- User Since
- Jun 28 2017, 5:32 PM (50 w, 5 d)
Tue, May 22
May 9 2018
Here's a comparison with D45842 for the 4th test @bit-check-combine-256 (which I understand is a long one, I'll try to see if I can reduce this.)
May 8 2018
After -instcombine with D46336
D46595 is a simpler but limited version of this. Note the test @bit-check-combine-256() doesn't get folded there as simple as here. But it doesn't modify the run loop and is less complex.
May 7 2018
After -instcombine -reassociate -instcombine -reassociate -instcombine with D45842
Thanks for finding that thread...somehow I find it difficult to search llvm-dev history...
May 3 2018
Re: bloat, is the consensus that we just don't add to instcombine any more?
FWIW, a nice thing is that this patch doesn't need to recognize and combine some specific patterns (eg. bitchecks, FoldPHIArgOrIntoPHI (D44626) or and-or-lshr (D45986)) or use some reassociation/canonicalization rules/orders based on some specific patterns (eg. some binops that operate on the same value for bitchecks, a certain definition of “matching pair” like D45842 or the ranking in the reassociate pass) (and those orders might conflict among them).
May 2 2018
Used utils/update_test_checks.py for bit-check-combine.ll.
May 1 2018
Mar 13 2018
Mar 12 2018
Mar 9 2018
Mar 6 2018
Mar 1 2018
Fixed bugs in the previous revision.
Ok, ready for reviews.
Feb 28 2018
I found a bug in the first revision. Please hold off reviews. Will update.
Feb 27 2018
Dec 12 2017
I think it will be an assert failure like before. Like said above, this patch isn't intended to handle that case.
Dec 11 2017
Going back to Diff 125662.
Dec 7 2017
Detect indirectbr critical non-MST edges in CFGMST and if none, avoid iterating
over the edges in FuncPGOInstrumentation().
Dec 6 2017
Split only the IndirectBr critical edges that need to be split rather than
all. This avoids changing the indirectbr.ll test.
Dec 5 2017
Split indirectbr critical edges in the PGO gen/use passes.
Dec 4 2017
Rebased after D40750.
Created https://reviews.llvm.org/D40750 for the refactoring part.
Dec 1 2017
Nov 30 2017
Nov 20 2017
Nov 14 2017
Rebased and addressed comments.
Nov 13 2017
Nov 2 2017
Oct 27 2017
Oct 25 2017
Changed the metadata name.
Added the metadata to the doc.
Oct 20 2017
Oct 17 2017
Oct 6 2017
To give more context, the motivating issue for us is the block frequency propagation of irreducible loops under PGO.
Oct 2 2017
A bit of cleanup and added an assert.
Sep 13 2017
The last note is that, as I suspect, we don't include a header in CodeGen/ into a file under Analysis/ which prevents the overload idea from working.
Sep 12 2017
Addressed the comment.