[kaffe] Host Porting Interface

Ronald Aigner ra3 at os.inf.tu-dresden.de
Wed Jun 18 09:31:01 PDT 2003


Hi Dalibor,

On Wednesday 28 May 2003 19:21, Ronald Aigner wrote:
> > The only reference to HPI I could find was
> > http://java.sun.com/j2me/docs/pdf/CDCFoundation_Porting_Guide.pdf . Is
> > that the one you are talking about.
>
> Well, what I tried to get hands on is probably that. I just read about it
> in a Paper and thought that this would be a good start to make Java-thread
> scheduling more transparent for the OS, since it had an explicit interface
> to control thread state.
>
> I wrote the authors of the paper about a week ago, but didn't get a
> response yet... The only reference of their paper I couldn't check for HPI
> was the "Inside JVM" book they mentioned. Maybe somebody here has it and
> can check if it mentions HPI.
I finally got an answer from the authors:
<cite>
    The description of the HPI interfaced that we present in the OSR article
is based on the study of the source code of Java 2 SDK, distributed by
SUN from its web page (http://java.sun.com).

    We didn't found articles that describe with details the implementation
of the HPI. So, we had to study its implementation starting from the
source code.
</cite>

> > Edouard's porting guide is a good start:
> >
> > http://egp.free.fr/port-kaffe/port-kaffe-0.2.html
> >
> > There is the config/cpu/os directory. That's where definitions of system
> > call macros, thread stack size, stack pointer offset, jit and interpreter
> > instruction set definitions, etc. go. If your platform doesn't offer a
> > unix'ish API, you'll have to implement your own threading system under
> > kaffe/kaffevm/system/. You'll also have to define threadsafe functions
> > for performing unix'ish operations defined in include/jsyscalls.h in
> > kaffe/kaffevm/system/your-system/syscall.c. If you want to use the AWT
> > and your platform does not have a supported AWT subsytem, you'll need to
> > port the AWT as well ;)
> >
> > If you go ahead, port kaffe to a new platform, and decide to write some
> > docs that complement those by Edouard, ... well, you know where to send
> > them ;)
>
> That's a good start. I (or rather Martin) did that ;)
That's what we'll use then.

Greetings, Ron.
-- 
Mit freundlichen Gruessen/With regards
ra3 @ os.inf.tu-dresden.de
http://os.inf.tu-dresden.de/~ra3/





More information about the kaffe mailing list