Chandler Carruth [Mon, 24 Sep 2012 00:34:20 +0000 (00:34 +0000)]
Address one of the original FIXMEs for the new SROA pass by implementing
integer promotion analogous to vector promotion. When there is an
integer alloca being accessed both as its integer type and as a narrower
integer type, promote the narrower access to "insert" and "extract" the
smaller integer from the larger one, and make the integer alloca
a candidate for promotion.
In the new formulation, we don't care about target legal integer or use
thresholds to control things. Instead, we only perform this promotion to
an integer type which the frontend has already emitted a load or store
for. This bounds the scope and prevents optimization passes from
coalescing larger and larger entities into a single integer.
llvm-svn: 164479
NAKAMURA Takumi [Mon, 24 Sep 2012 00:34:18 +0000 (00:34 +0000)]
unittests/AST/StmtPrinterTest.cpp: Suppress a LP64-assumed test, "0x100000000i128 => 4294967296L", for now.
LONG_MAX is 2147483647L on common 32 bit and LLP64 (Windows x64).
llvm-svn: 164478
Dmitri Gribenko [Sun, 23 Sep 2012 20:29:07 +0000 (20:29 +0000)]
As a followup for r164303, add some tests for printing literals that test
printing directly rather than through a complicated machinery of ObjC rewriter.
llvm-svn: 164477
Anton Korobeynikov [Sun, 23 Sep 2012 15:53:47 +0000 (15:53 +0000)]
Emit dtors into proper section while compiling in vcpp-compatible mode.
Patch by Kai!
llvm-svn: 164476
Chandler Carruth [Sun, 23 Sep 2012 11:43:14 +0000 (11:43 +0000)]
Switch to a signed representation for the dynamic offsets while walking
across the uses of the alloca. It's entirely possible for negative
numbers to come up here, and in some rare cases simply doing the 2's
complement arithmetic isn't the correct decision. Notably, we can't zext
the index of the GEP. The definition of GEP is that these offsets are
sign extended or truncated to the size of the pointer, and then wrapping
2's complement arithmetic used.
This patch fixes an issue that comes up with *no* input from the
buildbots or bootstrap afaict. The only place where it manifested,
disturbingly, is Clang's own regression test suite. A reduced and
targeted collection of tests are added to cope with this. Note that I've
tried to pin down the potential cases of overflow, but may have missed
some cases. I've tried to add a few cases to test this, but its hard
because LLVM has quite limited support for >64bit constructs.
llvm-svn: 164475
Nick Lewycky [Sun, 23 Sep 2012 03:58:21 +0000 (03:58 +0000)]
Don't do actual work inside an assert statement. Fixes PR11760!
llvm-svn: 164474
Rafael Espindola [Sun, 23 Sep 2012 03:05:41 +0000 (03:05 +0000)]
Implement ToolChain::IsUnwindTablesDefault to reduce code duplication a bit.
llvm-svn: 164473
Michael Liao [Sun, 23 Sep 2012 02:41:47 +0000 (02:41 +0000)]
Revise test to avoid using of 'grep'
llvm-svn: 164472
Craig Topper [Sun, 23 Sep 2012 02:12:10 +0000 (02:12 +0000)]
Add LLVM_OVERRIDE to methods that override their base classes.
llvm-svn: 164471
Rafael Espindola [Sat, 22 Sep 2012 22:58:57 +0000 (22:58 +0000)]
On Hexagon getArchName() never returns x86_64, simplify the function.
llvm-svn: 164470
Rafael Espindola [Sat, 22 Sep 2012 22:51:00 +0000 (22:51 +0000)]
Use pipes.
llvm-svn: 164469
Rafael Espindola [Sat, 22 Sep 2012 22:30:04 +0000 (22:30 +0000)]
Use FileCheck.
llvm-svn: 164468
Dmitri Gribenko [Sat, 22 Sep 2012 21:47:50 +0000 (21:47 +0000)]
Comment sema: warn when comment has \deprecated but declaration does not have a
deprecation attribute ('deprecated', 'availability' or 'unavailable').
This warning is under a separate flag, -Wdocumentation-deprecated-sync, so it
can be turned off easily while leaving other -Wdocumentation warnings on.
llvm-svn: 164467
Dmitri Gribenko [Sat, 22 Sep 2012 21:44:37 +0000 (21:44 +0000)]
Remove superfluous \namespace command.
llvm-svn: 164466
Michael Liao [Sat, 22 Sep 2012 21:07:59 +0000 (21:07 +0000)]
Enhance test case of atomic16 to verify inst encoding fixed in r164453.
llvm-svn: 164465
Dmitri Gribenko [Sat, 22 Sep 2012 17:52:29 +0000 (17:52 +0000)]
Fix cindex.py compatibility with older libclang.so
The issue is that we were calling clang_getCompletionBriefComment
unconditionally. New we check if this function is available before calling it.
llvm-svn: 164464
Rafael Espindola [Sat, 22 Sep 2012 15:26:38 +0000 (15:26 +0000)]
Use FileCheck.
llvm-svn: 164462
Rafael Espindola [Sat, 22 Sep 2012 15:04:11 +0000 (15:04 +0000)]
Fix pr13749. There is still a lot of code using getArchName that should be
using getArch, but I will try to fix them one at time to add tests.
llvm-svn: 164460
NAKAMURA Takumi [Sat, 22 Sep 2012 13:12:28 +0000 (13:12 +0000)]
ARMInstPrinter.cpp: Fix a warning in -Asserts. [-Wunused-variable]
llvm-svn: 164459
NAKAMURA Takumi [Sat, 22 Sep 2012 13:12:22 +0000 (13:12 +0000)]
Whitespace.
llvm-svn: 164458
Tim Northover [Sat, 22 Sep 2012 11:18:19 +0000 (11:18 +0000)]
Fix edge cases of ARM shift operands in arith instructions.
As before with load instructions, oddities like "asr #32", "rrx" could
be printed incorrectly.
Patch by Chris Lidbury.
llvm-svn: 164456
Tim Northover [Sat, 22 Sep 2012 11:18:12 +0000 (11:18 +0000)]
Fix the handling of edge cases in ARM shifted operands.
This patch fixes load/store instructions to handle less common cases
like "asr #32", "rrx" properly throughout the MC layer.
Patch by Chris Lidbury.
llvm-svn: 164455
Abramo Bagnara [Sat, 22 Sep 2012 09:05:22 +0000 (09:05 +0000)]
Avoid multiple atomic builtin declaration.
llvm-svn: 164454
Michael Liao [Sat, 22 Sep 2012 05:41:15 +0000 (05:41 +0000)]
Fix 16-bit atomic inst encoding and keep pseudo-inst starting with '#'
llvm-svn: 164453
Michael Liao [Sat, 22 Sep 2012 03:39:42 +0000 (03:39 +0000)]
Fix typo in r164357
llvm-svn: 164452
Andrew Trick [Sat, 22 Sep 2012 02:24:21 +0000 (02:24 +0000)]
Machine Model (-schedmodel only). Added SchedAliases.
Allow subtargets to tie SchedReadWrite types to processor specific
sequences or variants.
llvm-svn: 164451
Argyrios Kyrtzidis [Sat, 22 Sep 2012 01:58:06 +0000 (01:58 +0000)]
When importing a FunctionProtoType::ExtProtoInfo, its ExceptionSpecDecl can point to the
FunctionDecl that we are importing the FunctionProtoType for, in which case we'll have
infinite recursion when importing.
Initially create a FunctionProtoType with null ExceptionSpecDecl/ExceptionSpecTemplate and
update the type in ASTNodeImporter::VisitFunctionDecl after the FunctionDecl has been created.
llvm-svn: 164450
Jordan Rose [Sat, 22 Sep 2012 01:25:06 +0000 (01:25 +0000)]
[analyzer] Suppress bugs whose paths go through the return of a null pointer.
This is a heuristic intended to greatly reduce the number of false
positives resulting from inlining, particularly inlining of generic,
defensive C++ methods that live in header files. The suppression is
triggered in the cases where we ask to track where a null pointer came
from, and it turns out that the source of the null pointer was an inlined
function call.
This change brings the number of bug reports in LLVM from ~1500 down to
around ~300, a much more manageable number. Yes, some true positives may
be hidden as well, but from what I looked at the vast majority of silenced
reports are false positives, and many of the true issues found by the
analyzer are still reported.
I'm hoping to improve this heuristic further by adding some exceptions
next week (cases in which a bug should still be reported).
llvm-svn: 164449
Jordan Rose [Sat, 22 Sep 2012 01:25:00 +0000 (01:25 +0000)]
[analyzer] Track a null value back through FindLastStoreBRVisitor.
Also, tidy up the other tracking visitors so that they mark the right
things as interesting and don't do extra work.
llvm-svn: 164448
Jordan Rose [Sat, 22 Sep 2012 01:24:56 +0000 (01:24 +0000)]
[analyzer] Always allow BugReporterVisitors to see the bug path.
Before, PathDiagnosticConsumers that did not support actual path output
would (sensibly) cause the generation of the full path to be skipped.
However, BugReporterVisitors may want to see the path in order to mark a
BugReport as invalid.
Now, even for a path generation scheme of 'None' we will still create a
trimmed graph and walk backwards through the bug path, doing no work other
than passing the nodes to the BugReporterVisitors. This isn't cheap, but
it's necessary to properly do suppression when the first path consumer does
not support path notes.
In the future, we should try only generating the path and visitor-provided
path notes once, or at least only creating the trimmed graph once.
llvm-svn: 164447
Jordan Rose [Sat, 22 Sep 2012 01:24:53 +0000 (01:24 +0000)]
[analyzer] Allow a BugReport to be marked "invalid" during path generation.
This is intended to allow visitors to make decisions about whether a
BugReport is likely a false positive. Currently there are no visitors
making use of this feature, so there are no tests.
When a BugReport is marked invalid, the invalidator must provide a key
that identifies the invaliation (intended to be the visitor type and a
context pointer of some kind). This allows us to reverse the decision
later on. Being able to reverse a decision about invalidation gives us more
flexibility, and allows us to formulate conditions like "this report is
invalid UNLESS the original argument is 'foo'". We can use this to
fine-tune our false-positive suppression (coming soon).
llvm-svn: 164446
Jordan Rose [Sat, 22 Sep 2012 01:24:49 +0000 (01:24 +0000)]
[analyzer] Look through OpaqueValueExprs when tracking a nil value.
This allows us to show /why/ a particular object is nil, even when it is
wrapped in an OpaqueValueExpr.
llvm-svn: 164445
Jordan Rose [Sat, 22 Sep 2012 01:24:46 +0000 (01:24 +0000)]
[analyzer] Better path notes for null pointers passed as arguments.
Rather than saying "Null pointer value stored to 'foo'", we now say
"Passing null pointer value via Nth parameter 'foo'", which is much better.
The note is also now on the argument expression as well, rather than the
entire call.
This paves the way for continuing to track arguments back to their sources.
<rdar://problem/
12211490>
llvm-svn: 164444
Jordan Rose [Sat, 22 Sep 2012 01:24:42 +0000 (01:24 +0000)]
Use llvm::getOrdinalSuffix to print ordinal numbers in diagnostics.
Just a refactoring of common infrastructure. No intended functionality change.
llvm-svn: 164443
Jordan Rose [Sat, 22 Sep 2012 01:24:38 +0000 (01:24 +0000)]
[analyzer] Check that an ObjCIvarRefExpr's base is non-null even as an lvalue.
Like with struct fields, we want to catch cases like this early,
so that we can produce better diagnostics and path notes:
PointObj *p = nil;
int *px = &p->_x; // should warn here
*px = 1;
llvm-svn: 164442
Jordan Rose [Sat, 22 Sep 2012 01:24:33 +0000 (01:24 +0000)]
[analyzer] Check that a member expr is valid even when the result is an lvalue.
We want to catch cases like this early, so that we can produce better
diagnostics and path notes:
Point *p = 0;
int *px = &p->x; // should warn here
*px = 1;
llvm-svn: 164441
Jordan Rose [Sat, 22 Sep 2012 01:24:21 +0000 (01:24 +0000)]
Add llvm::getOrdinalSuffix to get the appropriate -st, -nd, -rd, -th suffix.
Used by clang to print parameter indexes.
llvm-svn: 164440
Jordan Rose [Sat, 22 Sep 2012 01:24:18 +0000 (01:24 +0000)]
Casting: assert that pointer arguments to isa<> are non-null.
This silences several analyzer warnings within LLVM, and provides a slightly
nicer crash experience when someone calls isa<>, cast<>, or dyn_cast<> with
a null pointer.
llvm-svn: 164439
Jordan Rose [Sat, 22 Sep 2012 01:24:16 +0000 (01:24 +0000)]
DenseMap: assert that we have found a bucket before we try to insert into it.
This silences literally dozens of analyzer warnings on LLVM (since DenseMap
is such a commonly-used class).
llvm-svn: 164438
Richard Smith [Sat, 22 Sep 2012 00:53:56 +0000 (00:53 +0000)]
Fix bug which sometimes resulted in further diagnostics being produced after a
fatal error. Previously, if a fatal error was followed by a diagnostic which
was suppressed due to a SFINAETrap, we'd forget that we'd seen a fatal error.
llvm-svn: 164437
Jim Ingham [Sat, 22 Sep 2012 00:14:47 +0000 (00:14 +0000)]
Missed a few places where I didn't delete the obsolete (commented out) version of the breakpoint creation.
llvm-svn: 164436
Akira Hatanaka [Sat, 22 Sep 2012 00:07:12 +0000 (00:07 +0000)]
MIPS DSP: Add immediate leaves.
llvm-svn: 164435
Akira Hatanaka [Sat, 22 Sep 2012 00:06:06 +0000 (00:06 +0000)]
MIPS DSP: Add predicates and instruction template.
llvm-svn: 164434
Jim Ingham [Sat, 22 Sep 2012 00:05:11 +0000 (00:05 +0000)]
Fix all the test case breakages caused by folks writing tests all over the place that depended explicitly
on the output of "break set". Please don't do this sort of thing!!!!!
llvm-svn: 164433
Jim Ingham [Sat, 22 Sep 2012 00:04:04 +0000 (00:04 +0000)]
Change the new breakpoint creation output (primarily from "break set") to something more useful.
<rdar://problem/
11333623>
llvm-svn: 164432
Akira Hatanaka [Fri, 21 Sep 2012 23:58:31 +0000 (23:58 +0000)]
Add MIPS DSP register classes. Set actions of DSP vector operations and override
TargetLowering's callback functions.
llvm-svn: 164431
Akira Hatanaka [Fri, 21 Sep 2012 23:52:47 +0000 (23:52 +0000)]
SelectionDAG node enums for MIPS DSP nodes.
llvm-svn: 164430
Akira Hatanaka [Fri, 21 Sep 2012 23:48:37 +0000 (23:48 +0000)]
Add MIPS accumulator and DSP control registers.
llvm-svn: 164429
Akira Hatanaka [Fri, 21 Sep 2012 23:41:49 +0000 (23:41 +0000)]
Add flags and feature bits for mips dsp.
llvm-svn: 164428
Chandler Carruth [Fri, 21 Sep 2012 23:36:40 +0000 (23:36 +0000)]
Fix a case where the new SROA pass failed to zap dead operands to
selects with a constant condition. This resulted in the operands
remaining live through the SROA rewriter. Most of the time, this just
caused some dead allocas to persist and get zapped by later passes, but
in one case found by Joerg, it caused a crash when we tried to *promote*
the alloca despite it having this dead use. We already have the
mechanisms in place to handle this, just wire select up to them.
llvm-svn: 164427
Eric Christopher [Fri, 21 Sep 2012 23:03:29 +0000 (23:03 +0000)]
Add an --enable-backtraces option to configure to determine
whether or not we want to print out backtrace information. Useful
for libraries that don't need backtrace information on a crash.
rdar://
11844710
llvm-svn: 164426
Dan Gohman [Fri, 21 Sep 2012 23:00:37 +0000 (23:00 +0000)]
Add an MDBuilder utility for creating !tbaa.struct nodes.
llvm-svn: 164425
Michael J. Spencer [Fri, 21 Sep 2012 22:49:06 +0000 (22:49 +0000)]
[CMake] Respect LLVM_BUILD_RUNTIME.
llvm-svn: 164424
Michael J. Spencer [Fri, 21 Sep 2012 22:39:39 +0000 (22:39 +0000)]
[CMake][tsan] Compile .S files for tsan as C instead of CXX.
Using CXX breaks if you add -stdlib=libc++ to CMAKE_CXX_FLAGS.
llvm-svn: 164422
Chad Rosier [Fri, 21 Sep 2012 22:22:39 +0000 (22:22 +0000)]
[ms-inline asm] Use mnemonicIsValid() in a helpful assert.
llvm-svn: 164421
Chad Rosier [Fri, 21 Sep 2012 22:21:26 +0000 (22:21 +0000)]
[ms-inline asm] Expose the mnemonicIsValid() function in the AsmParser.
llvm-svn: 164420
Eric Christopher [Fri, 21 Sep 2012 22:18:52 +0000 (22:18 +0000)]
Have the DbgVariable "isArtificial" and "isObjectPointer" not
care about it being an argument variable so that we can decide
that captured block and lambda vars that don't happen to
be arguments could be an argument pointer.
Add the object pointer for one case onto the subprogram die.
rdar://
12001329
llvm-svn: 164419
Eric Christopher [Fri, 21 Sep 2012 22:18:42 +0000 (22:18 +0000)]
Mark the passed in "self" for a block as an object pointer so
that the backend can mark it as the representative pointer for
the block.
rdar://
12001329
llvm-svn: 164418
Argyrios Kyrtzidis [Fri, 21 Sep 2012 22:17:13 +0000 (22:17 +0000)]
Fix build errors.
llvm-svn: 164417
Argyrios Kyrtzidis [Fri, 21 Sep 2012 22:04:02 +0000 (22:04 +0000)]
[AST importer] Do a correct import of a FunctionProtoType::ExtProtoInfo, don't pass
nodes from the imported ASTContext.
rdar://
12348924
llvm-svn: 164416
Chad Rosier [Fri, 21 Sep 2012 21:08:46 +0000 (21:08 +0000)]
Add comment.
llvm-svn: 164415
Chad Rosier [Fri, 21 Sep 2012 20:51:43 +0000 (20:51 +0000)]
Add comment.
llvm-svn: 164414
Bob Wilson [Fri, 21 Sep 2012 20:49:54 +0000 (20:49 +0000)]
Fix installheaders target to do what it did prior to r161760. rdar://
12348765
llvm-svn: 164413
Fariborz Jahanian [Fri, 21 Sep 2012 20:46:37 +0000 (20:46 +0000)]
objective-C: when diagnosing deprecated/unavailable usage of
setter or getter backing a deprecated/unavailable property,
also not location of the property. // rdar://
12324295
llvm-svn: 164412
Rafael Espindola [Fri, 21 Sep 2012 20:39:32 +0000 (20:39 +0000)]
Call CGM.SetLLVMFunctionAttributesForDefinition on thunks so that they get
attributes like uwtable. Without uwtable a stack unwinder would be unable
to go past the thunks.
llvm-svn: 164411
Simon Atanasyan [Fri, 21 Sep 2012 20:19:32 +0000 (20:19 +0000)]
Add four new command line options for MIPS CPU selection:
-mips32, -mips32r2, -mips64, -mips64r2.
The patch reviewed by Eric Christopher.
llvm-svn: 164410
Douglas Gregor [Fri, 21 Sep 2012 20:16:09 +0000 (20:16 +0000)]
Fix a typo in the PCH internals documentation
llvm-svn: 164409
Evan Cheng [Fri, 21 Sep 2012 20:04:28 +0000 (20:04 +0000)]
Fix a significant recent(?) regression. StackSlotColoring no longer did anything
because LiveStackAnalysis was not preserved by VirtRegWriter. This caused
big stack usage regression in some cases.
rdar://
12340383
llvm-svn: 164408
Sean Silva [Fri, 21 Sep 2012 19:48:16 +0000 (19:48 +0000)]
docs: Fix long-standing typo in yaml2obj.rst.
llvm-svn: 164407
Chad Rosier [Fri, 21 Sep 2012 19:25:59 +0000 (19:25 +0000)]
Whitespace.
llvm-svn: 164406
Howard Hinnant [Fri, 21 Sep 2012 19:21:37 +0000 (19:21 +0000)]
Rename class __lambda to __lambda_node to avoid clash with gcc. This fixes llvm.org/bugs/show_bug.cgi?id=13889
llvm-svn: 164405
Howard Hinnant [Fri, 21 Sep 2012 19:12:56 +0000 (19:12 +0000)]
Tweak use of dlopen to be a little more correct and higher performing.
llvm-svn: 164404
Enrico Granata [Fri, 21 Sep 2012 19:10:53 +0000 (19:10 +0000)]
Initial commit of a new testsuite feature: test categories.
This feature allows us to group test cases into logical groups (categories), and to only run a subset of test cases based on these categories.
Each test-case can have a new method getCategories(self): which returns a list of strings that are the categories to which the test case belongs.
If a test-case does not provide its own categories, we will look for categories in the class that contains the test case.
If that fails too, the default implementation looks for a .category file, which contains a comma separated list of strings.
The test suite will recurse look for .categories up until the top level directory (which we guarantee will have an empty .category file).
The driver dotest.py has a new --category <foo> option, which can be repeated, and specifies which categories of tests you want to run.
(example: ./dotest.py --category objc --category expression)
All tests that do not belong to any specified category will be skipped. Other filtering options still exist and should not interfere with category filtering.
A few tests have been categorized. Feel free to categorize others, and to suggest new categories that we could want to use.
All categories need to be validly defined in dotest.py, or the test suite will refuse to run when you use them as arguments to --category.
In the end, failures will be reported on a per-category basis, as well as in the usual format.
This is the very first stage of this feature. Feel free to chime in with ideas for improvements!
llvm-svn: 164403
Dan Gohman [Fri, 21 Sep 2012 18:41:30 +0000 (18:41 +0000)]
Say "bytes" instead of "address units", since that's what the
rest of LangRef uses.
llvm-svn: 164402
Ted Kremenek [Fri, 21 Sep 2012 18:33:56 +0000 (18:33 +0000)]
Experiment in BodyFarm of structuring AST creation calls in a hierarchy,
so that they visually look like an AST dump.
llvm-svn: 164401
Ted Kremenek [Fri, 21 Sep 2012 18:33:54 +0000 (18:33 +0000)]
Create helper method in BodyFarm for creating simple assignments.
llvm-svn: 164400
Ted Kremenek [Fri, 21 Sep 2012 18:33:52 +0000 (18:33 +0000)]
Add helper method in BodyFarm to create unary dereferences.
llvm-svn: 164399
Dan Gohman [Fri, 21 Sep 2012 18:21:48 +0000 (18:21 +0000)]
Document the new !tbaa.struct metadata.
llvm-svn: 164398
Ted Kremenek [Fri, 21 Sep 2012 18:13:27 +0000 (18:13 +0000)]
Add helper method to BodyFarm for creating lvalue-to-rvalue conversions.
llvm-svn: 164397
Ted Kremenek [Fri, 21 Sep 2012 18:13:23 +0000 (18:13 +0000)]
Add helper method to BodyFarm for creatinging integral casts.
llvm-svn: 164396
David Blaikie [Fri, 21 Sep 2012 18:03:02 +0000 (18:03 +0000)]
Add missing 'to' and rephrase an explanation of GCC's assumptions.
Wordsmithing by Matt Beaumont-Gay in response to r164389.
llvm-svn: 164395
DeLesley Hutchins [Fri, 21 Sep 2012 17:57:00 +0000 (17:57 +0000)]
Thread-safety analysis: better handling of unreachable blocks. Fixes a bug
where a call to function marked 'noreturn' is followed by unreachable
implicit destructor calls.
llvm-svn: 164394
Ted Kremenek [Fri, 21 Sep 2012 17:55:34 +0000 (17:55 +0000)]
Re-enable faux-bodies by default.
Try this again, now that r164392 is in place.
llvm-svn: 164393
Ted Kremenek [Fri, 21 Sep 2012 17:54:35 +0000 (17:54 +0000)]
Use helper method to create DeclRefExprs in BodyFarm, hopefully allevating
them being correctly constructed.
llvm-svn: 164392
Ted Kremenek [Fri, 21 Sep 2012 17:54:32 +0000 (17:54 +0000)]
Add some structuring comments. No functionality change.
llvm-svn: 164391
Chad Rosier [Fri, 21 Sep 2012 17:50:09 +0000 (17:50 +0000)]
Update comment.
llvm-svn: 164390
David Blaikie [Fri, 21 Sep 2012 17:47:36 +0000 (17:47 +0000)]
Document "do not use defaults in covered switch-over-enum" coding standard.
llvm-svn: 164389
Benjamin Kramer [Fri, 21 Sep 2012 17:27:23 +0000 (17:27 +0000)]
LoopIdiom: Give up when the loop is not in canonical form.
We rely on it when doing the transforms. This can happen when there is an
indirectbr in the loop.
Fixes PR13892.
llvm-svn: 164383
Rafael Espindola [Fri, 21 Sep 2012 17:15:24 +0000 (17:15 +0000)]
Use MapVectors to make the order we mark virtual functions used by a vtable
deterministic. Fixes pr13868.
llvm-svn: 164382
Chad Rosier [Fri, 21 Sep 2012 16:58:35 +0000 (16:58 +0000)]
[fast-isel] Fallback to SelectionDAG isel if we require strict alignment for
non-aligned i32 loads/stores.
rdar://
12304911
llvm-svn: 164381
Greg Clayton [Fri, 21 Sep 2012 16:31:20 +0000 (16:31 +0000)]
Full KDP process control with per CPU resume and step.
llvm-svn: 164378
Benjamin Kramer [Fri, 21 Sep 2012 16:26:41 +0000 (16:26 +0000)]
InstCombine: Make sure we use the pre-zext type when creating a constant of a value that is zext'd.
Fixes PR13250.
llvm-svn: 164377
Tobias Grosser [Fri, 21 Sep 2012 16:24:13 +0000 (16:24 +0000)]
Bailout if libpluto finds no schedule
Older versions of libpluto crashed, if no schedule was found. Recent
versions return NULL. We detect this and keep the original schedule.
llvm-svn: 164376
Bill Wendling [Fri, 21 Sep 2012 16:07:49 +0000 (16:07 +0000)]
Update for encapsulating the "construct*AlignmentFromInt" methods.
llvm-svn: 164374
Bill Wendling [Fri, 21 Sep 2012 16:07:28 +0000 (16:07 +0000)]
Encapsulate the "construct*AlignmentFromInt" functions.
llvm-svn: 164373
Michael Liao [Fri, 21 Sep 2012 16:03:03 +0000 (16:03 +0000)]
Fix a typo in r164357
llvm-svn: 164372
Dmitri Gribenko [Fri, 21 Sep 2012 15:26:34 +0000 (15:26 +0000)]
Clarify comment.
llvm-svn: 164371
Bill Wendling [Fri, 21 Sep 2012 15:26:31 +0000 (15:26 +0000)]
Make the 'get*AlignmentFromAttr' functions into member functions within the Attributes class. Now with fix.
llvm-svn: 164370
Benjamin Kramer [Fri, 21 Sep 2012 14:34:31 +0000 (14:34 +0000)]
BitcodeReader: Correctly insert blockaddress constant referring to a already parsed function.
We inserted a placeholder that was never replaced because the function was
already visited. Assert that all placeholders have been resolved when tearing
down the bitcode reader.
Fixes PR13895.
llvm-svn: 164369
NAKAMURA Takumi [Fri, 21 Sep 2012 12:00:42 +0000 (12:00 +0000)]
Revert r164364, "Flip "faux-bodies" in the analyzer on by default to flush out bugs."
It crashed test/Analysis/Output/blocks.m on some hosts.
llvm-svn: 164368