On 03/10/18 16:36, Jiong Wang wrote: > On 28/09/2018 14:36, Edward Cree wrote: > > But what you've described sounds interesting; perhaps it would also > > help later with loop-variable handling? > > Haven't considered how to use this for loop-variable handling, guess you mean > applying what I have described to your previous loop detection RFC? I will > look > into your RFC later.
Tbh I was thinking more of John Fastabend's version (I'm not sure if he ever got round to posting patches, but he discussed the design towards the end of https://www.mail-archive.com/netdev@vger.kernel.org/msg216285.html ) which is building 'proper compiler data structures' and thus might be interested in proper use-def chains. (Or it might not; I'm not really a compiler-guru so it's not immediately obvious to me.) My approach was much less interested in the 'provenance' of the induction variable, just that it was increasing appropriately, so use-def chains are not really relevant to it. -Ed