jecxz "jump out of range" --> first solution

Index > Compiler Internals > jecxz "jump out of range" --> first solution

Goto page 1, 2 Next

Author

Thread

ouadji

Joined: 24 Dec 2008
Posts: 1081
Location: Belgium

ouadji 02 Apr 2010, 22:41

jecxz : relative jump out of range.

sorry for my english, i will try to do my best.

It would be very interesting to also display the number of bytes exceeded.
This would allow to know if it's worth trying to optimize the code ... or not.
Indeed, if the excess is only a few bytes,
it's maybe possible to optimize the code so that the limits are no longer exceeded.
or ... if the number of bytes beyond the limits is too large,
it indicates that a simple optimization isn't enough to eliminate the problem.

jecxz : relative jump out of range (8bytes)

_________________
I am not young enough to know everything (Oscar Wilde)-

Last edited by ouadji on 19 Jun 2010, 12:03; edited 2 times in total

02 Apr 2010, 22:41

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20627
Location: In your JS exploiting you and your system

revolution 03 Apr 2010, 03:25

This can jump up to +-2GB.

Code:

test ecx,ecx
jz Somewhere

If you need to jump further than that then you are probably doing something wrong. Wink

03 Apr 2010, 03:25

sinsi

Joined: 10 Aug 2007
Posts: 794
Location: Adelaide

sinsi 03 Apr 2010, 03:34

If you need to optimize, using jecxz (and loop) are much slower than a test/jump (or a dec/jnz).

03 Apr 2010, 03:34

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4624
Location: Argentina

LocoDelAssembly 03 Apr 2010, 04:01

On AMD, JECXZ has a two cycles latency and it is a DirectPath instruction. Maybe JECXZ is not always as fast as TEST/J(N)Z is, but still, there may be situations in which you don't want to destroy the flags to find out ECX's current status.

03 Apr 2010, 04:01

Tyler

Joined: 19 Nov 2009
Posts: 1216
Location: NC, USA

Tyler 03 Apr 2010, 04:48

Tomasz uses loop in "convert_line" in preproce.inc

Code:

...
  mov     ebx,characters
        xor     ecx,ecx
      convert_symbol:
        lods    byte [esi]
        stos    byte [edi]
        xlat    byte [ebx]
        or      al,al
        loopnzd convert_symbol
        neg     ecx
        cmp     ecx,255         
...

03 Apr 2010, 04:48

sinsi

Joined: 10 Aug 2007
Posts: 794
Location: Adelaide

sinsi 03 Apr 2010, 04:52

>Tomasz uses loop in "convert_line" in preproce.inc
Also uses xlat...

Just an observation, not a hard-and-fast rule, since it depends on the cpu.
P4 is good with dec and bad with loop and jcxz. P3 is slower with dec.

03 Apr 2010, 04:52

ouadji

Joined: 24 Dec 2008
Posts: 1081
Location: Belgium

ouadji 03 Apr 2010, 06:35

revolution wrote:

This can jump up to +-2GB.
Code:
test ecx,ecx
jz Somewhere    
If you need to jump further than that then you are probably doing something wrong.

thank you revolution ... I didn't know that !
sorry, but i'm a beginner in asm. Laughing

Quote:

there may be situations in which you don't want to destroy the flags

good answer Wink

_________________
I am not young enough to know everything (Oscar Wilde)-

Last edited by ouadji on 03 Apr 2010, 06:43; edited 1 time in total

03 Apr 2010, 06:35

rugxulo

Joined: 09 Aug 2005
Posts: 2341
Location: Usono (aka, USA)

rugxulo 03 Apr 2010, 06:40

sinsi wrote:

>Tomasz uses loop in "convert_line" in preproce.inc
Also uses xlat...

Also the slow string instructions (lodsb, stosb). Wasn't he writing on a real 386 at one time back in the day?? I'm sure it's smarter to target slow processors that need the extra speed than fast processors that already have it. Wink

03 Apr 2010, 06:40

sinsi

Joined: 10 Aug 2007
Posts: 794
Location: Adelaide

sinsi 03 Apr 2010, 07:45

rugxulo, no argument here, although I haven't programmed recently on less then a P3.
I just latched onto the 'optimize' in the OP. Go to the MASM32 forum and look at all the optimization that goes on there... Rolling Eyes

03 Apr 2010, 07:45

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20627
Location: In your JS exploiting you and your system

revolution 03 Apr 2010, 08:47

The word "optimise" is so ambiguous. It can mean many things. Optimising for what? Speed? Size? Ease of coding? Ease of reading? Which platform? Which CPU? Under what usage conditions? Hehe, optimise is a dirty word!

03 Apr 2010, 08:47

ouadji

Joined: 24 Dec 2008
Posts: 1081
Location: Belgium

ouadji 03 Apr 2010, 10:38

no, "optimise" isn't a dirty word.
No matter why we search a optimization,
speed, size, ease of coding, ease of reading ...
it depends on the coding spirit of each developper
everyone has his own choice
but "car" is a dirty word .. bus, truck, van ? Wink

_________________
I am not young enough to know everything (Oscar Wilde)-

03 Apr 2010, 10:38

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20627
Location: In your JS exploiting you and your system

revolution 03 Apr 2010, 11:18

Please help me to optimise my code:

Code:

;pretend there is some code here

Okay, so what did I ask for there? No one can possibly know without further information.

03 Apr 2010, 11:18

bitshifter

Joined: 04 Dec 2007
Posts: 796
Location: Massachusetts, USA

bitshifter 03 Apr 2010, 11:25

I help you optimize it for size and readability and speed and everything else...

Code:

Hehe

03 Apr 2010, 11:25

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20627
Location: In your JS exploiting you and your system

revolution 03 Apr 2010, 11:29

Wow, it compiled nicely. Thanks bitshifter.

Quote:

flat assembler version 1.69.12 (152449 kilobytes memory)
1 passes, 0 bytes.

03 Apr 2010, 11:29

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4624
Location: Argentina

LocoDelAssembly 03 Apr 2010, 16:20

sinsi wrote:

P4 is good with dec

P4 is bad in pretty much everything including DEC, you must use "SUB reg/mem, 1" instead.

Agner Fog wrote:

16.2 INC and DEC (all Intel processors)

The INC and DEC instructions do not modify the carry flag but they do modify the other
arithmetic flags. Writing to only part of the flags register costs an extra µop on P4 and P4E.
It can cause a partial flags stalls on other Intel processors if a subsequent instruction reads
the carry flag or all the flag bits. On all processors, it can cause a false dependence on the
carry flag from a previous instruction.
Use ADD and SUB when optimizing for speed. Use INC and DEC when optimizing for size or
when no penalty is expected.

The part marked in bold I'm not sure if it is really correct.

03 Apr 2010, 16:20

ouadji

Joined: 24 Dec 2008
Posts: 1081
Location: Belgium

ouadji 04 Apr 2010, 08:04

LocoDelAssembly,
I don't quite understand, tell you about this?

Code:


mov eax,7
mov ebx,5
x_:
sub eax,1
dec ebx
jc somewhere
jmp x_

_________________
I am not young enough to know everything (Oscar Wilde)-

04 Apr 2010, 08:04

zhak

Joined: 12 Apr 2005
Posts: 501
Location: Belarus

zhak 04 Apr 2010, 10:19

yes, in this case jump will be taken if CF was set by SUB EAX, 1

04 Apr 2010, 10:19

SeproMan

Joined: 11 Oct 2009
Posts: 70
Location: Belgium

SeproMan 10 Apr 2010, 14:00

I find the original suggestion by Ouadji a very useful one.
In fact I've incorporated it just a few days ago into my own assembler language project by showing an error message like "Target out of reach by 2 bytes"

Thanks Ouadji

In FASM "jz Somewhere" has a range of +-2GB. This is not intuitive.
I advocate a more WYSIWYG approach
jz Somewhere ;[-128,127]
jzw Somewhere ;[-32768,32767]
jzd Somewhere ;[-2GB,2GB-1]

_________________
Real Address Mode.

10 Apr 2010, 14:00

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20627
Location: In your JS exploiting you and your system

revolution 10 Apr 2010, 14:02

in fasm syntax:

jz byte Somewhere ;[-128,127] [edit] byte doesn't compile. What is the byte override?
jz word Somewhere ;[-32768,32767]
jz dword Somewhere ;[-2GB,2GB-1]

The size override forces it to a particular length.

10 Apr 2010, 14:02

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8394
Location: Kraków, Poland

Tomasz Grysztar 10 Apr 2010, 14:08

Actually, this is not the fasm syntax. The fasm syntax is:

Code:

jz short word Somewhere ; [-128;127], EIP masked with 0FFFFh
jz short dword Somewhere ; [-128;127]
jz near word Somewhere ; [-32768,32767], EIP masked with 0FFFFh
jz near dword Somewhere ; [-2GB,2GB-1]

10 Apr 2010, 14:08

Goto page 1, 2 Next

< 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