llvm-6502

mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-10-04 00:54:53 +00:00

History

Andrew Trick 37aa33bc11 A new algorithm for computing LoopInfo. Temporarily disabled. -stable-loops enables a new algorithm for generating the Loop forest. It differs from the original algorithm in a few respects: - Not determined by use-list order. - Initially guarantees RPO order of block and subloops. - Linear in the number of CFG edges. - Nonrecursive. I didn't want to change the LoopInfo API yet, so the block lists are still inclusive. This seems strange to me, and it means that building LoopInfo is not strictly linear, but it may not be a problem in practice. At least the block lists start out in RPO order now. In the future we may add an attribute or wrapper analysis that allows other passes to assume RPO order. The primary motivation of this work was not to optimize LoopInfo, but to allow reproducing performance issues by decomposing the compilation stages. I'm often unable to do this with the current LoopInfo, because the loop tree order determines Loop pass order. Serializing the IR tends to invert the order, which reverses the optimization order. This makes it nearly impossible to debug interdependent loop optimizations such as LSR. I also believe this will provide more stable performance results across time. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@158790 91177308-0d34-0410-b5e6-96231b3b80d8		2012-06-20 05:23:33 +00:00
..
IPA	Handle intrinsics in GlobalsModRef. Fixes pr12351.	2012-03-28 21:31:24 +00:00
AliasAnalysis.cpp	Move the capture analysis from MemoryDependencyAnalysis to a more general place	2012-05-14 20:35:04 +00:00
AliasAnalysisCounter.cpp
AliasAnalysisEvaluator.cpp
AliasDebugger.cpp
AliasSetTracker.cpp
Analysis.cpp
BasicAliasAnalysis.cpp
BlockFrequencyInfo.cpp
BranchProbabilityInfo.cpp
CaptureTracking.cpp	Fix intendation.	2012-05-10 23:38:07 +00:00
CFGPrinter.cpp
CMakeLists.txt	Pull the implementation of the code metrics out of the inline cost	2012-03-16 05:51:52 +00:00
CodeMetrics.cpp	A pile of long over-due refactorings here. There are some very, very	2012-05-04 00:58:03 +00:00
ConstantFolding.cpp	Reapply r155682, making constant folding more consistent, with a fix to work	2012-04-27 17:50:22 +00:00
DbgInfoPrinter.cpp
DebugInfo.cpp	Actually support DW_TAG_rvalue_reference_type that we were trying	2012-05-19 01:36:37 +00:00
DIBuilder.cpp	Add support for enum forward declarations.	2012-06-01 00:22:32 +00:00
DominanceFrontier.cpp
DomPrinter.cpp
InlineCost.cpp	Fix typos found by http://github.com/lyda/misspell-check	2012-06-02 10:20:22 +00:00
InstCount.cpp
InstructionSimplify.cpp	Remove extraneous ';'.	2012-05-17 20:27:58 +00:00
Interval.cpp
IntervalPartition.cpp
IVUsers.cpp	Cleanup IVUsers::addUsersIfInteresting.	2012-03-22 17:47:33 +00:00
LazyValueInfo.cpp	allow LazyValueInfo::getEdgeValue() to reason about multiple edges from the same switch instruction by doing union of ranges (which may still be conservative, but it's more aggressive than before)	2012-05-18 21:02:10 +00:00
LibCallAliasAnalysis.cpp
LibCallSemantics.cpp
Lint.cpp	Always compute all the bits in ComputeMaskedBits.	2012-04-04 12:51:34 +00:00
LLVMBuild.txt
Loads.cpp
LoopDependenceAnalysis.cpp
LoopInfo.cpp	A new algorithm for computing LoopInfo. Temporarily disabled.	2012-06-20 05:23:33 +00:00
LoopPass.cpp	Take out the debug info probe stuff. It's making some changes to	2012-03-23 03:54:05 +00:00
Makefile
MemDepPrinter.cpp	Mark some static arrays as const.	2012-05-24 06:35:32 +00:00
MemoryBuiltins.cpp	remove calls to calloc if the allocated memory is not used (it was already being done for malloc)	2012-05-03 22:08:19 +00:00
MemoryDependenceAnalysis.cpp	Fix typos found by http://github.com/lyda/misspell-check	2012-06-02 10:20:22 +00:00
ModuleDebugInfoPrinter.cpp
NoAliasAnalysis.cpp
PathNumbering.cpp
PathProfileInfo.cpp
PathProfileVerifier.cpp
PHITransAddr.cpp
PostDominators.cpp
ProfileEstimatorPass.cpp
ProfileInfo.cpp
ProfileInfoLoader.cpp	Round 2 of dead private variable removal.	2012-06-06 19:47:08 +00:00
ProfileInfoLoaderPass.cpp	Round 2 of dead private variable removal.	2012-06-06 19:47:08 +00:00
ProfileVerifierPass.cpp
README.txt
RegionInfo.cpp	Rename the Region::block_iterator to Region::block_node_iterator, and	2012-05-04 20:55:23 +00:00
RegionPass.cpp	Rename the Region::block_iterator to Region::block_node_iterator, and	2012-05-04 20:55:23 +00:00
RegionPrinter.cpp	Rename the Region::block_iterator to Region::block_node_iterator, and	2012-05-04 20:55:23 +00:00
ScalarEvolution.cpp	Fix typos found by http://github.com/lyda/misspell-check	2012-06-02 10:20:22 +00:00
ScalarEvolutionAliasAnalysis.cpp
ScalarEvolutionExpander.cpp	LSR fix: add a missing phi check during IV hoisting.	2012-05-22 17:39:59 +00:00
ScalarEvolutionNormalization.cpp
SparsePropagation.cpp
Trace.cpp
TypeBasedAliasAnalysis.cpp
ValueTracking.cpp	Teach DeadStoreElimination to eliminate exit-block stores with phi addresses.	2012-05-10 18:57:38 +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))

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