help with calling API's in assembly

[Wayne Diamond]

could somebody show me a simple demo of how to invoke/call an API using inline assembly?
like, instead of

Declare Function MessageBoxA etc ..
FUNCTION PBMAIN() AS LONG
 Call MessageBoxA "Hello"
END FUNCTION

would somebody show me how to use nothing but inline assembly to invoke the MessageBoxA API ? i can already use that API directly with PB, but id love to know how to do it with PB's inline asm!

many thanks in advance for helping an asm newbie try and get more revs out of his now-standalone apps



------------------

[Edwin Knoppert]

I believe it's INVOKE..

What you can do is find the link to lcc compiler, from there you'll asm sites with this info.

------------------
[email protected]

[Steve Hutchesson]

Wayne,

The problem is that the calling of an API in assembler does not do it any
better, smaller or faster. In MASM a MessageBox call is as follows,

  invoke MessageBox,hWnd,ADDR MsgText,ADDR TitleTxt,MB_OK
  
  manually it looks like this,
  
      push MB_OK
      push offset TitleText
      push offset MsgText
      push hWnd
      call MessageBox
  
  Do the same call in PowerBASIC,
  
  MessageBox hWnd,"Hi Folks","Greetings",%MB_OK

and it is the same assembler so you do not pick up any advantage by doing so.

You pick up the simple advantages by using bits of the non critical code
like "! inc var", "! add var, 5", "! sub var 2". They are small, fast and
less typing.

It is supposed to politically and aerodynamically incorrect to use GOTO in
modern structured code so use the real thing "! jmp label" and you can
mess up "nice" well behaved code to make it go faster.

Loops are good value in assembler as well, get the swing of a few basics
and the overhead is minimal and the speed is very good. there are a
couple of tricks with loops that help as they get the instruction count
down which usually means higher speed.

If you have to increment an ADDRESS in a loop, calculate the number of
iterations by adding the count to the address and comparing to the result
to exit. If it is a fixed count loop, put the count in ECX and decrement
down to zero. at the end of the loop use JZ or JNZ and you save the
comparison to test for exit.

Something to keep in mind is what registers are to be preserved. If you
are using a mixed ASM/API/BASIC function or sub, you can use EAX, ECX and
EDX freely but remember that other called functions can do the same so
that if you are storing a value in one of these registers, push them onto
the stack before the function call and pop them back off after the
function call.

Remember to use "#REGISTER NONE" with inline asm code in the function/sub
as inline asm conflicts with the automatic compiler optimisation that
uses register variables and you can end up with some strange unpredictable
results if you forget to use it.

Regards & keep up the good work.

[email protected]

[This message has been edited by Steve Hutchesson (edited October 14, 2000).]

[Wayne Diamond]

Steve, Edwin, excellent - thanks guys! Makes a lot more sense now
asm is so far out of my league its not funny, but just being able to get a grip on the basics and actually use PB's inline assembly is fantastic
Mr Zale has added a huge new dimension to my programming, and I never thought i'd be dabbling in assembly _ever_ - until I got PB. the party starts here !




------------------