DomainKey-Signature:a=rsa-sha1; q=dns; c=nofws;
  s=s1024; d=yahoo.com.ar;
  h=Message-ID:X-YMail-OSG:Received:X-Mailer:Date:From:Subject:To:In-Reply-To:MIME-Version:Content-Type;
  b=258btLwt7Nk5EtHFom8sVJESjSWk+gJ5/Sj+2/Fj5xtsDcufsPX0eaiA7SCtQfkfmpxvcpfNkqk9/Pv0C2JTKye5oPpi0pUq1y26dcR4tjBpazfRQyK+t4fxE/FoOoHS1/TI/YxaK4uhnLqX8EEmdRBY0vlerOhz+++YU6kpxTI=;
Message-ID: <880318.53118.qm@web45107.mail.sp1.yahoo.com>
Date: Sun, 30 Jan 2011 12:18:11 -0800 (PST)
From: Pablo Marty <tigrepotrazosalvaje AT yahoo DOT com DOT ar>
Subject: Re: question of assembler
To: djgpp AT delorie DOT com
In-Reply-To: <if0r8s$ahk$1@speranza.aioe.org>
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary="0-328552469-1296418691=:53118"
Reply-To: djgpp AT delorie DOT com
Errors-To: nobody AT delorie DOT com
Precedence: bulk

--0-328552469-1296418691=:53118
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable

Mr Rod Pemberton
thank you very much, the solution you gave me succeded
=A0
What I found out is that what was wrong was not the linear framebuffer I al=
so succesfully compiled (but my game still ran slow under DOSBox) but DOSBo=
x itself
Why?
If the problem was the video system, something like an archive load into me=
mory would be normally fast; under Windows XP (whose problem is that does n=
ot support sound) my program loads very fast from the disk the files that c=
ontains the graphics, also the game runs fast. Under DOSBox both are slow. =
So the problem is not either the protected mode bank switching nor the line=
ar framebuffer. It's something about DOSBox speed. Yes,=A0I tried=A0setting=
 the CPU Cicles higher but it was the same.
=A0
Does someone have an idea about how to solve this??? If necessary, I can se=
nd my code, EXE, and all files needed
=A0
thank you
=A0
Pablo

Asesinar y comer un perro o un gato es lo mismo q hacerlo a una vaca, un po=
llo, un cerdo o un pez. El genocidio no discrimina. No seas complice de gen=
ocidas. No seas genocida. En la vida tarde o temprano todo vuelve. No comas=
 cadaveres. No seas necr=F3fago.=20

--- El vie 24-dic-10, Rod Pemberton <do_not_have AT notreplytome DOT cmm> escribi=
=F3:


De: Rod Pemberton <do_not_have AT notreplytome DOT cmm>
Asunto: Re: question of assembler
Para: djgpp AT delorie DOT com
Fecha: viernes, 24 de diciembre de 2010, 1:01


"Pablo Marty" <tigrepotrazosalvaje AT yahoo DOT com DOT ar> wrote in message
news:846345 DOT 79275 DOT qm AT web45108 DOT mail DOT sp1 DOT yahoo DOT com...
> hi guys and girls
> I'm still a the search of the VESA VBE protected mode bank switching
function in order to run my SuperMarioBros videogame normally fast and with
music under DOSBox
>
> this is the code I got from the DJGPP page about VESA 2.0
> the compiler throws an error at the set_vesa_bank_pm() function (at the
end of the code)
> In the last time I wrote to the list, Mr DJ Delorie told me that "you
can't clobber and use them", I suppose refering to the CPU registers. But h=
e
didn't tell me how to clobber them without using them ....
> I suppose that's what has to be done, because beside the code there is a
coment that says "clobber list ..."

There are three colons that can follow the inline assembly.=A0 The first is
for the input.=A0 The second is for the output.=A0 GAS understands that
registers used for the inputs and outputs get "clobbered", i.e., destroyed,
changed, modified...=A0 The third is for the "clobber" list.=A0 That's for
registers which get modified that are in the assembly which GAS was not tol=
d
about in the input and output sections.=A0 E.g., if you used %ecx in the
assembly, but it was not an input or output register, it'd go into the
clobber list.

> Can someone tell me how to clobber the list of registers correctly?
...

>
>This code will give you a pointer to the protected mode bank switching
function, but you cannot call this directly from C because it uses a specia=
l
register based argument passing convention. A little bit of inline >asm is
needed to make sure the parameters go into the correct registers, eg:=A0=A0=
=A0void
set_vesa_bank_pm(int bank_number)
> {
>=A0 =A0 =A0 asm (
>" call *%0 "

This assembly doesn't modify any registers.=A0 If it did, it would need to =
go
into the clobber list, if it isn't being used as an input or output.

>=A0 =A0 =A0 :=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=A0=A0=
/* no outputs */

There are no outputs.

>=A0 =A0 =A0 : "r" (pm_bank_switcher),=A0 =A0=A0=A0/* function pointer in a=
ny register */

There are inputs.=A0 This will clobber a register, but GCC understands whic=
h
register it chose.

> "b" (0),=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=A0=A0/* set %ebx to zero */

ebx is used.=A0 It shouldn't be in the clobber list below.

> "d" (bank_number)=A0 =A0 =A0 =A0 =A0 /* bank number in %edx */

edx is usd.=A0 It shouldn't be in the clobber list below.

>=A0 =A0 =A0 : "%eax",=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=A0=A0/* clobb=
er list (we have no way of */
> "%ebx",=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 /* knowing which registers=
 the VESA */

I think you need to remove the ebx line.

> "%ecx",=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 /* code is going to change=
, so we */
> "%edx",=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 /* have to assume the wors=
t and list */

I think you need to remove the edx line.

> "%esi",=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 /* them all here) */
> "%edi"
>=A0 =A0 =A0=A0=A0);
>=A0 =A0 }
>

You may also need to remove one more register from the list, perhaps eax, s=
o
that GCC can pick a register for the "r" input.=A0 Try something like that.


Rod Pemberton


=0A=0A=0A      
--0-328552469-1296418691=:53118
Content-Type: text/html; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable

<table cellspacing=3D"0" cellpadding=3D"0" border=3D"0" ><tr><td valign=3D"=
top" style=3D"font: inherit;"><DIV>Mr Rod Pemberton</DIV>
<DIV>thank you very much, the solution you gave me succeded</DIV>
<DIV>&nbsp;</DIV>
<DIV>What I found out is that what was wrong was not the linear framebuffer=
 I also succesfully compiled (but my game still ran slow under DOSBox) but =
DOSBox itself</DIV>
<DIV>Why?</DIV>
<DIV>If the problem was the video system, something like an archive load in=
to memory would be normally fast; under Windows XP (whose problem is that d=
oes not support sound) my program loads very fast from the disk the files t=
hat contains the graphics, also the game runs fast. Under DOSBox both are s=
low. So the problem is not either the protected mode bank switching nor the=
 linear framebuffer. It's something about DOSBox speed. Yes,&nbsp;I tried&n=
bsp;setting the CPU Cicles higher but it was the same.</DIV>
<DIV>&nbsp;</DIV>
<DIV>Does someone have an idea about how to solve this??? If necessary, I c=
an send my code, EXE, and all files needed</DIV>
<DIV>&nbsp;</DIV>
<DIV>thank you</DIV>
<DIV>&nbsp;</DIV>
<DIV>Pablo<BR><BR><FONT style=3D"BACKGROUND-COLOR: #8000ff" face=3D"arial, =
helvetica, sans-serif"><FONT color=3D#ffff00>Asesinar y comer un perro o un=
 gato es lo mismo q hacerlo a una vaca, un pollo, un cerdo o un pez. El gen=
ocidio no discrimina. No seas complice de genocidas. No seas genocida. En l=
a vida tarde o temprano todo vuelve. No comas cadaveres. No seas necr=F3fag=
o.</FONT> </FONT><BR><BR>--- El <B>vie 24-dic-10, Rod Pemberton <I>&lt;do_n=
ot_have AT notreplytome DOT cmm&gt;</I></B> escribi=F3:<BR></DIV>
<BLOCKQUOTE style=3D"PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: rgb(=
16,16,255) 2px solid"><BR>De: Rod Pemberton &lt;do_not_have AT notreplytome DOT cm=
m&gt;<BR>Asunto: Re: question of assembler<BR>Para: djgpp AT delorie DOT com<BR>Fe=
cha: viernes, 24 de diciembre de 2010, 1:01<BR><BR>
<DIV class=3DplainMail>"Pablo Marty" &lt;<A href=3D"http://ar.mc451.mail.ya=
hoo.com/mc/compose?to=3Dtigrepotrazosalvaje AT yahoo DOT com DOT ar" ymailto=3D"mailto=
:tigrepotrazosalvaje AT yahoo DOT com DOT ar">tigrepotrazosalvaje AT yahoo DOT com DOT ar</A>&gt;=
 wrote in message<BR>news:<A href=3D"http://ar.mc451.mail.yahoo.com/mc/comp=
ose?to=3D846345 DOT 79275 DOT qm AT web45108 DOT mail DOT sp1 DOT yahoo DOT com" ymailto=3D"mailto:846=
345 DOT 79275 DOT qm AT web45108 DOT mail DOT sp1 DOT yahoo DOT com">846345 DOT 79275 DOT qm AT web45108 DOT mail DOT sp1=
.yahoo.com</A>...<BR>&gt; hi guys and girls<BR>&gt; I'm still a the search =
of the VESA VBE protected mode bank switching<BR>function in order to run m=
y SuperMarioBros videogame normally fast and with<BR>music under DOSBox<BR>=
&gt;<BR>&gt; this is the code I got from the DJGPP page about VESA 2.0<BR>&=
gt; the compiler throws an error at the set_vesa_bank_pm() function (at the=
<BR>end of the code)<BR>&gt; In the last time I wrote to the list, Mr DJ De=
lorie told me that "you<BR>can't clobber and use them", I suppose refering =
to the
 CPU registers. But he<BR>didn't tell me how to clobber them without using =
them ....<BR>&gt; I suppose that's what has to be done, because beside the =
code there is a<BR>coment that says "clobber list ..."<BR><BR>There are thr=
ee colons that can follow the inline assembly.&nbsp; The first is<BR>for th=
e input.&nbsp; The second is for the output.&nbsp; GAS understands that<BR>=
registers used for the inputs and outputs get "clobbered", i.e., destroyed,=
<BR>changed, modified...&nbsp; The third is for the "clobber" list.&nbsp; T=
hat's for<BR>registers which get modified that are in the assembly which GA=
S was not told<BR>about in the input and output sections.&nbsp; E.g., if yo=
u used %ecx in the<BR>assembly, but it was not an input or output register,=
 it'd go into the<BR>clobber list.<BR><BR>&gt; Can someone tell me how to c=
lobber the list of registers correctly?<BR>...<BR><BR>&gt;<BR>&gt;This code=
 will give you a pointer to the protected mode bank
 switching<BR>function, but you cannot call this directly from C because it=
 uses a special<BR>register based argument passing convention. A little bit=
 of inline &gt;asm is<BR>needed to make sure the parameters go into the cor=
rect registers, eg:&nbsp;&nbsp;&nbsp;void<BR>set_vesa_bank_pm(int bank_numb=
er)<BR>&gt; {<BR>&gt;&nbsp; &nbsp; &nbsp; asm (<BR>&gt;" call *%0 "<BR><BR>=
This assembly doesn't modify any registers.&nbsp; If it did, it would need =
to go<BR>into the clobber list, if it isn't being used as an input or outpu=
t.<BR><BR>&gt;&nbsp; &nbsp; &nbsp; :&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbs=
p; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp;&nbsp;/* no=
 outputs */<BR><BR>There are no outputs.<BR><BR>&gt;&nbsp; &nbsp; &nbsp; : =
"r" (pm_bank_switcher),&nbsp; &nbsp;&nbsp;&nbsp;/* function pointer in any =
register */<BR><BR>There are inputs.&nbsp; This will clobber a register, bu=
t GCC understands which<BR>register it chose.<BR><BR>&gt; "b"
 (0),&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp;&n=
bsp;/* set %ebx to zero */<BR><BR>ebx is used.&nbsp; It shouldn't be in the=
 clobber list below.<BR><BR>&gt; "d" (bank_number)&nbsp; &nbsp; &nbsp; &nbs=
p; &nbsp; /* bank number in %edx */<BR><BR>edx is usd.&nbsp; It shouldn't b=
e in the clobber list below.<BR><BR>&gt;&nbsp; &nbsp; &nbsp; : "%eax",&nbsp=
; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp;&nbsp=
;/* clobber list (we have no way of */<BR>&gt; "%ebx",&nbsp; &nbsp; &nbsp; =
&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; /* knowing which registers=
 the VESA */<BR><BR>I think you need to remove the ebx line.<BR><BR>&gt; "%=
ecx",&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; =
/* code is going to change, so we */<BR>&gt; "%edx",&nbsp; &nbsp; &nbsp; &n=
bsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; /* have to assume the worst =
and list */<BR><BR>I think you need to remove the edx
 line.<BR><BR>&gt; "%esi",&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; =
&nbsp; &nbsp; &nbsp; /* them all here) */<BR>&gt; "%edi"<BR>&gt;&nbsp; &nbs=
p; &nbsp;&nbsp;&nbsp;);<BR>&gt;&nbsp; &nbsp; }<BR>&gt;<BR><BR>You may also =
need to remove one more register from the list, perhaps eax, so<BR>that GCC=
 can pick a register for the "r" input.&nbsp; Try something like that.<BR><=
BR><BR>Rod Pemberton<BR><BR><BR></DIV></BLOCKQUOTE></td></tr></table><br>=
=0A=0A=0A=0A=0A      &nbsp;
--0-328552469-1296418691=:53118--