[PATCH] D49385: [IPSCCP] Run Solve each time we resolved an undef in a function.

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

> Computationally the amount of work we are doing stays the same

I don't think this is right; each call to Solve() has a cost proportional to the size of the module, if I'm not mistaken.

> Maybe it would make sense to resolve undefs depending on the call graph, e.g. leaf functions first

We have scc_iterator, but I'm not sure resolving leaf functions first is actually more effective in general.



================
Comment at: test/Transforms/SCCP/ipsccp-basic.ll:253
   %call2 = call i64 @llvm.ctpop.i64(i64 %call1)
   ret void
 ; CHECK-LABEL: define void @test11b
----------------
What is this supposed to be testing?  ctpop is readnone.


https://reviews.llvm.org/D49385