TEST optimize

Index > Macroinstructions > TEST optimize

Author

Thread

l4m2

Joined: 15 Jan 2015
Posts: 670

l4m2 06 Sep 2015, 02:44

usually we use a dword value to store some bits and get one by

Code:

ATTR equ 0x2000
;...
test dword [esi], dword ATTR

but obviously it's a better choice if we use

Code:

test byte [esi+1], 0x20

So I wrote this. Bugs can be reported here.

Code:

macro test op1, op2 {
    local tmp1, tmp2
    tmp1 equ *
    match w[x], :op1 \{
        tmp1 equ x
    \}
    tmp2 equ op2
    match w x, op2 \{
        if w eqtype word
            tmp2 equ x
        end if
    \}
    if ~(tmp1 eq *) & tmp2 eqtype 0
        if ~(tmp2 and not 0xff)
            test byte [tmp1], tmp2
        else if ~(tmp2 and not 0xff00)
            test byte [tmp1+1], tmp2 shr 8
        else if ~(tmp2 and not 0xff0000)
            test byte [tmp1+2], tmp2 shr 16
        else if ~(tmp2 and not 0xff000000)
            test byte [tmp1+3], tmp2 shr 24
        else
            test op1, op2
        end if
    else
        test op1, op2
    end if
}

06 Sep 2015, 02:44

sinsi

Joined: 10 Aug 2007
Posts: 794
Location: Adelaide

sinsi 06 Sep 2015, 04:02

Unaligned memory access makes it slower. Why not use BT?

06 Sep 2015, 04:02

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20761
Location: In your JS exploiting you and your system

revolution 06 Sep 2015, 04:14

There are already some existing macros for this (and LEA) optimisations. They are from a few years ago but a bit of searching should be able to find them.

06 Sep 2015, 04:14

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20761
Location: In your JS exploiting you and your system

revolution 06 Sep 2015, 07:23

For your viewing convenience I found the old thread:

http://board.flatassembler.net/topic.php?t=1238

Please read the whole thread, there is a lot of thoughts and information presented there.

06 Sep 2015, 07:23

l4m2

Joined: 15 Jan 2015
Posts: 670

l4m2 06 Sep 2015, 10:47

sinsi wrote:

Unaligned memory access makes it slower. Why not use BT?

Code:

bt dword [ebx], 1  ;4 bytes
test byte [ebx], 2 ;3 bytes

06 Sep 2015, 10:47

l4m2

Joined: 15 Jan 2015
Posts: 670

l4m2 11 Sep 2015, 15:20

It's not a good idea that fasm itself compile TEST [EBX], DWORD 0x20 into TEST [EBX], BYTE 0x20. In a segment 0-18, if EBX is 16, the first one rise a fault while the second one doesn't.

11 Sep 2015, 15:20

< Last Thread | Next Thread >

Forum Rules:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum