Taewook Oh 9f93c9df69 Improve profile-guided heuristics to use estimated trip count. ... Summary: Existing heuristic uses the ratio between the function entry frequency and the loop invocation frequency to find cold loops. However, even if the loop executes frequently, if it has a small trip count per each invocation, vectorization is not beneficial. On the other hand, even if the loop invocation frequency is much smaller than the function invocation frequency, if the trip count is high it is still beneficial to vectorize the loop. This patch uses estimated trip count computed from the profile metadata as a primary metric to determine coldness of the loop. If the estimated trip count cannot be computed, it falls back to the original heuristics. Reviewers: Ayal, mssimpso, mkuper, danielcdh, wmi, tejohnson Reviewed By: tejohnson Subscribers: tejohnson, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D32451 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@305729 91177308-0d34-0410-b5e6-96231b3b80d8		2017-06-19 18:48:58 +00:00
..
ADCE	Fix a test to check assembly output instead of bitcode.	2017-03-31 23:22:19 +00:00
AddDiscriminators	[AddDiscriminators] Assign discriminators to MemIntrinsic calls.	2017-04-11 19:07:30 +00:00
AlignmentFromAssumptions
ArgumentPromotion	[ArgPromotion] Add a testcase for PR32917	2017-05-05 08:40:24 +00:00
AtomicExpand	Add address space mangling to lifetime intrinsics	2017-04-10 20:18:21 +00:00
BBVectorize	Replace hardcoded intrinsic list with speculatable attribute.	2017-05-03 02:26:10 +00:00
BDCE	[Analysis] Support bitreverse in -demanded-bits pass	2017-04-13 16:44:25 +00:00
BranchFolding
CodeExtractor	[PartialInlining] Support shrinkwrap life_range markers	2017-06-11 20:46:05 +00:00
CodeGenPrepare	[CGP, x86] add tests for potential memcmp expansion; NFC	2017-06-08 20:40:39 +00:00
ConstantHoisting	[ConstHoisting] Add BFI in constanthoisting pass and select the best insertion	2017-04-21 15:50:16 +00:00
ConstantMerge	Don't merge global constants with non-dbg metadata.	2017-03-09 00:03:37 +00:00
ConstProp	[X86] Replace 'REQUIRES: x86' in tests with 'REQUIRES: x86-registered-target' which seems to be the correct way to make them run on an x86 build.	2017-06-04 08:21:58 +00:00
Coroutines	[coroutines] PR33271: Remove stray coro.save intrinsics during CoroSplit	2017-06-02 02:18:36 +00:00
CorrelatedValuePropagation	Add address space mangling to lifetime intrinsics	2017-04-10 20:18:21 +00:00
CountingFunctionInserter
CrossDSOCFI	[cfi] CFI-ICall for ThinLTO.	2017-06-16 00:18:29 +00:00
DCE	[InstSimplify] Don't constant fold or DCE calls that are marked nobuiltin	2017-06-09 23:18:11 +00:00
DeadArgElim	Replace hardcoded intrinsic list with speculatable attribute.	2017-05-03 02:26:10 +00:00
DeadStoreElimination	Add address space mangling to lifetime intrinsics	2017-04-10 20:18:21 +00:00
EarlyCSE	[EarlyCSE] Make PhiToCheck in removeMSSA() a set.	2017-06-14 19:29:53 +00:00
EliminateAvailableExternally
Float2Int
ForcedFunctionAttrs
FunctionAttrs	Support arbitrary address space pointers in masked gather/scatter intrinsics.	2017-05-03 12:28:54 +00:00
FunctionImport	Revert r295861, "[ModuleSummaryAnalysis] Don't crash when referencing unnamed globals."	2017-05-03 00:18:48 +00:00
GCOVProfiling
GlobalDCE	[PM] Teach the PGO instrumentation pasess to run GlobalDCE before	2017-05-25 07:15:09 +00:00
GlobalMerge	Align definition of DW_OP_plus with DWARF spec [3/3]	2017-06-14 13:14:38 +00:00
GlobalOpt	[GlobalOpt] Correctly update metadata when localizing a global.	2017-04-27 18:39:08 +00:00
GlobalSplit	Fix one-after-the-end type metadata handling in globalsplit.	2017-03-07 22:18:48 +00:00
GuardWidening
GVN	Revert rL305578. There is still some buildbot failure to be fixed.	2017-06-16 23:14:35 +00:00
GVNHoist	Disable GVN Hoist due to still more bugs being found in it. There is	2017-04-27 00:28:03 +00:00
GVNSink	Changed a comparison operator for std::stable_sort to implement strict weak ordering.	2017-06-08 17:27:40 +00:00
IndVarSimplify	[IndVars] Add an option to be able to disable LFTR	2017-06-09 06:11:59 +00:00
InferAddressSpaces	InferAddressSpaces: Avoid assertion failure with replacing identical	2017-06-08 13:20:55 +00:00
InferFunctionAttrs	Add argmononly attribute to strlen and wcslen, i.e. they only read memory (string) passed to them.	2017-06-18 03:10:26 +00:00
Inline	Inliner: Don't remove calls to readnone+nounwind (but not always_inline) functions in the AlwaysInliner	2017-06-12 23:01:17 +00:00
InstCombine	[InstCombine] Make sure AddReachableCodeToWorklist sets MadeIRChange	2017-06-19 18:00:27 +00:00
InstMerge	Remove the load hoisting code of MLSM, it is completely subsumed by GVNHoist	2017-01-24 19:55:36 +00:00
InstNamer	Add basic test case for -instnamer	2017-05-08 23:18:46 +00:00
InstSimplify	[InstSimplify] Constant fold the new GEP in SimplifyGEPInst.	2017-06-06 10:17:14 +00:00
InterleavedAccess	Added tests for X86InterleavedStore.	2017-06-06 21:08:00 +00:00
Internalize
IPConstantProp	[IPCP] Don't propagate return value for naked functions.	2017-02-04 19:44:14 +00:00
IRCE	[IRCE] Canonicalize pre/post loops after the blocks are added into parent loop	2017-06-06 14:54:01 +00:00
JumpThreading	[JumpThreading] Safely replace uses of condition	2017-05-23 13:36:25 +00:00
LCSSA
LICM	Add argmononly attribute to strlen and wcslen, i.e. they only read memory (string) passed to them.	2017-06-18 03:10:26 +00:00
LoadCombine	[LoadCombine] Avoid analysing dead basic blocks	2017-04-12 08:07:55 +00:00
LoadStoreVectorizer	Skip bitcasts while looking for GEP in LoadStoreVectorizer	2017-04-25 18:00:08 +00:00
LoopDataPrefetch
LoopDeletion	[Loop Deletion] Delete loops that are never executed	2017-05-03 11:47:11 +00:00
LoopDistribute	Move test of lazy BFI with ORE to a generic directory	2017-01-13 00:16:23 +00:00
LoopIdiom	[Atomics] Rename and change prototype for atomic memcpy intrinsic	2017-06-16 14:43:59 +00:00
LoopInterchange
LoopLoadElim	[PM] Port LoopLoadElimination to the new pass manager and wire it into	2017-01-27 01:32:26 +00:00
LoopPredication	Loop predication expand both sides of the widened condition	2017-02-27 15:44:49 +00:00
LoopReroll
LoopRotate	Cleanup tests to not share a DISubprogram between multiple Functions.	2017-05-04 16:24:31 +00:00
LoopSimplify	[BasicBlockUtils] Use getFirstNonPHIOrDbg to set debugloc for instructions created in SplitBlockPredecessors	2017-02-14 21:10:40 +00:00
LoopSimplifyCFG
LoopStrengthReduce	Revert r304824 "Fix PR23384 (part 3 of 3)"	2017-06-19 17:57:15 +00:00
LoopUnroll	[LoopUnroll] Fix a test. REQUIRE should be REQUIRES.	2017-05-12 15:30:58 +00:00
LoopUnswitch	[LoopUnswitch] Don't remove instructions with side effects.	2017-04-29 00:12:18 +00:00
LoopVectorize	Improve profile-guided heuristics to use estimated trip count.	2017-06-19 18:48:58 +00:00
LoopVersioning	[LAA] Correctly return a half-open range in expandBounds	2017-04-05 09:24:26 +00:00
LoopVersioningLICM	[LAA] Correctly return a half-open range in expandBounds	2017-04-05 09:24:26 +00:00
LowerAtomic
LowerExpectIntrinsic	Fix builin_expect lowering bug	2017-06-07 18:32:24 +00:00
LowerGuardIntrinsic
LowerInvoke
LowerSwitch	LowerSwitch: Fix producing invalid IR on unreachable code	2017-04-21 23:54:12 +00:00
LowerTypeTests	Change YAML traits for vector<string> to flow_vector.	2017-06-16 20:50:41 +00:00
Mem2Reg	Add test coverage for mem2reg dbg.declare lowering.	2017-04-21 22:13:55 +00:00
MemCpyOpt	Add address space mangling to lifetime intrinsics	2017-04-10 20:18:21 +00:00
MergeFunc	MergeFunctions: Preserve debug info in thunks, under option -mergefunc-preserve-debug-info	2017-01-21 02:02:56 +00:00
MetaRenamer	[MetaRenamer] Don't rename library functions.	2017-03-23 23:21:07 +00:00
NameAnonGlobals
NaryReassociate
NewGVN	NewGVN: Fix PR 33461, caused by slightly overzealous verification.	2017-06-19 00:24:00 +00:00
ObjCARC	Replace hardcoded intrinsic list with speculatable attribute.	2017-05-03 02:26:10 +00:00
PartiallyInlineLibCalls
PGOProfile	[PGO] Update VP metadata after memory intrinsic optimization	2017-06-13 20:44:08 +00:00
PhaseOrdering	Make globalaa-retained.ll test catching more cases.	2017-04-18 13:29:26 +00:00
PlaceSafepoints
PreISelIntrinsicLowering
PruneEH
Reassociate	[Reassociate] Support some reassociation of vector xors	2017-06-19 16:23:46 +00:00
Reg2Mem
RewriteStatepointsForGC	[RS4GC] Drop invalid metadata after pointers are relocated	2017-06-12 21:26:53 +00:00
SafeStack	Align definition of DW_OP_plus with DWARF spec [3/3]	2017-06-14 13:14:38 +00:00
SampleProfile	Do not early-inline recursive calls in sample profile loader.	2017-06-08 20:11:57 +00:00
Scalarizer	[Scalarizer] Handle scalar arguments in vector GEP	2017-03-31 06:29:49 +00:00
SCCP	[SCCP] Resolve indirect branch target when possible.	2017-04-10 00:33:25 +00:00
SeparateConstOffsetFromGEP	AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel	2017-03-21 21:39:51 +00:00
SimpleLoopUnswitch	[PM/Unswitch] Fix a bug in the domtree update logic for the new unswitch	2017-05-25 06:33:36 +00:00
SimplifyCFG	[ValueTracking] Teach isSafeToSpeculativelyExecute() about the speculatable attribute	2017-04-28 21:13:09 +00:00
Sink	[Sink] Fix predicate in legality check	2017-06-09 19:31:10 +00:00
SLPVectorizer	[X86][SLM] Add SLM arithmetic vectorization tests	2017-06-10 19:16:09 +00:00
SpeculativeExecution	Revert r301950: SpeculativeExecution: Stop using whitelist for costs	2017-05-10 12:30:07 +00:00
SROA	[SROA] Add support for non-integral pointers	2017-06-17 20:28:13 +00:00
StraightLineStrengthReduce	AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel	2017-03-21 21:39:51 +00:00
StripDeadPrototypes
StripSymbols	[StripDeadDebug/DIFinder] Track inlined SPs	2017-04-11 13:32:11 +00:00
StructurizeCFG	Revert "StructurizeCFG: Directly invert cmp instructions"	2017-04-24 20:25:01 +00:00
TailCallElim	[TRE] Improve code motion in TRE, use AA to tell whether a load can be moved before a call that writes to memory.	2017-06-19 15:21:18 +00:00
ThinLTOBitcodeWriter	Write summaries for merged modules when splitting modules for ThinLTO.	2017-06-08 23:01:49 +00:00
Util	PredicateInfo: Don't insert conditional info when a conditional branch jumps to the same target regardless of condition	2017-06-14 21:19:52 +00:00
WholeProgramDevirt	Skip CFI for dead functions.	2017-06-02 18:24:23 +00:00