llvm dev - Apr 2017 - Reducing printing deducible information in MIR based codegen testing

Hi Matthias,

Sorry I took some time.
I have checked code for MIRPrinter.cpp and I think we can control what will
be converted to YAML from there. So Do you mean to control information from
there? Or some thing smarter like using mapping traits information (i.e
optional values can be turned off) and/or flag to indicate how much
information should be printed?

Please guide me if I am going in wrong direction.

Sincerely,
Vivek
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170417/8b18e640/attachment.html>

> On Apr 16, 2017, at 12:05 PM, vivek pandya <vivekvpandya at
gmail.com> wrote:
> 
> Hi Matthias,
> 
> Sorry I took some time.
> I have checked code for MIRPrinter.cpp and I think we can control what will
be converted to YAML from there. So Do you mean to control information from
there? Or some thing smarter like using mapping traits information (i.e optional
values can be turned off) and/or flag to indicate how much information should be
printed?
Whatever works (and results in nicer code). Just send patches, this also can be
split into multiple patches or starting with a proof of concept for a single
value that we can discuss on phabricator.

It seems like the yam traits allow to specify default values. Though I assume
that only works for statically known values so you may need extra code in some
cases where the “default” changes dynamically (I would expect the default
function alignment to change based on target and other things for example).
I’m not convinced “levels” make sense here (that would mean we are actually
loosing information and I can’t come up with good use cases for that).

- Matthias