Message-ID: <3f34b5c7$0$164$cc7c7865@news.luth.se>
From: Martin Str|mberg <ams AT speedy DOT ludd DOT luth DOT se>
Subject: Re: Inline assembly error
Newsgroups: comp.os.msdos.djgpp
References: <84e4e2a9 DOT 0308072140 DOT 47872db2 AT posting DOT google DOT com> <84e4e2a9 DOT 0308082024 DOT 59e5816b AT posting DOT google DOT com>
User-Agent: tin/1.4.6-20020816 ("Aerials") (UNIX) (NetBSD/1.6Q (alpha))
Date: 09 Aug 2003 08:50:15 GMT
Lines: 83
NNTP-Posting-Host: speedy.ludd.luth.se
X-Trace: 1060419015 news.luth.se 164 130.240.16.13
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Joel <jbs30000 AT aol DOT com> wrote:
: jbs30000 AT aol DOT com (Joel) wrote in message news:<84e4e2a9 DOT 0308072140 DOT 47872db2 AT posting DOT google DOT com>...
:> One of the SVGA graphics routines I'm making has given me an inline
:> error message.
:> What I'm doing is making a routine to call the protected mode verson
:> of
:> Function 9 -Set/Get Palette Data
:> BL = 0 - Set Palette Data / 80H Set Palette Data During Vertical
:> Retrace
:> CX = Number of Palette Registers to Update
:> DX = First of the Palette Registers to Update
:> ES:EDI = Table of Palette Values
:> So, here's my routine
:> unsigned short PM_Palette_Data(unsigned char Flag, unsigned short
:> PCount, unsigned short FirstPal, unsigned short TableSeg, unsigned
:> long TableOff)
:> {
:>         asm("movw %%ax, %%es\n\t"
:>             "call *%0"
:>             :
:>             : "r" (PMPaletteData), "b" (Flag), "C" (PCount), "a"
:> (TableSeg), "D" (TableOff)
:>         );
:> }
:> 
:> And here's the error I get
:> Error: inconsistent operand constraints in an `asm'
:> Referring of course, to the "movw %%ax, %%es\n\t"
:> I'm not sure why this is happening.  movw is move word ax and es are
:> words, although maybe it's confusing es as esi, but then what would I
:> use to represent ES?
:> Anyway, I'm still getting use to AT&T style DJGPP extended inline
:> assembly, so any help is appreciated.  Thanks.

: All right, nobody is answering me, so it must be that the answer is
: obvious, and you're waiting for me to figure it out on my own.  But

No, it's hard. Or too obvious for me.

: going over everything, I shouldn't be having a problem.  movw moves
: words, and ax and es are words.  I used two %% with each register
: which is required in extended inline assembly.  The source and dest
: are in the correct order, movw %%ax, %%es should move ax into ex, and
: I put the required \n\t at the end since I had two lines of assembly. 

But you're wrong; it's not movw that makes gcc complain; there are two
bugs too (``"C"''; DX?). (And shouldn't "TableSeg" be "TableSel"?)


I tried to figure it out. And it's the second constraint that messes
up. If I remove that one (Flag), the next one (PCount) makes it
complain. This doesn't make sense.

Things I did: 

1. Change every non int/long to int.

2. ``"C" (PCount)'' -> ``"c" (PCount)''.

3. Removal of "movw %%ax, %%es\n\t".

4. "movw" -> "movl" -> "mov".

5. "movw %%ax, %%es\n\t" -> "movl %%eax, %%es\n\t". Repeat 4.

6. Removal of "c" (PCount).

7. Removal of every constraint except "r" (PMPaletteData).

With 1 and 7 it the warning vanishes at least. Not sure if 1 is
necessary.

Most of the tries above is just to understand what is going on. Some
of the 4 and 5 changes are broken anyway.


Sum: there's someting strange going on.


Right,

						MartinS