shrq popq %rdi, $1, %ebx $1, %rdi pcount r, %rax urn 0; urn (x & 1) Type x & 1 Callee-saved %rax Recursive urn value pushq movq %rdi, andl $1, %ebx shrq $1, %rdi popq pcount r, %rax CS429 Slideset 9: 47 CS429 Slideset 9: 48

13 Recursive Function Result Recursive Function Completion urn 0; urn (x & 1) Type x & 1 Callee-saved %rax Return value pushq movq %rdi, andl $1, %ebx shrq $1, %rdi pcount r popq, %rax urn 0; urn (x & 1) Type %rax Return value pushq movq %rdi, andl $1, %ebx shrq $1, %rdi pcount r, %rax popq CS429 Slideset 9: 49 CS429 Slideset 9: 50 Observations About Recursion Handled without Special Consideration frames mean that each function has private storage Saved registers and local variables Saved urn pointer Register saving conventions prevent one function from corrupting another s data Unless the C code explicitly does so (e.g., buffer overflow) discipline follows /urn pattern If P s Q, then Q urns before P Last-In, First-Out Also works for mutual recursion P s Q; Q s P x86-64 Procedure Summary Important Points is the right data structure for procedure / urn; if P s Q, then Q urns before P Recursion (and mutual recursion) are handled by normal ing conventions Can safely store values in local stack frame and in ee-saved registers Put function arguments at top of stack Result urn in %rax Pointers are addresses of values (on stack or global) Caller frame Arguments 7+ Return addr Frame pointer Old %rbp %rbp (optional) Saved Register + Local vars Argument pointer Build (optional) CS429 Slideset 9: 51 CS429 Slideset 9: 52

CS429: Computer Organization and Architecture

CS429: Computer Organization and Architecture Dr. Bill Young Department of Computer Sciences University of Texas at Austin Last updated: February 28, 2018 at 06:32 CS429 Slideset 9: 1 Mechanisms in Procedures