[llvm-dev] Debug information in NVPTX backend

[Artem Belevich via llvm-dev]

It's on our todo list, though not at the top of it yet.

I've tried enabling dwarf generation some time back and quickly ran into
the fact that debug constructs get emitted at locations where PTX syntax
does not allow it. At the very least dwarf generation would need to be
tweaked to deal with PTX's syntax peculiarities. That's about as far as I
went.

So if you're ready to work on it, it would indeed be very useful.

--Artem


On Mon, Jan 11, 2016 at 2:34 PM, Jingyue Wu <jingyue at google.com> wrote:

> +Artem
>
> Agreed -- it's for sure nice to have. I haven't worked anything on this
> direction. I have the impression that NVPTX can already generate ".loc"
> directives for debugging, but not dwarf.
>
> On Mon, Jan 11, 2016 at 8:26 AM, Justin Holewinski <jholewinski at nvidia.com
> > wrote:
>
>> Hi Arpith!
>>
>> Full DWARF emission in the NVPTX backend would be a very welcome
>> addition!  We (NVIDIA) have this internally, but are unable to upstream it
>> at this time.  I’m not aware of anyone currently working on it, though
>> Google has been working on many other aspects of the backend.  Adding
>> Jingyue as a point of contact there.
>>
>> On Jan 11, 2016, at 11:03 AM, Arpith C Jacob <acjacob at us.ibm.com> wrote:
>>
>> Hi Justin,
>>
>> I understand you are the maintainer of the NVPTX backend so I have a
>> question for you.
>>
>> I would like to add DWARF info to the generated PTX. Do you know if
>> people are already working on this? Has this been tried before? I would
>> appreciate your thoughts and suggestions before I attempt this.
>>
>> Thanks for your time in advance.
>>
>> Regards,
>> Arpith
>>
>>
>> ------------------------------
>> This email message is for the sole use of the intended recipient(s) and
>> may contain confidential information.  Any unauthorized review, use,
>> disclosure or distribution is prohibited.  If you are not the intended
>> recipient, please contact the sender by reply email and destroy all copies
>> of the original message.
>> ------------------------------
>>
>
>


-- 
--Artem Belevich
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160111/96813c68/attachment.html>