Re: compiler back-end development?

"jacob navia" <jacob@nospam.org> wrote in message
> BGB / cr88192 wrote:
>>
>> but, yes, it would be better to have GCC working, but for now I have
>> ended
>> up using MSVC as my backend (static) compiler (when doing 64-bit
>> stuff...).
>>
>
> lcc-win64 has 64 bit support.
>


yep, I had thought of it, but I couldn't find much info about it, I wasn't
sure how complete its 64-bit support was, ...


similarly, I have little idea how its (external) behavior (such as
command-line options, suppported debuggers and profilers, ...) would compare
with GCC and MSVC.


I was also unsure about things like matters of liscense, ...




GCC and MSVC are "safer" bets, although in this case, GCC exhibited some
bugs that made me rather uncertain about using it, leading me to use MSVC
despite it being much less convinient.


I could try to use my own own compiler, but at this stage its Win64 support
is probably far worse than GCC (I am not to the point yet of actually
trusting my compiler's output in general, although, for what things it does,
there are not a whole lot of other options...).


I guess I am saying this, as I am aware of the tangled messes of horror that
exist in the core of my currently in-use codegen... (I had started on a
reworked codegen, but for now my old codegen needs to be used, as the new
one uses a very different API, and is targetted by a new IL compiler that is
itself not-yet-complete...).




partly have my doubts, as can be noted by last night, me going and finding
'WinDbg', and using it mostly to track down some fairly significant bugs.
after finding that some of my Win64 auto-thunks (code built and assembled at
runtime), were not complete, does not help matters much.


likewise goes for finding several bugs in my assembler, ... (one showing up
as I had in a few places been using it as an alternative to an inline
assembler, ...).


for example, the thunks for gluing my object system to the Win64 ABI were
bad, initially causing the thing to blow up whenever method-calls were
issued, ...




luckily... I caught this, mostly because I have tests rigged up to test the
core operation of my object system (that classes, instances, fields,
methods, interfaces, ... generally behave as expected...).


a lot more of my code needs proper tests, and a lot more of my code needs
more comprehensive tests, but oh well...




but, I guess bugs are to be expected when transitioning, not just the
compiler and arch used, but also over to using a good amount of more-or-less
untested code.


I have little idea as of yet what my compiler's output will be like on
Win64, as I have not yet tested this part, me still off fixing more basic
stuff...




or such...

Return to the comp.compilers page.
Search the comp.compilers archives again.