Florian Hahn fc7654a67b [Local] Keep K's range if K does not move when combining metadata. ... As K has to dominate I, IIUC I's range metadata must be a subset of K's. After Eli's recent clarification to the LangRef, loading a value outside of the range is undefined behavior. Therefore if I's range contains elements outside of K's range and we would load one such value, K would cause undefined behavior. In cases like hoisting/sinking, we still want the most generic range over all code paths to/from the hoist/sink point. As suggested in the patches related to D47339, I will refactor the handling of those scenarios and try to decouple it from this function as follow up, once we switched to a similar handling of metadata in most of combineMetadata. I updated some tests checking mostly the merging of metadata to keep the metadata of to dominating load. The most interesting one is probably test8 in test/Transforms/JumpThreading/thread-loads.ll. It contained a comment about the alias metadata preventing us to eliminate the branch, but it seem like the actual problem currently is that we merge the ranges of both loads and cannot eliminate the icmp afterwards. With this patch, we manage to eliminate the icmp, as the range of the first load excludes 8. Reviewers: efriedma, nlopes, davide Reviewed By: efriedma Differential Revision: https://reviews.llvm.org/D51629 llvm-svn: 345456		2018-10-27 16:53:45 +00:00
..
2007-07-25-DominatedLoop.ll	…
2007-07-25-InfiniteLoop.ll	…
2007-07-25-Loop.ll	…
2007-07-25-NestedLoop.ll	…
2007-07-25-SinglePredecessor.ll	…
2007-07-26-InterlockingLoops.ll	…
2007-07-26-NonRedundant.ll	…
2007-07-26-PhiErasure.ll	NewGVN: Update a number of xfailed tests to either be correct or note	2017-02-12 23:28:06 +00:00
2007-07-30-PredIDom.ll	…
2007-07-31-NoDomInherit.ll	…
2007-07-31-RedundantPhi.ll	…
2008-02-12-UndefLoad.ll	…
2008-02-13-NewPHI.ll	…
2008-07-02-Unreachable.ll	…
2008-12-09-SelfRemove.ll	Revert "NewGVN: Make unreachable blocks be marked with unreachable"	2017-01-30 17:06:55 +00:00
2008-12-12-RLE-Crash.ll	…
2008-12-14-rle-reanalyze.ll	…
2008-12-15-CacheVisited.ll	…
2009-01-21-SortInvalidation.ll	…
2009-01-22-SortInvalidation.ll	…
2009-03-10-PREOnVoid.ll	…
2009-07-13-MemDepSortFail.ll	…
2009-11-12-MemDepMallocBitCast.ll	…
2010-03-31-RedundantPHIs.ll	…
2010-05-08-OneBit.ll	…
2010-11-13-Simplify.ll	…
2011-04-27-phioperands.ll	…
2011-07-07-MatchIntrinsicExtract.ll	NewGVN: We were not really failing this testcase, because the instructions it was looking for are unused. GVN value numbers unused instructions, NewGVN does not. Fix the instructions to be used, so we eliminate the redundancies it's checking for, and un-XFAIL it	2017-03-06 20:01:31 +00:00
2011-09-07-TypeIdFor.ll	…
2012-05-22-PreCrash.ll	…
2016-08-30-MaskedScatterGather.ll	Support arbitrary address space pointers in masked gather/scatter intrinsics.	2017-05-03 12:28:54 +00:00
MemdepMiscompile.ll	…
assume-equal.ll	[Devirtualization] MemDep returns non-local !invariant.group dependencies	2017-01-12 11:33:58 +00:00
assumes.ll	[NewGVN] Fix handling of assumes	2018-05-22 17:38:22 +00:00
basic-cyclic-opt.ll	NewGVN: Clean up how we handle the INITIAL class so that everything in	2017-02-11 12:48:50 +00:00
basic-undef-test.ll	…
basic.ll	NewGVN: We weren't properly simplifying selects with equal arguments due to a thinko.	2017-08-24 02:43:17 +00:00
big-endian.ll	…
bitcast-of-call.ll	NewGVN: Update a number of xfailed tests to either be correct or note	2017-02-12 23:28:06 +00:00
br-identical.ll	…
calloc-load-removal.ll	NewGVN: Handle coercion of constant stores, loads, memory insts.	2017-04-02 13:23:44 +00:00
calls-nonlocal.ll	NewGVN: Update a number of xfailed tests to either be correct or note	2017-02-12 23:28:06 +00:00
calls-readonly.ll	…
commute.ll	…
completeness.ll	NewGVN: Fix PR 34473, by not using ExactlyEqualsExpression for finding	2017-09-30 23:51:55 +00:00
cond_br.ll	…
cond_br2.ll	Add address space mangling to lifetime intrinsics	2017-04-10 20:18:21 +00:00
condprop-xfail.ll	Add missing condprop-xfail.ll that contains the remaining xfail'd tests	2017-03-14 01:46:51 +00:00
condprop.ll	PredicateInfo: Support switch statements	2017-02-22 22:20:58 +00:00
crash-no-aa.ll	…
crash.ll	…
cyclic-phi-handling.ll	NewGVN: Fix PR 31501.	2017-01-07 00:01:42 +00:00
dbg-redundant-load.ll	[DebugInfo] Add DILabel metadata and intrinsic llvm.dbg.label.	2018-05-09 02:40:45 +00:00
deadstore.ll	NewGVN: Add basic dead and redundant store elimination	2017-01-27 02:37:11 +00:00
debugloc.ll	Do not propagate DebugLoc across basic blocks	2017-01-31 20:57:13 +00:00
edge.ll	PredicateInfo: Support switch statements	2017-02-22 22:20:58 +00:00
eliminate-callsite-inline.ll	[NewGVN] Mark function as changed if we erase instructions.	2018-09-07 11:41:34 +00:00
equivalent-phi.ll	NewGVN: Add a test case for equivalent phis.	2017-01-02 19:55:13 +00:00
fence.ll	[AliasAnalysis] Fences do not modify constant memory location	2017-01-20 00:21:33 +00:00
flags-simplify.ll	[NewGVN] Add tests for r340031.	2018-08-17 14:39:53 +00:00
flags.ll	NewGVN: Apply the fast math flags fix in r267113 to NewGVN as well.	2017-02-12 22:25:20 +00:00
fold-const-expr.ll	[NewGVN] Now that load coercion is enable, we pass this test.	2017-08-02 19:47:37 +00:00
fpmath.ll	…
funclet.ll	…
int_sideeffect.ll	Add an @llvm.sideeffect intrinsic	2017-11-08 21:59:51 +00:00
invariant.group.ll	Implement strip.invariant.group	2018-07-02 04:49:30 +00:00
invariant.start.ll	…
lifetime-simple.ll	Add address space mangling to lifetime intrinsics	2017-04-10 20:18:21 +00:00
load-constant-mem.ll	NewGVN: Rewrite DCE during elimination so we do it as well as old GVN did.	2017-03-10 00:32:33 +00:00
load-from-unreachable-predecessor.ll	…
loadforward.ll	[InstSimplify] Constant fold the new GEP in SimplifyGEPInst.	2017-06-06 10:17:14 +00:00
malloc-load-removal.ll	NewGVN: Handle coercion of constant stores, loads, memory insts.	2017-04-02 13:23:44 +00:00
memory-handling.ll	Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)	2018-01-19 17:13:12 +00:00
metadata-nonnull.ll	[GVN,NewGVN] Keep nonnull if K does not move.	2018-08-07 15:36:11 +00:00
metadata-simplify.ll	[NewGVN] Add tests for r340031.	2018-08-17 14:39:53 +00:00
no_speculative_loads_with_asan.ll	Hardware-assisted AddressSanitizer (llvm part).	2017-12-09 00:21:41 +00:00
noalias.ll	…
non-integral-pointers.ll	[GVN] Don't coerce non-integral pointers to integers or vice versa	2017-04-19 18:21:09 +00:00
non-local-offset.ll	…
nonescaping-malloc.ll	Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)	2018-01-19 17:13:12 +00:00
null-aliases-nothing.ll	…
opt-remarks.ll	[DebugInfo] Add DILabel metadata and intrinsic llvm.dbg.label.	2018-05-09 02:40:45 +00:00
pair_jumpthread.ll	[InstSimplify] fold extracting from std::pair (2/2)	2018-08-03 05:39:48 +00:00
phi-edge-handling.ll	NewGVN: Fix PHI evaluation bug exposed by new verifier. We were checking whether the incoming block was reachable instead of whether the specific edge was reachable	2017-03-18 15:41:36 +00:00
phi-of-ops-move-block.ll	[NewGVN] Split OpPHI detection and creation.	2018-04-20 16:37:13 +00:00
phi-translate-partial-alias.ll	…
pr10820.ll	NewGVN: Update a number of xfailed tests to either be correct or note	2017-02-12 23:28:06 +00:00
pr12979.ll	…
pr14166.ll	NewGVN: Update a number of xfailed tests to either be correct or note	2017-02-12 23:28:06 +00:00
pr17732.ll	Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)	2018-01-19 17:13:12 +00:00
pr17852.ll	…
pr24397.ll	…
pr24426.ll	…
pr25440.ll	…
pr28562.ll	…
pr31472.ll	Don't use our own incorrect version of isTriviallyDeadInstruction in NewGVN. Fixes PR/31472	2016-12-26 18:44:36 +00:00
pr31483.ll	NewGVN: Add forgotten testcase for PR 31483	2017-01-02 19:49:20 +00:00
pr31491.ll	NewGVN: Fix PR 31491 by ensuring that we touch the right instructions. Change to one based numbering so we can assert we don't cause the same bug again.	2016-12-29 22:15:12 +00:00
pr31501.ll	NewGVN: Fix PR 31501.	2017-01-07 00:01:42 +00:00
pr31573.ll	NewGVN: Fix PR 31573, a failure to verify memory congruency due to	2017-01-09 05:34:29 +00:00
pr31594.ll	NewGVN: Instead of changeToUnreachable, insert an instruction SimplifyCFG will turn into unreachable when it runs	2017-01-30 18:12:56 +00:00
pr31613.ll	NewGVN: This patch makes memory congruence work for all types of	2017-04-06 18:52:50 +00:00
pr31682.ll	NewGVN: Rewrite DCE during elimination so we do it as well as old GVN did.	2017-03-10 00:32:33 +00:00
pr31758.ll	NewGVN: Rewrite DCE during elimination so we do it as well as old GVN did.	2017-03-10 00:32:33 +00:00
pr32403.ll	NewGVN: Fix PR/33187. This is a bug caused by two things:	2017-06-06 17:15:28 +00:00
pr32607.ll	[InstCombine, NewGVN] remove FP undef from tests	2018-03-08 14:57:08 +00:00
pr32836.ll	[NewGVN] Update additionalUsers when we simplify to a value.	2017-05-24 02:30:24 +00:00
pr32838.ll	NewGVN: Fix PR32838.	2017-05-19 20:22:20 +00:00
pr32845.ll	[NewGVN] Actually check the NewGVN output.	2017-05-21 20:55:53 +00:00
pr32852.ll	[NewGVN] Don't derive incorrect implications.	2017-05-01 22:26:28 +00:00
pr32897.ll	NewGVN: Fix PR/33187. This is a bug caused by two things:	2017-06-06 17:15:28 +00:00
pr32934.ll	[NewGVN] Re-enable test now that the nondeterminism has been fixed.	2017-05-16 22:27:06 +00:00
pr32945.ll	[NewGVN] Replace predicate info leftovers.	2017-05-18 21:43:23 +00:00
pr32952.ll	[NewGVN] Fix a consistent order for phi nodes operands.	2017-05-09 16:58:28 +00:00
pr33014.ll	[NewGVN] Break infinite recursion in singleReachablePHIPath().	2017-05-18 23:22:44 +00:00
pr33086.ll	[NewGVN] Delete the old store when we find congruent to a load.	2017-05-19 04:06:10 +00:00
pr33116.ll	NewGVN: Fix PR 33116, the memoryphi version of bug 32838.	2017-05-21 23:41:58 +00:00
pr33185.ll	[InstrSimplify,NewGVN] Add option to ignore additional instr info when simplifying.	2018-08-17 14:39:04 +00:00
pr33187.ll	NewGVN: Fix PR/33187. This is a bug caused by two things:	2017-06-06 17:15:28 +00:00
pr33196.ll	[NewGVN] Fix the case where we have a phi-of-ops which goes away.	2017-08-03 21:17:49 +00:00
pr33204.ll	NewGVN: Fix PR33204 - We need to add memory users when we bypass memorydefs for loads, not just when we do it for stores.	2017-08-26 07:37:11 +00:00
pr33305.ll	[PredicateInfo] Use custom mangling to support ssa_copy with unnamed types.	2018-07-24 14:49:52 +00:00
pr33367.ll	NewGVN: Fix PR/33367, which was causing us to delete non-copy intrinsics accidentally in some rare cases	2018-01-09 20:12:42 +00:00
pr33432.ll	Fix PR/33305. caused by trying to simplify expressions in phi of ops that should have no leaders.	2017-09-02 02:18:44 +00:00
pr33461.ll	NewGVN: Evaluate phi of ops expressions before creating phi node	2017-09-30 23:51:54 +00:00
pr33720.ll	[NewGVN] Check for congruency of memory accesses.	2017-07-11 19:49:12 +00:00
pr34135.ll	NewGVN: Make sure we don't incorrectly use PredicateInfo when doing PHI of ops	2017-09-01 19:20:18 +00:00
pr34430.ll	NewGVN: Fix PR 34430 - we need to look through predicateinfo copies to detect self-cycles of phi nodes. We also need to not ignore certain types of arguments when testing whether the phi has a backedge or was originally constant.	2017-09-05 02:17:43 +00:00
pr34452.ll	NewGVN: Fix PR 34452 by passing instruction all the way down when we do aggregate value simplification	2017-09-05 02:17:42 +00:00
pr35074.ll	[NewGVN] Add ops as dependency if we cannot find a leader for ValueOp.	2018-04-19 15:05:47 +00:00
pr35125.ll	[NewGVN] Stop assuming PHI args ordering when looking at phi-of-ops.	2017-10-30 20:20:16 +00:00
pre-compare.ll	…
pre-new-inst.ll	…
predicates.ll	NewGVN: Start making use of predicateinfo pass.	2017-02-18 23:06:50 +00:00
propagate-ir-flags.ll	NewGVN: Fix PR 31686 and PR 31698 by rewriting store leader handling.	2017-01-20 21:04:30 +00:00
range.ll	[Local] Keep K's range if K does not move when combining metadata.	2018-10-27 16:53:45 +00:00
readattrs.ll	[BasicAA] Take attributes into account when requesting modref info for a call site	2017-03-01 13:19:51 +00:00
refine-stores.ll	NewGVN: This patch makes memory congruence work for all types of	2017-04-06 18:52:50 +00:00
rle-must-alias.ll	…
rle-no-phi-translate.ll	…
rle-nonlocal.ll	NewGVN: We pass rle-nonlocal, we just perform the replacement in a way that keeps the old name instead of the new one	2017-03-13 22:43:30 +00:00
rle.ll	Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)	2018-01-19 17:13:12 +00:00
stale-loop-info.ll	…
storeoverstore.ll	NewGVN: Evaluate phi of ops expressions before creating phi node	2017-09-30 23:51:54 +00:00
tbaa.ll	[Analysis] Generate more precise TBAA tags when one access encloses the other	2017-12-18 20:05:20 +00:00
unreachable_block_infinite_loop.ll	…
verify-memoryphi.ll	[NewGVN] Fix verification of MemoryPhis in verifyMemoryCongruency().	2017-05-15 18:50:53 +00:00
volatile-nonvolatile.ll	AliasAnalysis: Be less conservative about volatile than atomic.	2017-04-07 01:28:36 +00:00