1 / 19

# CSE 131B – Compiler Construction II - PowerPoint PPT Presentation

CSE 131B – Compiler Construction II. Discussion 7: Short-Circuiting, Pointers/Records, and Arrays 2/28/2007. Overview. Phase 2 Phase 3. Short-Circuiting. & and OR are "short-circuiting" operators. In A & B, if A evaluates to FALSE, B is not evaluated.

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.

## PowerPoint Slideshow about ' CSE 131B – Compiler Construction II' - dylan

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

### CSE 131B – Compiler Construction II

Discussion 7: Short-Circuiting, Pointers/Records, and Arrays

2/28/2007

• Phase 2

• Phase 3

• & and OR are "short-circuiting" operators.

• In A & B, if A evaluates to FALSE, B is not evaluated.

• In A OR B, if A evaluates to TRUE, B is not evaluated.

• Think of how you handle an if-else statement.

• Short-circuiting follows the same principle:

• In the A & B case, branch to the false case if A is FALSE. Else, check B.

• In the A OR B case, branch to the true case if A is TRUE. Else, check B.

! e2 is TRUE, so result TRUE

mov 1, %l5

ba endlabel

nop

flabel:

mov 0, %l5

endlabel:

! Result is now in %l5

• Oberon: e1 & e2:

! Load e1 and check if FALSE

ld [%fp-4], %l0 ! e1’s temp

cmp %l0, %g0

be flabel

nop

! e1 is TRUE, so check e2

ld [%fp-8], %l0 ! e2’s temp

cmp %l0, %g0

be flabel

nop

• When you declare an array, one possible way to allocate space for it is to allocate an entire chunk in BSS and have the variable label at the end of it:

VAR X : ARRAY 7 OF INTEGER;

.section “.bss”

.align 4

.skip 28 ! 7 * sizeof(int)

X:

Now X[0] is at X-4, X[1] is at X-8, and so on.

• Similarly, if you are taking the single “globals” label approach, you just need to .skip/.space the total size of the array structure.

• A useful attribute to have for Arrays and Records is “size”, so you know how much space to allocation.

• Offsets would also be useful. For arrays, the offsets are simply multiples of the element’s size. For records, the offsets are the collective sizes of the preceding fields.

• a := X[b] + 7; ! X is array of INTEGER

set X, %l1 ! X  base address

set b, %l0 ! b

sll %l0, 2, %l0 ! (b++) * 4  offset

add %l1, %l0, %l0 ! Base + offset

ld [%l0], %l0 ! X[b]’s value

add %l0, 7, %l0 ! X[b] + 7

set a, %l1

st %l0, [%l1] ! a := X[b] + 7

• Very similar to array usage.

• You need to start at the base address of where the record is located.

• Then, you have to move some offset to get to a specific field you are interested in.

• Once at that location, you either load or store, depending on what you wanted to do.

• Arrays must be passed by reference (VAR), where the address to the address is sent to the procedure.

PROCEDURE baz (VAR a : arrType)

baz(myArr);

• In the above case, you would store the base address + offset of the first element in %o0. Once in baz, %i0 will have the address of the first element. All other elements will be accessed by some offset from that first element address.

• Since Records always have a pointer to them, Records can be passed by reference (VAR) or value.

FUNCTION bar1 (VAR r : recPtrType)

FUNCTION bar2 (r : recPtrType)

bar1(myRecPtr);

bar2(myRecPtr);

• The behavior is essentially the same, but when using VAR, there is a double dereference (one for the VAR, one for the record’s . operator)

• If you have any doubts about value vs. VAR parameters or parameter passing, please look at the following URL:

• http://www.cse.ucsd.edu/classes/wi07/cse131b/VARvsValue.pdf

• Quick note: Pointers always have a size of 4 bytes. Pointers will point to another address, which can be of a different size (the size of the record object once dynamically created). So, the value of a pointer variable is an address, and that address is the base of the object being pointed to.

• Consider p := q;

• This is just copying the address that is in q into p.

ld [%fp-4], %l0 ! Assuming q is in fp-4

st %l0, [%fp-8]

• Consider x := myRPtr.field;

ld [%fp-4], %l0 ! Load myRPtr

! %l0 now contains the base address of the record

set 4, %l1 ! Field at offset of 4

ld [%l2], %l0 ! Load field’s value

st %l0, [%fp-8] ! Store value in x

TYPE ptrType = POINTER TO RECORD x,y : INTEGER; END;

VAR myGlobal : ptrType;

FUNCTION foo() : INTEGER;

VAR myLocal : ptrType;

BEGIN

NEW(myLocal);

myLocal.y := 420;

RETURN 7;

END foo;

BEGIN

NEW(myGlobal);

myGlobal.x := foo();

WRITE myGlobal.x;

END.

call calloc ! NEW

nop

st %o0, [%fp-4]

! myLocal.y := 420

set 420, %l0

ld [%fp-4], %l1

set 4, %l2 ! y’s offset

st %l0, [%l1]

! RETURN 7

set 7, %i0

ret

restore

foo.SAVE = -(92 + 4) & -8

.global main

main:

save %sp, -96, %sp

set globals, %g7

! NEW(myGlobal)

(same code as other NEW)

! myGlobal.x := foo()

call foo

nop

ld [%g7-4], %l0

set 0, %l1 ! x’s offset

st %o0, [%l0]

! OUTPUT myGlobal.x

set ifmt, %o0

ld [%g7-4], %l0

set 0, %l1 ! x’s offset

ld [%l0], %o1

call printf

nop

.section ".bss"

.align 4

.skip 4

globals:

.section ".rodata"

.align 4

ifmt: .asciz "%d"

.section ".text"

.align 4

.global foo

foo:

set foo.SAVE, %g1

save %sp, %g1, %sp

! NEW(myLocal)

set 1, %o0 ! numelem

set 8, %o1 ! sizeof(rec)

• Finish Phase 2.

• Start of Phase 3.

• Thoroughly test and re-test Phase 1, 2, and 3.

• Come to lab hours and ask questions.

• Anything else you would like me to go over now?

• Anything in particular you would like to see next week?