llvm-6502/test/Transforms
Jingyue Wu 1241b83d61 [NVPTX] enable SpeculativeExecution in NVPTX
Summary:
SpeculativeExecution enables a series straight line optimizations (such
as SLSR and NaryReassociate) on conditional code. For example,

  if (...)
    ... b * s ...
  if (...)
    ... (b + 1) * s ...

speculative execution can hoist b * s and (b + 1) * s from then-blocks,
so that we have

  ... b * s ...
  if (...)
    ...
  ... (b + 1) * s ...
  if (...)
    ...

Then, SLSR can rewrite (b + 1) * s to (b * s + s) because after
speculative execution b * s dominates (b + 1) * s.

The performance impact of this change is significant. It speeds up the
benchmarks running EigenFloatContractionKernelInternal16x16
(ba68f42fa6/unsupported/Eigen/CXX11/src/Tensor/TensorContractionCuda.h (cl-526))
by roughly 2%. Some internal benchmarks that have the above code pattern
are improved by up to 40%. No significant slowdowns are observed on
Eigen CUDA microbenchmarks.

Reviewers: jholewinski, broune, eliben

Subscribers: llvm-commits, jholewinski

Differential Revision: http://reviews.llvm.org/D11201

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@242437 91177308-0d34-0410-b5e6-96231b3b80d8
2015-07-16 20:13:48 +00:00
..
ADCE
AddDiscriminators
AlignmentFromAssumptions
ArgumentPromotion
AtomicExpand/ARM
BBVectorize
BDCE
BranchFolding
CodeExtractor
CodeGenPrepare
ConstantHoisting
ConstantMerge
ConstProp
CorrelatedValuePropagation
DeadArgElim
DeadStoreElimination
EarlyCSE
EliminateAvailableExternally Don't change the visibility when converting a definition to a declaration. 2015-07-13 14:18:22 +00:00
Float2Int
FunctionAttrs
GCOVProfiling
GlobalDCE
GlobalOpt
GVN GVN: tolerate an instruction being replaced without existing in the leaderboard 2015-07-14 21:03:18 +00:00
IndVarSimplify [IndVars] Try to use existing values in RewriteLoopExitValues. 2015-07-09 18:46:12 +00:00
Inline Rename llvm.frameescape and llvm.framerecover to localescape and localrecover 2015-07-07 22:25:32 +00:00
InstCombine [InstCombine] Generalize sub of selects optimization to all BinaryOperators 2015-07-14 22:39:23 +00:00
InstMerge
InstSimplify [InstSimplify] Teach InstSimplify how to simplify extractelement 2015-07-13 01:15:53 +00:00
Internalize Internalize: internalize comdat members as a group, and drop comdat on such members. 2015-07-16 17:42:21 +00:00
IPConstantProp
IRCE
JumpThreading
LCSSA
LICM [LICM] Don't try to sink values out of loops without any exits 2015-07-12 03:53:05 +00:00
LoadCombine
LoopDeletion
LoopDistribute [LAA] Merge memchecks for accesses separated by a constant offset 2015-07-08 09:16:33 +00:00
LoopIdiom Remove unnecessary lines from the test in r242068. 2015-07-13 21:50:35 +00:00
LoopInterchange
LoopReroll
LoopRotate Disable loop re-rotation for -Oz (patch by Andrey Turetsky) 2015-07-10 10:37:09 +00:00
LoopSimplify
LoopStrengthReduce [LSR] don't attempt to promote ephemeral values to indvars 2015-07-13 03:28:53 +00:00
LoopUnroll Tidy-up test case from r242257. 2015-07-15 01:51:51 +00:00
LoopUnswitch
LoopVectorize Renamed some uses of unroll to interleave in the vectorizer. 2015-07-11 00:31:11 +00:00
LowerAtomic
LowerBitSets
LowerExpectIntrinsic
LowerInvoke
LowerSwitch
Mem2Reg
MemCpyOpt
MergeFunc Fix mergefunc infinite loop 2015-07-15 21:51:33 +00:00
MetaRenamer
NaryReassociate
ObjCARC
PartiallyInlineLibCalls
PhaseOrdering
PlaceSafepoints Rename llvm.frameescape and llvm.framerecover to localescape and localrecover 2015-07-07 22:25:32 +00:00
PruneEH
Reassociate
Reg2Mem
RewriteStatepointsForGC
SafeStack
SampleProfile
Scalarizer
ScalarRepl
SCCP [SCCP] Turn loads of null into undef instead of zero initialized values 2015-07-01 05:37:57 +00:00
SeparateConstOffsetFromGEP
SimplifyCFG
Sink
SLPVectorizer [SLPVectorizer] Try different vectorization factors for store chains 2015-07-08 23:40:55 +00:00
SpeculativeExecution
SROA [SROA] Don't de-atomic volatile loads and stores 2015-07-14 06:19:58 +00:00
StraightLineStrengthReduce [NVPTX] enable SpeculativeExecution in NVPTX 2015-07-16 20:13:48 +00:00
StripSymbols
StructurizeCFG
TailCallElim
TailDup
Util