This is an archive of the discontinued LLVM Phabricator instance.

[llvm-reduce] Attempt to inline calls
Changes PlannedPublic

Authored by aeubanks on Oct 16 2022, 10:01 PM.

Details

Reviewers
arsenm

Diff Detail

Event Timeline

aeubanks created this revision.Oct 16 2022, 10:01 PM
Herald added a project: Restricted Project. · View Herald TranscriptOct 16 2022, 10:01 PM
aeubanks requested review of this revision.Oct 16 2022, 10:01 PM
Herald added a project: Restricted Project. · View Herald TranscriptOct 16 2022, 10:01 PM
arsenm added a subscriber: arsenm.Oct 16 2022, 10:03 PM
arsenm added inline comments.
llvm/tools/llvm-reduce/deltas/InlineCalls.cpp
25

Can you get away without building this vector if you use make_early_inc on the block and instruction ranges?

aeubanks updated this revision to Diff 468261.Oct 17 2022, 11:00 AM

use make_early_inc_range()
initially I only wanted to do one level of inlining everywhere, but we'd go through all of them anyway if we try running the pass again

arsenm added inline comments.Oct 17 2022, 11:13 AM
llvm/tools/llvm-reduce/deltas/InlineCalls.cpp
20

Unused variable

llvm/tools/llvm-reduce/deltas/InlineCalls.h
1

Missing C++ mode comment

arsenm accepted this revision.Oct 17 2022, 11:13 AM
This revision is now accepted and ready to land.Oct 17 2022, 11:13 AM
aeubanks planned changes to this revision.Oct 17 2022, 1:22 PM

I realized that this may badly handle mutually recursive functions, need to think about how to handle that