Aligned memory comparison

[hutch--]

This algo is a DWORD compare that drops to a byte compare to test which byte is different, it exits with any of 3 return value to tell you the two bytes are equal, one is greater than the other or less than the other. It appeas to run OK and returns the correct results so far but it could probably do with some optimisation to get a bit more grunt out of it.

Code Select Expand


; «««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««

align 4

OPTION PROLOGUE:NONE
OPTION EPILOGUE:NONE

cmpmema proc buf1:DWORD,buf2:DWORD,bcnt:DWORD

; ---------------------------------------
; compare 2 4 byte aligned memory buffers
; ---------------------------------------

    push esi
    push edi

    mov edx, [esp+20]               ; bcnt
    shr edx, 2                      ; div by 4

    mov esi, [esp+12]               ; buf1
    mov edi, [esp+16]               ; buf2
    xor ecx, ecx

  align 4
  @@:
    mov eax, [esi+ecx]              ; DWORD compare main file
    cmp eax, [edi+ecx]
    jne nxt
    add ecx, 4
    sub edx, 1
    jnz @B

    mov edx, [esp+20]               ; bcnt  calculate any remainder
    and edx, 3
    jz match                        ; exit if its zero
    xor eax, eax
    jmp @F

  nxt:
    mov edx, 4                      ; set counter for 4 bytes
    xor eax, eax                    ; clear EAX for partial writes

  @@:
    mov al, [esi+ecx]               ; BYTE compare tail
    cmp al, [edi+ecx]
    jg greater
    jl lessthan
    add ecx, 1
    sub edx, 1
    jnz @B

    jmp match

  greater:
    mov eax, 1
    jmp quit

  lessthan:
    mov eax, -1
    jmp quit

  match:
    xor eax, eax                    ; return ZERO on match

  quit:
    pop edi
    pop esi

    ret 12

; -------------------------------------------------

cmpmema endp

OPTION PROLOGUE:PrologueDef
OPTION EPILOGUE:EpilogueDef

; «««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««««


[daydreamer]

thanks Hutch, got me started, now see if I can beat you
9 opcodes=16 compares of equal or less than or greater than

could we go even faster?

Code Select Expand

data
align 16
twos: db 16 dup (2)
minus1:db 16 dup (-1)
mask:db 16 dup (0FFh)
.code


;top of loop

movdqa xmm0,[esi+ecx] ;source1
movdqa xmm1,xmm0
;greater than or less than
pcmpgtb xmm1,[edi+ecx] ;source2
pand xmm1,twos
paddb xmm1,minus1 ; 2+ -1 = 1 or 0+-1 == -1
check for equal
pcmpeqb,xmm0,[edi+ecx] ;source2FF's = equal, 0 = nonequal
pxor, xmm0,mask ; FF becomes 0 (not)
pandb xmm1,xmm0 ;mask out when zero, all thats equal becomes zero
movdqa result,xmm1
;endwhile


[MichaelW]

I managed to reduce the execution time by 25%, using 486-compatible code, and I think without disturbing the functionality, but bcnt must be a multiple of 16.


[attachment deleted by admin]