Date: Mon, 19 Jan 1998 12:26:41 -0800 (PST)
Message-Id: <199801192026.MAA04264@adit.ap.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: "elizabeth anne dominy" <msunguli AT global DOT co DOT za>, djgpp AT delorie DOT com
From: Nate Eldredge <eldredge AT ap DOT net>
Subject: Re: Strange errors after calling an ASM routine
Precedence: bulk

At 06:37  1/18/1998 GMT, elizabeth anne dominy wrote:
>Hi,
>
>I have a routine that loads data from disk into an array of chars (variable
>size) that works properly as far as I can tell. The thing is I have another
>routine that clears the graphics screen, and uses the following ASM code
>(looped if I have to clear more than one bank, depending on screen mode, NB
>not working with VBE2.0+) --
>
>__asm__ __volatile__ ("
> movw	%0,%%es
> movl	$0,%%edi
> movw	$65535, %%cx
> movb	$0,%%al
> rep
> stosb
>" : : "g" (graphics_selector) : "memory");
>
>After calling this routine, or in fact any ASM routine, simple assignment
>such as 
>
> f1 = f2; //both are structs of the same type
>
>cause Segmentation Violation Exceptions! Any ideas why?
You need to tell GCC what registers you have clobbered, and you need to
explicitly save and restore %es. Fixed version:

asm volatile("movw %%es, %%dx ;" /* I don't use `push' and `pop' because of
                                    a problem with -fomit-frame-pointer */
             "movw %0,%%es ;"
             "xorl %%edi, %%edi ;" /* faster/smaller than movl $0,... */
             "movw $65535,%%cx ;"
             "rep; stosb ;"
             "movw %%dx,%%es"
             :
             : "g" (graphics_selector)
             : "cx", "dx", "edi", "memory");

Nate Eldredge
eldredge AT ap DOT net