llvm-mirror

mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 03:33:20 +01:00

History

Renato Golin bf4b959200 [ARM] Adding IEEE-754 SIMD detection to loop vectorizer Some SIMD implementations are not IEEE-754 compliant, for example ARM's NEON. This patch teaches the loop vectorizer to only allow transformations of loops that either contain no floating-point operations or have enough allowance flags supporting lack of precision (ex. -ffast-math, Darwin). For that, the target description now has a method which tells us if the vectorizer is allowed to handle FP math without falling into unsafe representations, plus a check on every FP instruction in the candidate loop to check for the safety flags. This commit makes LLVM behave like GCC with respect to ARM NEON support, but it stops short of fixing the underlying problem: sub-normals. Neither GCC nor LLVM have a flag for allowing sub-normal operations. Before this patch, GCC only allows it using unsafe-math flags and LLVM allows it by default with no way to turn it off (short of not using NEON at all). As a first step, we push this change to make it safe and in sync with GCC. The second step is to discuss a new sub-normal's flag on both communitues and come up with a common solution. The third step is to improve the FastMath flags in LLVM to encode sub-normals and use those flags to restrict NEON FP. Fixes PR16275. llvm-svn: 266363		2016-04-14 20:42:18 +00:00
..
AliasAnalysis.cpp	NFC: make AtomicOrdering an enum class	2016-04-06 21:19:33 +00:00
AliasAnalysisEvaluator.cpp
AliasSetTracker.cpp	NFC: make AtomicOrdering an enum class	2016-04-06 21:19:33 +00:00
Analysis.cpp	[ThinLTO] Move summary computation from BitcodeWriter to new pass	2016-04-11 13:58:45 +00:00
AssumptionCache.cpp
BasicAliasAnalysis.cpp	Don't IPO over functions that can be de-refined	2016-04-08 00:48:30 +00:00
BlockFrequencyInfo.cpp
BlockFrequencyInfoImpl.cpp
BranchProbabilityInfo.cpp	Const correctness for BranchProbabilityInfo (NFC)	2016-04-07 21:59:28 +00:00
CallGraph.cpp
CallGraphSCCPass.cpp
CallPrinter.cpp
CaptureTracking.cpp
CFG.cpp
CFGPrinter.cpp
CFLAliasAnalysis.cpp	[CFLAA] Fix up code style a bit. NFC.	2016-04-13 23:27:37 +00:00
CGSCCPassManager.cpp
CMakeLists.txt	[ThinLTO] Move summary computation from BitcodeWriter to new pass	2016-04-11 13:58:45 +00:00
CodeMetrics.cpp
ConstantFolding.cpp	Propagate Undef in llvm.cos Intrinsic	2016-04-08 18:21:11 +00:00
CostModel.cpp	[CodeGen] Teach LLVM how to lower @llvm.{min,max}num to {MIN,MAX}NAN	2016-04-14 07:13:24 +00:00
Delinearization.cpp
DemandedBits.cpp
DependenceAnalysis.cpp	[DependenceAnalysis] Check if result of getConstantPart is null	2016-04-04 18:13:18 +00:00
DivergenceAnalysis.cpp	[DivergenceAnalysis] Treat PHI with incoming undef as constant	2016-04-14 17:42:47 +00:00
DominanceFrontier.cpp
DomPrinter.cpp
EHPersonalities.cpp
GlobalsModRef.cpp	Don't IPO over functions that can be de-refined	2016-04-08 00:48:30 +00:00
InlineCost.cpp	Return immediately from analyzeCall if analyzeBlock returns false.	2016-04-13 21:20:22 +00:00
InstCount.cpp
InstructionSimplify.cpp	[InstCombine] We folded an fcmp to an i1 instead of a vector of i1	2016-04-13 06:55:52 +00:00
Interval.cpp
IntervalPartition.cpp
IteratedDominanceFrontier.cpp
IVUsers.cpp
LazyCallGraph.cpp
LazyValueInfo.cpp
Lint.cpp
LLVMBuild.txt
Loads.cpp	Don't IPO over functions that can be de-refined	2016-04-08 00:48:30 +00:00
LoopAccessAnalysis.cpp	[SCEV][LAA] Add tests for SCEV expression transformations performed during LAA	2016-04-14 16:08:45 +00:00
LoopInfo.cpp
LoopPass.cpp
LoopPassManager.cpp
LoopUnrollAnalyzer.cpp
MemDepPrinter.cpp
MemDerefPrinter.cpp
MemoryBuiltins.cpp	Calculate __builtin_object_size when pointer depends on a condition	2016-04-13 12:25:25 +00:00
MemoryDependenceAnalysis.cpp	NFC: make AtomicOrdering an enum class	2016-04-06 21:19:33 +00:00
MemoryLocation.cpp
ModuleDebugInfoPrinter.cpp
ModuleSummaryAnalysis.cpp	[ThinLTO] Move summary computation from BitcodeWriter to new pass	2016-04-11 13:58:45 +00:00
ObjCARCAliasAnalysis.cpp
ObjCARCAnalysisUtils.cpp
ObjCARCInstKind.cpp
OrderedBasicBlock.cpp
PHITransAddr.cpp
PostDominators.cpp
PtrUseVisitor.cpp
README.txt
RegionInfo.cpp
RegionPass.cpp
RegionPrinter.cpp
ScalarEvolution.cpp	[SCEV][LAA] Add tests for SCEV expression transformations performed during LAA	2016-04-14 16:08:45 +00:00
ScalarEvolutionAliasAnalysis.cpp
ScalarEvolutionExpander.cpp	Re-commit [SCEV] Introduce a guarded backedge taken count and use it in LAA and LV	2016-04-08 14:29:09 +00:00
ScalarEvolutionNormalization.cpp
ScopedNoAliasAA.cpp
SparsePropagation.cpp
StratifiedSets.h	[CFLAA] Fix up code style a bit. NFC.	2016-04-13 23:27:37 +00:00
TargetLibraryInfo.cpp	[NVPTX] Infer __nvvm_reflect as nounwind, readnone	2016-03-31 21:29:57 +00:00
TargetTransformInfo.cpp	[ARM] Adding IEEE-754 SIMD detection to loop vectorizer	2016-04-14 20:42:18 +00:00
Trace.cpp
TypeBasedAliasAnalysis.cpp
ValueTracking.cpp	Simplify strlen to a subtraction for certain cases.	2016-04-13 14:31:06 +00:00
VectorUtils.cpp	[InstCombine] We folded an fcmp to an i1 instead of a vector of i1	2016-04-13 06:55:52 +00:00

README.txt

Analysis Opportunities:

//===---------------------------------------------------------------------===//

In test/Transforms/LoopStrengthReduce/quadradic-exit-value.ll, the
ScalarEvolution expression for %r is this:

  {1,+,3,+,2}<loop>

Outside the loop, this could be evaluated simply as (%n * %n), however
ScalarEvolution currently evaluates it as

  (-2 + (2 * (trunc i65 (((zext i64 (-2 + %n) to i65) * (zext i64 (-1 + %n) to i65)) /u 2) to i64)) + (3 * %n))

In addition to being much more complicated, it involves i65 arithmetic,
which is very inefficient when expanded into code.

//===---------------------------------------------------------------------===//

In formatValue in test/CodeGen/X86/lsr-delayed-fold.ll,

ScalarEvolution is forming this expression:

((trunc i64 (-1 * %arg5) to i32) + (trunc i64 %arg5 to i32) + (-1 * (trunc i64 undef to i32)))

This could be folded to

(-1 * (trunc i64 undef to i32))

//===---------------------------------------------------------------------===//