Chris Lattner [Wed, 21 Jan 2009 07:43:11 +0000 (07:43 +0000)]
Add a bit to IdentifierInfo that acts as a simple predicate which
tells us whether Preprocessor::HandleIdentifier needs to be called.
Because this method is only rarely needed, this saves a call and a
bunch of random checks. This drops the time in HandleIdentifier
from 3.52ms to .98ms on cocoa.h on my machine.
Ted Kremenek [Wed, 21 Jan 2009 06:57:53 +0000 (06:57 +0000)]
Get RegionStore to work with the retain/release checker and its test cases.
Because the RegionStore can reason about values beyond the reasoning power of BasicStore, this patch splits some of the test cases for the retain/release checker to have versions that are handled by RegionStore (more warnings) and BasicStore (less warnings).
Ted Kremenek [Tue, 20 Jan 2009 23:28:34 +0000 (23:28 +0000)]
Fix: <rdar://problem/6510344> [pth] PTH slows down regular lexer considerably (when it has substantial work)
Changes to IdentifierTable:
- High-level summary: StringMap never owns IdentifierInfos. It just
references them.
- The string map now has StringMapEntry<IdentifierInfo*> instead of
StringMapEntry<IdentifierInfo>. The IdentifierInfo object is
allocated using the same bump pointer allocator as used by the
StringMap.
Changes to IdentifierInfo:
- Added an extra pointer to point to the
StringMapEntry<IdentifierInfo*> in the string map. This pointer
will be null if the IdentifierInfo* is *only* used by the PTHLexer
(that is it isn't in the StringMap).
Algorithmic changes:
- Non-PTH case:
IdentifierInfo::get() will always consult the StringMap first to
see if we have an IdentifierInfo object. If that StringMapEntry
references a null pointer, we allocate a new one from the BumpPtrAllocator
and update the reference in the StringMapEntry.
- PTH case:
We do the same lookup as with the non-PTH case, but if we don't get
a hit in the StringMap we do a secondary lookup in the PTHManager for
the IdentifierInfo. If we don't find an IdentifierInfo we create a
new one as in the non-PTH case. If we do find and IdentifierInfo
in the PTHManager, we update the StringMapEntry to refer to it so
that the IdentifierInfo will be found on the next StringMap lookup.
This way we only do a binary search in the PTH file at most once
for a given IdentifierInfo. This greatly speeds things up for source
files containing a non-trivial amount of code.
Performance impact:
While these changes do add some extra indirection in
IdentifierTable to access an IdentifierInfo*, I saw speedups even
in the non-PTH case as well.
Non-PTH: For -fsyntax-only on Cocoa.h, we see a 6% speedup.
PTH (with Cocoa.h in token cache): 11% speedup.
I also did an experiment where we did -fsyntax-only on a source file
including a large header and Cocoa.h, but the token cache did not
contain the larger header. For this file, we were seeing a performance
*regression* when using PTH of 3% over non-PTH. Now we are seeing
a performance improvement of 9%!
Tests:
The serialization tests are now failing. I looked at this extensively,
and I my belief is that this change is unmasking a bug rather than
introducing a new one. I have disabled the serialization tests for now.
Sebastian Redl [Tue, 20 Jan 2009 22:23:13 +0000 (22:23 +0000)]
Provide a placement new taking an ASTContext argument.
This allows more concise syntax when allocating an object using the ASTContext's allocator.
Convert a few allocations to this operator to for test purposes.
Daniel Dunbar [Tue, 20 Jan 2009 21:29:14 +0000 (21:29 +0000)]
ccc: Allow downstream tools to be aware of final output name.
- This is a hack to allow the Darwin linker to get -final_output when
doing universal builds; the mechanism should be generalized.
Chris Lattner [Tue, 20 Jan 2009 19:11:22 +0000 (19:11 +0000)]
Optimize Declarator to avoid malloc/free traffic for the argument list of a
function DeclaratorChunk in common cases. This uses a fixed array in
Declarator when it is small enough for the first function declarator chunk
in a declarator.
This eliminates all malloc/free traffic from DeclaratorChunk::getFunction
when running on Cocoa.h except for five functions: signal/bsd_signal/sigset,
which have multiple Function DeclChunk's, and
CFUUIDCreateWithBytes/CFUUIDGetConstantUUIDWithBytes, which take more than
16 arguments.
Douglas Gregor [Tue, 20 Jan 2009 04:25:11 +0000 (04:25 +0000)]
Remove the TopLevelDecls from TranslationUnit, since all of those decls are owned by the ASTContext's TranslationUnitDecl. There are definitely some leaking Decls now that I'll tackle tomorrow
Douglas Gregor [Tue, 20 Jan 2009 01:17:11 +0000 (01:17 +0000)]
Remove ScopedDecl, collapsing all of its functionality into Decl, so
that every declaration lives inside a DeclContext.
Moved several things that don't have names but were ScopedDecls (and,
therefore, NamedDecls) to inherit from Decl rather than NamedDecl,
including ObjCImplementationDecl and LinkageSpecDecl. Now, we don't
store empty DeclarationNames for these things, nor do we try to insert
them into DeclContext's lookup structure.
The serialization tests are temporarily disabled. We'll re-enable them
once we've sorted out the remaining ownership/serialiazation issues
between DeclContexts and TranslationUnion, DeclGroups, etc.
Ted Kremenek [Tue, 20 Jan 2009 00:47:45 +0000 (00:47 +0000)]
Dead stores checker: Fix <rdar://problem/6506065> by being more selective when say that a store is dead even though the computed value is used in the enclosing expression.
Ted Kremenek [Mon, 19 Jan 2009 23:13:15 +0000 (23:13 +0000)]
PTH: Emitted tokens now consist of 12 bytes that are loaded used 3 32-bit loads. This reduces user time but increases system time because of the slightly larger PTH file. Although there is no performance win on Cocoa.h and -Eonly, overall this seems like a good step.
Douglas Gregor [Mon, 19 Jan 2009 19:26:10 +0000 (19:26 +0000)]
Centralize error reporting of improper uses of incomplete types in the
new DiagnoseIncompleteType. It provides additional information about
struct/class/union/enum types when possible, either by pointing to the
forward declaration of that type or by pointing to the definition (if
we're in the process of defining that type).
Fixes <rdar://problem/6500531>.
Daniel Dunbar [Mon, 19 Jan 2009 18:50:49 +0000 (18:50 +0000)]
ccc: Add installation of ccc; based on patch from Mike Stump.
- This doesn't follow normal installation procedure of python
code, but no sense trying too hard since ccc will be moved to
C++.
Chris Lattner [Mon, 19 Jan 2009 08:02:45 +0000 (08:02 +0000)]
rearrange SourceManager methods to group them by kind. Remove the
SourceManager::getFileCharacteristic(FileID) method: we need a
full location to know if it is a system header etc in the future.
Chris Lattner [Mon, 19 Jan 2009 06:46:35 +0000 (06:46 +0000)]
Make SourceLocation::getFileLoc private to reduce the API exposure of
SourceLocation. This requires making some cleanups to token pasting
and _Pragma expansion.
Sebastian Redl [Sun, 18 Jan 2009 18:53:16 +0000 (18:53 +0000)]
Convert a few expression actions to smart pointers.
These actions are extremely widely used (identifier expressions and literals); still no performance regression.
Chris Lattner [Sun, 18 Jan 2009 09:39:41 +0000 (09:39 +0000)]
Switch MinimalAction from new/delete'ing its TypeNameInfo to
allocating them from a recycling bump pointer allocator. This
reduces malloc/free traffic of parse-noop (but no other mode),
which makes sharking -parse-noop more meaningful.
Chris Lattner [Sun, 18 Jan 2009 02:57:21 +0000 (02:57 +0000)]
rearrange GetIdentifierInfo so that the fast path can be partially inlined into PTHLexer::Lex. This speeds up the user time of PTH -Eonly by another 2ms (4.4%)
Chris Lattner [Sun, 18 Jan 2009 02:52:26 +0000 (02:52 +0000)]
Avoid malloc thrashing on the std::vector for ConditionalStack. Because
there is one of these per header, this almost always gets alloc/free'd for
each #ifdef.
Chris Lattner [Sun, 18 Jan 2009 02:19:16 +0000 (02:19 +0000)]
On i386 and x86-64, just do unaligned loads
instead of assembling from bytes. This speeds up -Eonly PTH reading
of cocoa.h by about 2ms, which is 4.2%.
projects / clang / log