Extending my jumps?

[Metallic-Red]

I'm using assembly but my code has become too large that the jumps won't work because they are too far away. How can I fix this or set my code up so that it will work?

[IRBMe]

By giving your code a good structure. You shouldn't need to be jumping over huge sections of code left right and centre. That totally ruins all of the optimisation features built into the CPU to take advantage of the phenomenon of program locality.

For a start, try to split some of your lengthy bits of code into neat functions which can be called with a single instruction

Code: Select all

jmp start

foo:
  ; some code
  ret

bar:
  ; some code
  ret

start:
  call foo
  call bar
  ; and so on..

[Perica]

..

[Metallic-Red]

I am using the NASM compiler and my code is part of a DOS style OS I've been working on.

I have the code to set up my commands:

MOV SI, strCommand
MOV DI, cmdCls
MOV CX, 4
REPE CMPSB
JE COMMAND_CLS

MOV SI, strCommand
MOV DI, cmdDate
MOV CX, 5
REPE CMPSB
JE COMMAND_DATE

etc.

But I have 6 functions and some of them have lengthy code so some of these tests can't reach their designated functions. I hope I make sense?

[Perica]

..

[Metallic-Red]

Thanks! That worked. Now all my code works perfectly.

[bubach]

Maybe you should have a table with all commands and another table with the function address to each command?
That way you can have a single loop to compare the commandline against the command-table (and keep count), then call the address in the other table.
Do you understand what i mean?

[IRBMe]

instead of:

Code: Select all

JE COMMAND_CLS

Make COMMAND_CLS a function, and call it instead. I also agree that you should do what bubach suggested.

[Candy]

If I've understood you correctly, then placing the word NEAR after your jump instructions should do the trick. For example:

Code: Select all

je NEAR jump_destination

Converting them to inverse logic and using a normal jump would also work:

Code: Select all

jne next_try_1
jmp jump_destination

next_try_1:

[Brendan]

Hi,

If I've understood you correctly, then placing the word NEAR after your jump instructions should do the trick. For example:

Code: Select all

je NEAR jump_destination

Turning on NASM's multi-pass optimizer also works (or at least it does for versions released in the last few years). Check the NASM manual ("Section 2.1.16 The -On Option: Specifying Multipass Optimization").

I generally use something like this in my makefiles:
NASM = nasm -i$(INCDIR) -O99 -w+orphan-labels -f bin

It's easier than manually adding the NEAR keyword each time NASM gives an error. The optimizer doesn't change your instructions, it only finds better encodings for the instructions - the final binary will still match a disassembly line for line (useful for Bochs debugging) .


Cheers,

Brendan