Date:	Wed, 11 Nov 1998 12:20:06 +0200 (IST)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender:	eliz AT is
To:	Jonathan Villani <jonathan DOT villani AT videotron DOT ca>
cc:	djgpp AT delorie DOT com
Subject:	Re: Question regarding the use of __djgpp_nearptr_enable()
In-Reply-To:	<so222.447$4S.2902@weber.videotron.net>
Message-ID:	<Pine.SUN.3.91.981111121946.4151L-100000@is>
MIME-Version:	1.0
Reply-To:	djgpp AT delorie DOT com

On Tue, 10 Nov 1998, Jonathan Villani wrote:

> I have seen tutorials on the web for programming the DMA but all them use
> the __djgpp_nearptr_enable function.

I would guess that this is due to three main reasons:

  1) Near pointers make the coding somewhat easier, especially if the
     programmer is used to access absolute addresses with normal C
     pointers and data structures.

  2) Near pointers make it easier to convert real-mode programs to
     DJGPP without understanding too much how does protected mode
     work.

  3) People tend to think (incorrectly, IMHO) that near pointers make
     your program a whole lot faster.

> I don't want to use that because it disable the memory protection.

You are right.

> Here's some code from the tutorials:
> --------------------------------------------------
> 
> void *MK_FP(DWORD seg, DWORD ofs)
> {
>  if (!(_crt0_startup_flags & _CRT0_FLAG_NEARPTR))
>   if
> (!__djgpp_nearptr_enable()) ------------------------------------------------
> > Oh no ! no memory protection!
>    return (void*)0;
>   return (void *) (seg*16+ofs+__djgpp_conventional_base);
> }

This was taken directly from the FAQ, where this code is supposed to
help those who have a lot of real-mode code using the MK_FP macro.  I
never meant for this to proliferate, I just wanted to help those who
need a one-time quick-and-dirty hack to make some legacy program work
with minimal effort.

Note that the __djgpp_nearptr_enable/disable functions call a DPMI
service, which will tremendously slow down your program if you call
MK_FP in some inner loop or in a time-critical section of your code.

>  memset((BYTE *)MK_FP(SegInfo.rm_segment,0),0,BUFFSIZE); ----------> Set
> buffer to 0 using nearpointer!

`dosmemput' would make the same job with the same speed, but without
compromising memory protection.

> (A)  __djgpp_nearptr_enable();     ------------------> Does this line will
> slow down my real time app.??

It does, a lot!

> /* COPY using DOS memory using Nearpointers */
>  memset(...[DMAbuffer -> in DOS memory])
> ....
> (B)  __djgpp_nearptr_disable();   ------------------> Does this line will
> slow down my real time app.??

Yes, it will be very slow.  Just use `dosmemput' instead of near
pointers, and you will get a program that's not only safer, but also
much faster.

- Raw text -