maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/Inline
History
Chandler Carruth bd9c29039e [PM] Finish implementing and fix a chain of bugs uncovered by testing 
the invalidation propagation logic from an SCC to a Function.

I wrote the infrastructure to test this but didn't actually use it in
the unit test where it was designed to be used. =[ My bad. Once
I actually added it to the test case I discovered that it also hadn't
been properly implemented, so I've implemented it. The logic in the FAM
proxy for an SCC pass to propagate invalidation follows the same ideas
as the FAM proxy for a Module pass, but the implementation is a bit
different to reflect the fact that it is forwarding just for an SCC.

However, implementing this correctly uncovered a surprising "bug" (it
was conservatively correct but relatively very expensive) in how we
handle invalidation when splitting one SCC into multiple SCCs. We did an
eager invalidation when in reality we should be deferring invaliadtion
for the *current* SCC to the CGSCC pass manager and just invaliating the
newly constructed SCCs. Otherwise we end up invalidating too much too
soon. This was exposed by the inliner test case that I've updated. Now,
we invalidate *just* the split off '(test1_f)' SCC when doing the CG
update, and then the inliner finishes and invalidates the '(test1_g,
test1_h)' SCC's analyses. The first few attempts at fixing this hit
still more bugs, but all of those are covered by existing tests. For
example, the inliner should also preserve the FAM proxy to avoid
unnecesasry invalidation, and this is safe because the CG update
routines it uses handle any necessary adjustments to the FAM proxy.

Finally, the unittests for the CGSCC pass manager needed a bunch of
updates where we weren't correctly preserving the FAM proxy because it
hadn't been fully implemented and failing to preserve it didn't matter.

Note that this doesn't yet fix the current crasher due to MemSSA finding
a stale dominator tree, but without this the fix to that crasher doesn't
really make any sense when testing because it relies on the proxy
behavior.

llvm-svn: 307487
 		2017-07-09 03:59:31 +00:00
..
AArch64 [AArch64] Inline callee if its target-features are a subset of the caller 2017-06-27 22:27:32 +00:00
X86 [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
2003-09-14-InlineValue.ll
…
2003-09-22-PHINodeInlineFail.ll
…
2003-09-22-PHINodesInExceptionDest.ll
…
2003-09-22-PHINodesInNormalInvokeDest.ll
…
2003-10-13-AllocaDominanceProblem.ll
…
2004-04-15-InlineDeletesCall.ll
…
2004-04-20-InlineLinkOnce.ll
…
2004-10-17-InlineFunctionWithoutReturn.ll
…
2006-01-14-CallGraphUpdate.ll
…
2006-07-12-InlinePruneCGUpdate.ll
…
2006-11-09-InlineCGUpdate-2.ll
…
2006-11-09-InlineCGUpdate.ll
…
2007-04-15-InlineEH.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
2007-06-25-WeakInline.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
2007-12-19-InlineNoUnwind.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
2008-09-02-NoInline.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
2009-01-08-NoInlineDynamicAlloca.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
2009-01-13-RecursiveInlineCrash.ll
…
2009-05-07-CallUsingSelfCrash.ll
…
2010-05-12-ValueMap.ll
…
PR4909.ll
…
align.ll
…
alloca-bonus.ll Add address space mangling to lifetime intrinsics 2017-04-10 20:18:21 +00:00
alloca-dbgdeclare-merge.ll
…
alloca-dbgdeclare.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
alloca-in-scc.ll
…
alloca-merge-align.ll
…
alloca_test.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
always-inline.ll Inliner: Don't remove calls to readnone+nounwind (but not always_inline) functions in the AlwaysInliner 2017-06-12 23:01:17 +00:00
arg-attr-propagation.ll Revert r296366 "[InlineFunction] add nonnull assumptions based on argument attributes" 2017-02-27 22:33:02 +00:00
array-alloca.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
array_merge.ll
…
attributes.ll [PM] Teach the inliner in the new PM to merge attributes after inlining. 2016-12-27 03:39:54 +00:00
basictest.ll Inliner: Don't touch indirect calls 2017-06-09 03:29:20 +00:00
bfi-update.ll Fix a bug in caller's BFI update code after inlining. 2017-02-14 22:49:28 +00:00
blockaddress.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
byval-tail-call.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
byval.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
byval_lifetime.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
callgraph-update.ll
…
casts.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
cfg_preserve_test.ll [Inliner] Modernize all of the inliner tests that were using grep. 2016-12-27 02:47:37 +00:00
cgscc-incremental-invalidate.ll [PM] Finish implementing and fix a chain of bugs uncovered by testing 2017-07-09 03:59:31 +00:00
cgscc-invalidate.ll [PM/Inliner] Make the new PM's inliner process call edges across an 2017-03-09 11:35:40 +00:00
cgscc-update.ll [PM] Teach the inliner's call graph update to handle inserting new edges 2016-12-28 03:13:12 +00:00
clear-analyses.ll [PH] Replace uses of AssertingVH from members of analysis results with 2017-01-24 12:55:57 +00:00
comdat-ipo.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
crash-lifetime-marker.ll Add address space mangling to lifetime intrinsics 2017-04-10 20:18:21 +00:00
crash.ll
…
crash2.ll
…
debug-info-duplicate-calls.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
debug-invoke.ll
…
delete-call.ll
…
deopt-bundles.ll
…
deoptimize-intrinsic-cconv.ll
…
deoptimize-intrinsic.ll
…
devirtualize-2.ll [PM] Introduce a devirtualization iteration layer for the new PM. 2016-12-28 11:07:33 +00:00
devirtualize-3.ll
…
devirtualize.ll
…
dynamic_alloca_test.ll
…
ephemeral.ll [InlineCost] Reduce inline thresholds to compensate for cost changes 2016-11-28 11:07:37 +00:00
externally_available.ll [Inliner] Modernize all of the inliner tests that were using grep. 2016-12-27 02:47:37 +00:00
frameescape.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
function-count-update-2.ll Improve PGO support for the new inliner 2017-01-20 22:44:04 +00:00
function-count-update-3.ll Improve PGO support for the new inliner 2017-01-20 22:44:04 +00:00
function-count-update.ll Improve PGO support for the new inliner 2017-01-20 22:44:04 +00:00
guard-intrinsic.ll
…
gvn-inline-iteration.ll
…
ignore-debug-info.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inalloca-not-static.ll [PM] Wire up another test to the new pass manager. 2016-12-27 06:46:16 +00:00
inline-assume.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-byval-bonus.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-cold-callee.ll Improve PGO support for the new inliner 2017-01-20 22:44:04 +00:00
inline-cold-callsite-pgo.ll [NewPM/Inliner] Reduce threshold for cold callsites in the non-PGO case 2017-06-27 23:11:18 +00:00
inline-cold-callsite.ll [NewPM/Inliner] Reduce threshold for cold callsites in the non-PGO case 2017-06-27 23:11:18 +00:00
inline-cold.ll Decrease inlinecold-threshold to 45 2017-05-11 21:36:28 +00:00
inline-constexpr-addrspacecast-argument.ll ConstantFold: Handle gep nonnull, undef as well 2017-05-08 17:37:33 +00:00
inline-fast-math-flags.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-fp.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-funclets.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-hot-callee.ll [Inliner] Do not mix callsite and callee hotness based updates. 2017-05-16 21:18:09 +00:00
inline-hot-callsite-2.ll Improve PGO support for the new inliner 2017-01-20 22:44:04 +00:00
inline-hot-callsite.ll [PM] Add ProfileSummaryAnalysis as a required pass in the new pipeline. 2017-05-04 16:58:45 +00:00
inline-indirect.ll
…
inline-invoke-tail.ll [Inliner] Fix a test where I typo'ed 'CHECK' as 'CHCEK' when converting 2017-01-04 11:15:01 +00:00
inline-invoke-with-asm-call.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-musttail-varargs.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-optnone.ll
…
inline-optsize.ll [InlineCost] Remove skew when calculating call costs 2016-11-14 11:14:41 +00:00
inline-probe-stack.ll Add a "probe-stack" attribute 2017-06-21 18:46:50 +00:00
inline-stack-probe-size.ll Define behavior of "stack-probe-size" attribute when inlining. 2017-06-22 23:22:36 +00:00
inline-tail.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline-threshold.ll
…
inline-vla.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_cleanup.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_constprop.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_dbg_declare.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_dce.ll [Inliner] Modernize all of the inliner tests that were using grep. 2016-12-27 02:47:37 +00:00
inline_invoke.ll [PM] Move the collection of call sites to a more appropriate place 2016-12-27 01:24:50 +00:00
inline_minisize.ll
…
inline_prune.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_returns_twice.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_ssp.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_stats.ll [ThinLTO] Correct counting of functions in inliner stats 2017-03-24 17:59:06 +00:00
inline_unreachable-2.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
inline_unreachable.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
internal-scc-members.ll [PM/Inliner] Fix a bug in r297374 where we would leave stale calls in 2017-03-16 10:45:42 +00:00
invoke-cleanup.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
invoke-combine-clauses.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
invoke-cost.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
invoke_test-1.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
invoke_test-2.ll [Inliner] Modernize all of the inliner tests that were using grep. 2016-12-27 02:47:37 +00:00
invoke_test-3.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
last-call-bonus.ll Do not apply redundant LastCallToStaticBonus 2017-02-14 17:30:05 +00:00
last-callsite.ll [PM] Provide an initial, minimal port of the inliner to the new pass manager. 2016-12-20 03:15:32 +00:00
lifetime-no-datalayout.ll Add address space mangling to lifetime intrinsics 2017-04-10 20:18:21 +00:00
lifetime.ll Add address space mangling to lifetime intrinsics 2017-04-10 20:18:21 +00:00
local-as-metadata-undominated-use.ll Revert "Strip debug info when inlining into a nodebug function." 2017-03-07 17:28:57 +00:00
monster_scc.ll [PM/Inliner] Add a test case that encapsulates the core issue addressed 2017-03-16 10:13:55 +00:00
nested-inline.ll [PM] Provide an initial, minimal port of the inliner to the new pass manager. 2016-12-20 03:15:32 +00:00
noalias-calls.ll
…
noalias-cs.ll
…
noalias.ll
…
noalias2.ll
…
noinline-recursive-fn.ll [PM] Add one of the features left out of the initial inliner patch: 2016-12-27 06:46:20 +00:00
noinline.ll
…
nonnull.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
null-function.ll
…
optimization-remarks-passed-yaml.ll [InlineCost] Remove skew when calculating call costs 2016-11-14 11:14:41 +00:00
optimization-remarks-with-hotness.ll [Inliner] Fold analysis remarks into missed remarks 2017-01-30 16:22:45 +00:00
optimization-remarks-yaml.ll [ORE] Add diagnostics hotness threshold 2017-06-30 23:14:53 +00:00
optimization-remarks.ll [Inliner] Fold analysis remarks into missed remarks 2017-01-30 16:22:45 +00:00
parallel-loop-md.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
partial-inline-act.ll Renable test that was disabled due to cost analysis 2017-05-14 02:58:39 +00:00
pr21206.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
pr22285.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
pr26698.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
pr28298.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
pr33637.ll [Cloner] Re-map simplfied cloned instructions. 2017-07-01 03:29:33 +00:00
prof-update-instr.ll Fix update VP metadata after inlining for instrumentation PGO 2017-05-22 20:28:18 +00:00
prof-update-sample.ll Fix update VP metadata after inlining for instrumentation PGO 2017-05-22 20:28:18 +00:00
profile-meta.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
ptr-diff.ll [InlineCost] Remove skew when calculating call costs 2016-11-14 11:14:41 +00:00
recursive.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
store-sroa.ll
…
switch.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
vector-bonus.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00
zero-cost.ll [PM] Turn on the new PM's inliner in addition to the current one for 2016-12-27 07:18:43 +00:00