llvm dev - Sep 2018 - Serializing LLVM IR in custom fashion

Hello.

I need to write a pass that converts a whole LLVM module into a bytecode
for some other VM. VM's instructions resemble LLVM ones pretty much, so it
is almost 1-to-1 translation. Basically, I need to do what BitWriterPass
does, but using different format for the resulting bytecode.

The obvious solution is just to iterate over all Instructions and serialize
them using giant switch(i->getOpcode()), however this involves writing much
boilerplate code. And I would eventually need to write a deserializer too,
so that doubles the boilerplate work.

Any ideas how to do that easier way? How much of existing LLVM code can I
reuse?

Thanks in advance.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180906/6dfc7c7b/attachment.html>

Hi,

[InstVisitor](http://llvm.org/doxygen/classllvm_1_1InstVisitor.html) will
give you that switch. Derive from that and you get to handle the different
instruction types appropriately.

Cheers,
Philip

On Thu, Sep 6, 2018 at 3:41 PM Gleb Popov via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Hello.
>
> I need to write a pass that converts a whole LLVM module into a bytecode
> for some other VM. VM's instructions resemble LLVM ones pretty much, so
it
> is almost 1-to-1 translation. Basically, I need to do what BitWriterPass
> does, but using different format for the resulting bytecode.
>
> The obvious solution is just to iterate over all Instructions and
> serialize them using giant switch(i->getOpcode()), however this involves
> writing much boilerplate code. And I would eventually need to write a
> deserializer too, so that doubles the boilerplate work.
>
> Any ideas how to do that easier way? How much of existing LLVM code can I
> reuse?
>
> Thanks in advance.
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180906/9875b33d/attachment.html>

To really reduce boilerplate you might be able to define the right macros
and then #include include/llvm/IR/Instruction.def.

On 6 September 2018 at 23:16, Philip Pfaffe via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Hi,
>
> [InstVisitor](http://llvm.org/doxygen/classllvm_1_1InstVisitor.html) will
> give you that switch. Derive from that and you get to handle the different
> instruction types appropriately.
>
> Cheers,
> Philip
>
> On Thu, Sep 6, 2018 at 3:41 PM Gleb Popov via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
>> Hello.
>>
>> I need to write a pass that converts a whole LLVM module into a
bytecode
>> for some other VM. VM's instructions resemble LLVM ones pretty
much, so it
>> is almost 1-to-1 translation. Basically, I need to do what
BitWriterPass
>> does, but using different format for the resulting bytecode.
>>
>> The obvious solution is just to iterate over all Instructions and
>> serialize them using giant switch(i->getOpcode()), however this
involves
>> writing much boilerplate code. And I would eventually need to write a
>> deserializer too, so that doubles the boilerplate work.
>>
>> Any ideas how to do that easier way? How much of existing LLVM code can
I
>> reuse?
>>
>> Thanks in advance.
>> _______________________________________________
>> LLVM Developers mailing list
>> llvm-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180907/d302a8db/attachment.html>