[llvm-dev] RFC: Inlining report

Easwaran Raman via llvm-dev llvm-dev at lists.llvm.org
Wed Oct 28 15:39:30 PDT 2015

This is certainly very useful! My comments are related to determining the
inline reason.

*   -> INLINE: stat (35<=487)*
> *      <<Callee is single basic block>>*
In the above example, the callee would have been inlined even if the single
BB bonus is not applied to the threshold (since 35 <= 225). The fact that
it is single BB didn't tip the scales. If the stat were, say, (300 <= 487),
then the fact that it is a single BB is crucial. Perhaps, there could be a
way to specify a primary reason and many secondary reasons?

*   -> INLINE: setExit (15<=225)*
> *      <<Inlining is profitable>>*

I think it is more useful to know why it is profitable. At the minimum it
will be useful to differentiate between a callee that is so small that it
will get inlined at any callsite vs a callee that is profitable to inline
at this callsite due to instruction simplification.

> Thanks,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20151028/d5944cd4/attachment.html>

More information about the llvm-dev mailing list