Shellsort: Write NASM code from C - c

Question

I am trying to convert a C function to NASM in 32-bit.
Here is the C function
void shellsort (int *A, int n)
{
int gap, i, j, temp;
for (gap = n/2; gap > 0; gap /=2)
{
for (i= gap; i<n; i++)
{
for (j=i-gap; j>=0 && A[j] > A[j+gap]; j-=gap)
{
temp = A[j];
A[j] = A[j+gap]
A[j+gap] = temp;
}
}
}
}
Here is what I have done so far
%include "asm_io.inc"
SECTION .data
SECTION .bss
SECTION .text
extern printf
global sort
sort:
push ebp ;set up stack frame
mov ebp, esp
sub esp, 0x10 ;assign 16byte space for local variables
firstloop:
mov eax, [ebp+8] ;moves value of n into eax
mov edx, 0 ;prepares for division, remainder init
mov ecx, 2 ;divisor
div ecx ;division, store gap in eax
cmp eax, 0 ;compare if gap with zero
jle firstloopDone ;firstloopDone
firstloopDone:
div ecx
jmp done
secondloop:
mov ecx, eax ;copy value of gap into ecx, i=gap
cmp ecx, [ebp+8] ;compare i with 1st parameter n
jge secondloopDone ;jump to secondloop if greater-equal
inc ecx ;increment i
jmp thirdloop ;jump to thirdloop
secondloopDone:
inc ecx ;increment i
jmp firstloop ;jump to firstloop
thirdloop:
mov edx, ecx ;save i value
sub ecx, eax ;subtract gap from i and store in ecx
cmp ecx, 0 ;compare j with zero
jl thirdloopDone ;if not j>=0, then skip to end of loop
cmp [ebp+12+ecx], [ebp+12+ecx+eax] ;compare A[j] and A[j+gap]
jle thirdloopDone
sub ecx, eax ;subtract gap from j and store in ecx
jmp swap
thirdloopDone:
sub ecx, eax ;j-gap
jmp secondloop ;jump to second loop
swap:
mov edx, [ebp+12+ecx] ;copy A[j] to temp
mov [ebp+12+ecx], [ebp+12+ecx+eax] ;A[j]=A[j+gap]
mov [edp+12+ecx+eax], edx ;A[j+gap]= temp
jmp thirdloop
done:
leave ;destroy stack frame
ret
Needless to say, it doesn't work.
The error message says:
"error: beroset-p-603-invalid effective address"
at line 58 (thirdloop)
cmp [ebp+12+ecx], [ebp+12+ecx+eax]
and at line 71 (swap)
mov edx, [ebp+12+ecx]
I understand that this may be the wrong method. But from the nested loop in C code, I have too many variables to keep so I cant spare any of the registers.
Also, I suspect I may not have a proper understanding of the stack frame, which may show in my code.
Any help, or bug-discovery, would be greatly appreciated. Thanks.

Answer 1

You say you don't have enough registers to play with but you don't even use EBX, ESI, nor EDI. You made provisions for local variables yet you don't use any.
firstloop:
mov eax, [ebp+8] ;moves value of n into eax
mov edx, 0 ;prepares for division, remainder init
mov ecx, 2 ;divisor
div ecx ;division, store gap in eax
The previous code does not need the division. Simplify it with a shift operation. shr eax,1
The next code will always terminate. Use jle done
cmp eax, 0 ;compare if gap with zero
jle firstloopDone ;firstloopDone
firstloopDone:
div ecx
jmp done
Aren't parameters pushed from right to left?
mov ebx, [ebp+8] ;pointer to array A
firstloop:
mov eax, [ebp+12] ;number of elemnets n
You would typically have to code
swap:
mov edx, [ebx+ecx*4] ;TEMP=A[j]
add ecx,eax
xchg edx, [ebx+ecx*4] ;Swap A[j+gap] with TEMP
sub ecx, eax
mov [ebx+ecx*4], edx ;A[j]=TEMP
jmp thirdloop