• V. Krishnakumar (8/8) Jul 30 2002 Jan,
• Jan Knepper (10/18) Jul 30 2002 Hi Krish,
• V. Krishnakumar (17/25) Jul 31 2002 into

"V. Krishnakumar" <lvimala eth.net> writes:

Jan,

How do you plan to implement object-messaging ? using VTABLES ?
How about using a runtime messenger like Objective-C ? Using a runtime
messenger makes the backend much easier and simpler to write. Besides, type
introspection and message forwarding can be achieved easily. I think looking
at the GNU Objective-C sources should reveal valuable insight.

cheers,
-Krish

Jan Knepper <jan smartsoft.cc> writes:

Hi Krish,

I have not gone that far yet. Yesterday I started looking more serious into
other compiler front-end implementations to see how we coule reuse existing
technology.
Of course initially we are going to use the GCC back-end. We are just writing a
GLUE layer to glue the D front-end to the GCC back-end. Any help and insight is
more than welcome.
I certainly will take a look at the GCC Objective-C implementation.

Jan



"V. Krishnakumar" wrote:

 Jan,

 How do you plan to implement object-messaging ? using VTABLES ?
 How about using a runtime messenger like Objective-C ? Using a runtime
 messenger makes the backend much easier and simpler to write. Besides, type
 introspection and message forwarding can be achieved easily. I think looking
 at the GNU Objective-C sources should reveal valuable insight.

 cheers,
 -Krish

"V. Krishnakumar" <lvimala eth.net> writes:

"Jan Knepper" <jan smartsoft.cc> wrote in message
news:3D46E25F.3642255D smartsoft.cc...
 Hi Krish,

 I have not gone that far yet. Yesterday I started looking more serious

into
 other compiler front-end implementations to see how we coule reuse

existing
 technology.

 Of course initially we are going to use the GCC back-end. We are just

writing a
 GLUE layer to glue the D front-end to the GCC back-end. Any help and

insight is
 more than welcome.

I'm talking about the backend. The GCC backend as I assume, provides an
infrastructure for code generation. You still have to map the oop stuff,
isnt it ? (or is my ignorance speaking here ?)  AFAIK, there are 3 models of
OOP impl:
1. the "SELF" approach
2. the VTABLE approach
3. the runtime messenger.

Option (3) seems to be the most flexible, but slightly slower.

 I certainly will take a look at the GCC Objective-C implementation.

And also look at some of the old archives of the objective-c mailing list.
It contains a lot of valuable info.

-Krish