(Headers first)
It has been generally agreed that we need to do this reorg, and now
seems like the perfect time. Some major pass reorganization is in the
works.
This does not have to be the final word on the matter. The consensus
among those working on the code is that it's much better than what we
had and a better starting point for future bike shedding.
Note that the previous organization was designed to allow separate
analysis and optimization libraries. It turns out this is an
artificial distinction and not an important goal.
Running sil-opt with the new call graph printing pass (which uses
llvm::outs) was resulting in some spew about "LLVM IO Error" because we
could potentially close stdout twice.
Swift SVN r32041
Till now, a SIL module would be only verified if an optimization has changed it. But if there were no changes, then no verification would happen and some SIL module format errors would stay unnoticed. This was happening in certain cases when reading a textual SIL module representation, which turned out to be broken, but SIL verifier wouldn't catch it.
Swift SVN r31863
To invoke the front-end on a SIL with whole-module optimizations enabled, execute:
swiftc -frontend myfile.sil
To invoke the front-end on a SIL without whole-module optimizations enabled, add a -primary-file option:
swiftc -frontend -primary-file myfile.sil
To invoke a sil-opt with whole-module optimizations enabled, use the -wmo option:
sil-opt myfile.sil -wmo
This change was need to be able to write SIL unit tests which should be compiled in the WMO mode.
Swift SVN r31862
I also added a macro called INITIALIZE_LLVM(argc, argv) which moves this logic
into one place and should be used at the beginning of *all* binaries. It
initializes an LLVM shutdown object, sets up the pretty stack trace, and then
initializes all of the parts of LLVM. This will make it easy to update this in
the future.
The reason why a macro was required was because of llvm_shutdown_obj, an RAII
object that cleans up LLVM. It has to be at the function level scope of the main
function.
Swift SVN r31815
This allows one by editing one place to create an Analysis kind and ensure that
your analysis is properly registered in all of the relevant places for use with
swift/sil-opt.
If we add tools like a sil bugpoint, this will make it easy to add analysis
without having to know about all of these locations.
It also standardizes the create***Analysis API to take a SILModule and a
SILPassManager. All passes take in a SILModule and some take in a
SILPassManager. By standardizing the API and in the cases where SILPassManager
is not used, just making it a dead argument enables metaprogramming.
Swift SVN r31420
This was the last bit of boilerplate associated with adding a new pass that had
not been added to Passes.def. Now to add a new pass, all you need to do is:
1. Add the pass to Passes.def.
2. Implement the pass in SILPasses and update the CMake files.
Swift SVN r29856
These classes don't show up well in generated headers (rdar://problem/20855568),
can't actually be allocated from Objective-C (rdar://problem/17184317), and
make the story of "what is exposed to Objective-C" more complicated. Better
to just disallow them.
All classes are still "id-compatible" in that they can be converted to
AnyObject and passed to Objective-C, they secretly implement NSObjectProtocol
(via our SwiftObject root class), and their members can still be individually
exposed to Objective-C.
The frontend flag -disable-objc-attr-requires-foundation-module will disable
this requirement as well, which is still necessary for both the standard
library and a variety of tests I didn't feel like transforming.
Swift SVN r29760
I was always confused by the inconsistency of -sil-print-all and the other -sil-print options.
So I thought it would be a good idea to make -sil-print-all also an llvm option.
Together with some other internal options which are only used in the pass manager.
Swift SVN r29365
Teach dominator based simplifications to also thread dominated edges.
The code now handles cond_br and switch_enum terminators for both value based
simplifications (where the use is dominated) and jump threading edges (the edge
is dominated).
Update simplify_cfg.sil test cases for split edges.
This also handles the test case from rdar://20390647.
Swift SVN r27843
The GlobalPropertyOpt pass performs a static analysis over the whole module.
If it can prove that an array property call (_getArrayPropertyIsNativeNoTypeCheck) always yiels true,
then it replaces the call with a literal-true.
The pass runs on the high-level SIL using the array semantics calls.
Currently it only handles the isNativeNoTypeCheck array property, but in future it might handle additinal properties
(therefore I chose this general name for it).
It gives +24% on DeltaBlue.
Swift SVN r27361
To set the PassKind automatically, I needed to refactor some code of the pass manager and the pass definitions.
The main changes are:
1) SILPassManager now has an add-function for each pass: PM.add(createP()) -> PM.addP()
2) I removed the ARGS argument in Passes.def, which we didn't use anyway.
Swift SVN r26756
This commit splits DominanceAnalysis into two analysis (Dom and PDom) that
can be cached and invalidates using the common FunctionAnalysisBase interface
independent of one another.
Swift SVN r26643
I completely missed that one of the CondFailOpt optimization was already implemented in SimplifyCFG.
I move the other optimization also into SimplifyCFG because both share some code.
Swift SVN r26626
Separate InputFileKind from SourceFileKind, FrontendOptions will now use
InputFileKind, while Module will use SourceFileKind.
This is in preparation for adding an input file kind for LLVM IR.
rdar://19048891
Swift SVN r25555
Also into a separate file.
Before (swift/Serialization/SerializedModuleLoader.h):
ModuleStatus
SerializedModuleLoader::ValidationInfo
SerializedModuleLoader::ExtendedValidationInfo
SerializedModuleLoader::isSerializedAST
SerializedModuleLoader::validateSerializedAST
After (swift/Serialization/Validation.h):
serialization::Status
serialization::ValidationInfo
serialization::ExtendedValidationInfo
serialization::isSerializedAST
serialization::validateSerializedAST
No functionality change, just a lot of renaming and a bit of reorganizing.
Swift SVN r25226
Main changes:
*) Instruction costs are not counted for blocks which are dead after inlining
*) Terminator instructions which get constant after inlining increase the threshold
*) Calls inside loops increase the threshold
In theory this should be a step towards making the performance not so dependent on the inlining heuristic.
But I must admit that I still did some fine tuning of all the parameters to get the best results.
Improvements in the benchmarks:
-O:
Chars: +11%
CommonMarkRender: +11%
DollarReduce: +22%
ForLoops: +22%
Forest: +10%
HeapSort: +36%
ImageProc: +14%
StrCat: +14%
StrComplexWalk: +70%
StrToInt: +11%
StringWalk: +99%
-Ounchecked:
Ary: +40%
Ary2: +30%
EditDistance: +22%
Forest: +18%
HeapSort: +50%
Histogram: +11%
StrCat: +12%
StrComplexWalk: +63%
StrSplitter: +11%
StrToInt: +17%
StringWalk: +75%
Regressions (I will file radars for them):
-Ounchecked:
PolymorphicCalls: -21%
QuickSort: -22%
Rectangles: -12%
Code size of the PerfTests_O decreased by 8%
Code size of the PerfTests_Ounchecked increased by 1%
Swift SVN r24801
