[Info-vax] Intel proposal to simplify x86-64

[Arne Vajhøj]

On 6/13/2023 8:32 AM, Simon Clubley wrote:
> On 2023-06-12, Arne Vajhøj <arne at vajhoej.dk> wrote:
>> On 6/12/2023 9:00 PM, John Dallman wrote:
>>> In article <u689pl$38idu$1 at dont-email.me>, arne at vajhoej.dk (Arne Vajhøj)
>>> wrote:
>>>> Does Dibol compile to native? If yes then they would need
>>>> to integrate with LLVM backend.
>>>
>>> Looks like a bytecode. So the run-time system for that would need to be
>>> re-built, along with the compiler.
>>
>> source code to byte code compiler + byte code interpreter
>>
>> should be easier to port than:
>>
>> source code to native code compiler
> 
> It's bytecode (unless it has changed recently), but the build sequence
> on Alpha (the last architecture I used Synergex Dibol on) resulted in
> normal VMS executables created using the VMS linker. It looked like the
> bytecode was embedded within the executable and the executable was linked
> against the Dibol RTL.

So they will need to:
* build the interpreter module
* build the RTL
* build a stub
* build a compiler that embed stub + interpreter module + byte code

Work.

But I don't think as much work as for a native compiler.

I am sure they wish that .NET was available on VMS.   :-)

> You can also call directly (for example) C subroutines and VMS system
> services from within your Dibol code (I used to write some low-level
> stuff in C that was too slow when implemented directly in Dibol), so
> the executables can be regarded as some hybrid of native VMS code and
> Dibol bytecode.

That is pretty standard. Java, Python etc. all has such capabilities.

Arne