Willow Schlanger wrote:
>
> How are you doing DT? I can't find sources for DT anywhere.. even though
> I downloaded Plex86, I'm new to CVS, and I can't get DocBook or whatever
> to work......
There is a docs tarball on the FTP site, for those who don't want
to use DocBook to generate the files. HTML and PS are pre-formatted
for you.
Also, look in dt-testbed/proto2/README.
What I'm working on currently is a very lightweight DT. It's
probably more aptly called "code filtering", a terminology from
Ramon. Mostly, the goal is to expand out some instructions
that can be handled in ring3 space so we don't often invoke the
expensive exception handling by the monitor.
The focus isn't like a normal DT strategy where you have a lot
of data flow analysis, register allocation, state management, etc.
> I can help with the theory. I'd
> like to give you a program I wrote: it takes in a script file that is
> about 140 lines or so, and it expands from that all possible binary
> encodings for all Pentium instructions (minux MMX and FPU and anything
> later than the Pentium) along with the coresponding instruction names,
> and assembly language operands. It is extremely easy to use (as I have
> done) the resulting binary tables for disassembling, and with work,
> assembling. If we could devise up a pseudo-code language that models
> exactly what each systems-level x86 instruction does, in terms of
> certain 'axiom instructions' that we will make up, then, theoretically,
> we could write a program that uses that data to, automatically, generate
> C source code for a program that serves as a DT. Wouldn't that be nifty?
We already have an instruction decoder and disassembler.
-Kevin
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Kevin Lawton [EMAIL PROTECTED]
MandrakeSoft, Inc. Plex86 developer
http://www.linux-mandrake.com/ http://www.plex86.org/
