NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/os/linux/22182

Internet Message Format | 1992-12-30 | 3.6 KB

Xref: sparky comp.os.linux:22182 comp.windows.x.i386unix:184
Newsgroups: comp.os.linux,comp.windows.x.i386unix
Path: sparky!uunet!gatech!paladin.american.edu!darwin.sura.net!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!cbnewsc!cbfsb!att-out!cbnewsj!dwex
From: [e-mail] (david.e.wexelblat)
Subject: Re: XFree 1.1 with T8900; mixed results.
Organization: AT&T
Date: Wed, 30 Dec 1992 19:50:53 GMT
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]> <[e-mail]>
Lines: 62

In article <[e-mail]> [e-mail] (David Dawes) writes:
> In article <[e-mail]> [e-mail] (Thomas Roell) writes:
> >>  2. why does the 256 color Trident display driver only support
> >>     T8900C chips? What's this annoying difference between the
> >>     T8900B and C?
> >
> >That's a very trivial reason. For some X stuff you need to have access
> >to a source byte and a destination byte at the same time. If you deal
> >with SVGA's which cannot map their whole 1MB video memory into the
> >CPU's address space, you have to do banking. If you also have to
> >fullfill the upper constraint, then you see that you need to have two
> >banks which can be used independantly (since source and destination of
> >the operation might not be in the same bank).
> >
> >The TVGA8900B supports only ONE bank, while the TVGA8900C supports TWO
> >banks.
>
> The XFree86 tvga driver actually uses the TVGA8900C in single-bank mode.
> Apparently it benchmarked quicker in that mode than in the dual-bank
> mode (don't ask me why -- I was suprised at the results).  We use a
> different bitblt routine in this case.
>

David is right, and Thomas is wrong (rare though that is :->).  XFree86
uses only a single bank for the TVGA8900.  For whatever reason, one bank
is faster than two (verified with three different boards on two different
machines, under 3 different OSs).

> For this reason I expected that our driver might work on the 8900B, but
> it seems that it doesn't.  The only other difference that I know of
> between the 8900B and 8900C is that some registers are read-only on the B
> and R/W on the C.  If someone with an 8900B and X source on-line wants
> to look into this, please let us know.

As one of the two developers of XFree86's Trident driver, I can explain
this.  There are two registers on the TVGA8900 that indicate the amount
of installed memory.  One is always right, and the other is generally
wrong.  The value in this second register must be set to the correct memory
amount or the memory addressing will be wrong.  On the TVGA8900B, this
register cannot be written, hence the correction cannot be made.  It
MIGHT be possible to use a TVGA89000B with only 512k memory, but I doubt
it.  Rather than have things fail in bizarre ways, we just disallowed
the use of the 8900B at startup.

Sorry.


>
> David
> --
> ------------------------------------------------------------------------------
>  David Dawes ([e-mail])    DoD#210      | Phone: [phone]
>  School of Physics, University of Sydney, Australia   | Fax:   [phone]
> ------------------------------------------------------------------------------


--
David Wexelblat <[e-mail]>  [phone]
AT&T Bell Laboratories, 200 Laurel Ave - 3F-428, Middletown, NJ  07748

"The meaning of life?  That's simple.  Try to be happy, try not to hurt
 other people, and hope to fall in love."  -- Mallory Keaton