Benjamin Kramer [Sun, 3 Feb 2013 19:59:25 +0000 (19:59 +0000)]
CodeGen: Implement hint values for dynamic_cast as described in the Itanium C++ ABI.
This can yield dramatic speedups of dynamic_cast for simple inheritance trees,
at least with libsupc++. Neither libcxxabi nor libcxxrt make use of this
hint currently, it was never implemented because clang didn't support it.
There was some concern about the number of class hierarchy walks this change
introduces. If it turns out to be an issue we can add caching either at the cast
pair level or even deeper, but we also do a lot of walks in Sema so this
codepath is probably fairly optimized already.
NAKAMURA Takumi [Sun, 3 Feb 2013 12:52:54 +0000 (12:52 +0000)]
clang/test/CodeGenCXX/debug-info-class.cpp: Tweak to unbreak test for a few targets.
- Relax a expression for arm-gnueabi.
- Exclude msvc to limit target triplets to add limited a few targets.
Feel free to remove actions if guys thought they redundant.
Jordan Rose [Sat, 2 Feb 2013 05:15:53 +0000 (05:15 +0000)]
Revert "[analyzer] Model trivial copy/move ctors with an aggregate bind."
...again. The problem has not been fixed and our internal buildbot is still
getting hangs.
This reverts r174212, originally applied in r173951, then reverted in r174069.
Will not re-apply until the entire project analyzes successfully on my
local machine.
Richard Smith [Sat, 2 Feb 2013 02:14:45 +0000 (02:14 +0000)]
PR15132: Replace "address expression must be an lvalue or a function
designator" diagnostic with more correct and more human-friendly "cannot take
address of rvalue of type 'T'".
For the case of & &T::f, provide a custom diagnostic, rather than unhelpfully
saying "cannot take address of rvalue of type '<overloaded function type>'".
For the case of &array_temporary, treat it just like a class temporary
(including allowing it as an extension); the existing diagnostic wording
for the class temporary case works fine.
David Blaikie [Sat, 2 Feb 2013 00:34:26 +0000 (00:34 +0000)]
Generalize DebugInfo tests by avoiding explicit metadata numbers
This addresses several (not all) debug info tests that use explicit metadata
numbers. Wherever the same number appeared more than once in a test I used
a named match to ensure the same number appeared in all those cases (this may
still be overly constraining test cases as they may not have actually cared
about that relationship). For one-off numbers I just replaced them with an
unnamed regex.
This may underconstrain poorly written test cases that were interested in
checking that certain metadata nodes were related but didn't actually match
on all the related nodes numbers.
Anna Zaks [Sat, 2 Feb 2013 00:30:04 +0000 (00:30 +0000)]
[analyzer] Always inline functions with bodies generated by BodyFarm.
Inlining these functions is essential for correctness. We often have
cases where we do not inline calls. For example, the shallow mode and
when reanalyzing previously inlined ObjC methods as top level.
Nick Lewycky [Sat, 2 Feb 2013 00:25:55 +0000 (00:25 +0000)]
This patch makes "&Cls::purevfn" not an odr use. This isn't what the standard
says, but that's a defect (to be filed). "Cls::purevfn()" is still an odr use.
Also fixes a bug that caused us to not mark the function referenced just
because we didn't want to mark it odr used.
Douglas Gregor [Fri, 1 Feb 2013 23:45:03 +0000 (23:45 +0000)]
Merge "special" types from different modules in the AST reader.
Different modules may have different views of the various "special"
types in the AST, such as the redefinition type for "id". Merge those
types rather than only considering the redefinition types for the
first AST file loaded.
Jordan Rose [Fri, 1 Feb 2013 19:49:57 +0000 (19:49 +0000)]
[analyzer] Reuse a LazyCompoundVal if its type matches the new region.
This allows us to keep from chaining LazyCompoundVals in cases like this:
CGRect r = CGRectMake(0, 0, 640, 480);
CGRect r2 = r;
CGRect r3 = r2;
Previously we only made this optimization if the struct did not begin with
an aggregate member, to make sure that we weren't picking up an LCV for
the first field of the struct. But since LazyCompoundVals are typed, we can
make that inference directly by comparing types.
This is a pure optimization; the test changes are to guard against possible
future regressions.
David Blaikie [Fri, 1 Feb 2013 19:09:49 +0000 (19:09 +0000)]
Fix exception handling line table problems introduced by r173593
r173593 made us a little too eager to associate all code at the end of a
function with the user-written 'return' line. This caused problems with
breakpoints as they'd be set in exception handling code preceeding the
actual non-exception return handling code, leading to the breakpoint never
being hit in non-exceptional execution.
This change restores the pre-r173593 exception handling line information where
the cleanup code is associated with the '}' not the return line.
[modules] Introduce ModuleFile::DirectImportLoc which is the source location
where the module was explicitly or implicitly imported in the local translation unit.
Bill Schmidt [Fri, 1 Feb 2013 15:34:29 +0000 (15:34 +0000)]
Fix PR14881 by implementing conversion rules between int and complex int.
Prior to the patch, Clang does not properly promote types when a complex
integer operand is combined with an integer via a binary operator, or when
one is assigned to the other in either order. This patch detects when
promotion is needed (and permissible) and generates the necessary code.
The test assmes no target has the same size operands for "char" and
"long long," and that no target performs arithmetic on char operands without
extending them to a larger format first. If there are any targets for
which this is not the case, they should be XFAILed.
Manuel Klimek [Fri, 1 Feb 2013 13:41:35 +0000 (13:41 +0000)]
Re-design the convenience interfaces on MatchFinder.
First, this implements a match() method on MatchFinder; this allows us
to get rid of the findAll implementation, as findAll is really a special
case of recursive matchers on match.
Instead of findAll, provide a convenience function match() that lets
users iterate easily over the results instead of needing to implement
callbacks.
Daniel Jasper [Fri, 1 Feb 2013 11:00:45 +0000 (11:00 +0000)]
Revamp of the basic layouting algorithm in clang-format.
In order to end up with good solutions, clang-format needs to try
"all" combinations of line breaks, evaluate them and select the
best one. Before, we have done this using a DFS with memoization
and cut-off conditions. However, this approach is very limited
as shown by the huge static initializer in the attachment of
llvm.org/PR14959.
Instead, this new implementation uses a variant of Dijkstra's
algorithm to do a prioritized BFS over the solution space.
Some numbers:
lib/Format/TokenAnnotator.cpp: 1.5s -> 0.15s
Attachment of PR14959: 10min+ (didn't finish) -> 10s
Bill Schmidt [Fri, 1 Feb 2013 02:14:03 +0000 (02:14 +0000)]
Enable -fno-altivec, -mno-altivec for PowerPC.
Introduces these negation forms explicitly and uses them to control a new
"altivec" target feature for PowerPC. This allows avoiding generating
Altivec instructions on processors that support Altivec.
The new test case verifies that the Altivec "lvx" instruction is not
used when -fno-altivec is present on the command line.
Michael Han [Fri, 1 Feb 2013 01:19:17 +0000 (01:19 +0000)]
[Sema][Attr]Fix alignment attribute printing.
Remove "IsMSDeclspec" argument from Align attribute since the arguments in Attr.td should
only model those appear in source code. Introduce attribute Accessor, and teach TableGen
to generate syntax kind accessors for Align attribute, and use those accessors to decide
if an alignment attribute is a declspec attribute.
[Comment parsing] Add support for recognizing
\headerfile command and representing it in an xml
document. Patch reviewed by Dmitri Gribenko.
// rdar://12397511
Jordan Rose [Thu, 31 Jan 2013 22:06:02 +0000 (22:06 +0000)]
[analyzer] Don't track autorelease pools created by +new.
This matches our behavior for autorelease pools created by +alloc. Some
people like to create autorelease pools in one method and release them
somewhere else.
If you want safe autorelease pool semantics, use the new ARC-compatible
syntax: @autoreleasepool { ... }
Daniel Jasper [Thu, 31 Jan 2013 14:59:26 +0000 (14:59 +0000)]
Several improvements to the formatting of static initializers.
1. Never avoid bin packing in static initializers as this can
lead to terrible results.
2. If an element has to be broken over multiple lines, break after
the following comma.
This should be a step forward, but there are still many cases
especially with nested static initializers that we handle badly.
More patches will follow.
Tim Northover [Thu, 31 Jan 2013 12:13:10 +0000 (12:13 +0000)]
Add support for AArch64 target.
In cooperation with the LLVM patch, this should implement all scalar front-end
parts of the C and C++ ABIs for AArch64.
This patch excludes the NEON support also reviewed due to an outbreak of
batshit insanity in our legal department. That will be committed soon bringing
the changes to precisely what has been approved.
Evgeniy Stepanov [Thu, 31 Jan 2013 09:53:29 +0000 (09:53 +0000)]
[msan] Run more optimizations after MemorySanitizer pass.
MSan instrumentation is driven by the original code. We take every
incoming instruction and emit another instruction (or ten) next to
it, operating on the shadow values (but sometimes on the real values,
too). Two programs in one, essentially. There can be any kinds of
redundancies in the second one, so we just run whatever is normally
run at -O2, and then exclude some passes that do not help much with
benchmarks.
Richard Smith [Thu, 31 Jan 2013 06:13:11 +0000 (06:13 +0000)]
Remove incorrect warning signs from QualType::getUnqualifiedType(). These
statements were only true for 65 CLs after they were introduced two years ago.
Douglas Gregor [Thu, 31 Jan 2013 05:50:40 +0000 (05:50 +0000)]
When we're emitting a constructor or destructor call from a delegating
constructor, retrieve our VTT parameter directly. Fixes PR14588 /
<rdar://problem/12867962>.