This is a re-commit of r190764, with an extra check to make sure that we're not
performing the transformation on illegal types (a small test case has been
added for this as well).
Original commit message:
The PPC backend uses a target-specific DAG combine to turn unaligned Altivec
loads into a permutation-based sequence when possible. Unfortunately, the
target-specific DAG combine is not always called on all loads of interest
(sometimes the routines in DAGCombine call CombineTo such that the new node and
users are not added to the worklist); allowing the combine to trigger early
(before type legalization) mitigates this problem. Because the autovectorizers
only create legal vector types, I don't expect a lot of cases where this
optimization is enabled by type legalization in practice.
llvm-svn: 190771
Summary:
The '?' flag uses the last section group if the last had a section
group. We treat combining an explicit section group and the '?' as a
hard error.
This fixes PR17198.
Reviewers: rafael, bkramer
Reviewed By: bkramer
CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1686
llvm-svn: 190768
For alignment purposes, the instruction array will always have an even
number of entries, with the final entry potentially unused (in which
case the array will be one longer than indicated by the count of unwind
codes field).
Reviewed by Anton Korobeynikov, Charles Davis and Nico Rieck.
llvm-svn: 190767
data structures.
The Win64 EH data structures must be of type IMAGE_REL_AMD64_ADDR32NB
instead of IMAGE_REL_AMD64_ADDR32. This is easiely achieved by adding
the VK_COFF_IMGREL32 modifier to the symbol reference.
Change also references to start and end of the SEH range of a function
as offsets to start of the function.
Reviewed by Jim Grosbach, Charles Davis and Nico Rieck.
llvm-svn: 190766
This is causing test-suite failures.
Original commit message:
The PPC backend uses a target-specific DAG combine to turn unaligned Altivec
loads into a permutation-based sequence when possible. Unfortunately, the
target-specific DAG combine is not always called on all loads of interest
(sometimes the routines in DAGCombine call CombineTo such that the new node and
users are not added to the worklist); allowing the combine to trigger early
(before type legalization) mitigates this problem. Because the autovectorizers
only create legal vector types, I don't expect a lot of cases where this
optimization is enabled by type legalization in practice.
llvm-svn: 190765
The PPC backend uses a target-specific DAG combine to turn unaligned Altivec
loads into a permutation-based sequence when possible. Unfortunately, the
target-specific DAG combine is not always called on all loads of interest
(sometimes the routines in DAGCombine call CombineTo such that the new node and
users are not added to the worklist); allowing the combine to trigger early
(before type legalization) mitigates this problem. Because the autovectorizers
only create legal vector types, I don't expect a lot of cases where this
optimization is enabled by type legalization in practice.
llvm-svn: 190764
DAGCombiner::isAlias can be called with SrcValue1 or SrcValue2 null, and we
can't use AA in this case (if we try, then the casting code in AA will assert).
llvm-svn: 190763
Unlike C++11's "thread_local" keyword, C11's "_Thread_local" is in the
reserved namespace, meaning we provide it unconditionally; it is marked
as KEYALL in TokenKinds.def.
This means that like all the other C11 keywords, we can expose its
presence through __has_extension().
llvm-svn: 190755
This is a fix to PR12778: in erroneous code an allocation function
can be declared with no arguments, quering the first argument in this case
causes assertion violation.
llvm-svn: 190751
This pass was based on the previous (essentially unused) profiling
infrastructure and the assumption that by ordering the basic blocks at
the IR level in a particular way, the correct layout would happen in the
end. This sometimes worked, and mostly didn't. It also was a really
naive implementation of the classical paper that dates from when branch
predictors were primarily directional and when loop structure wasn't
commonly available. It also didn't factor into the equation
non-fallthrough branches and other machine level details.
Anyways, for all of these reasons and more, I wrote
MachineBlockPlacement, which completely supercedes this pass. It both
uses modern profile information infrastructure, and actually works. =]
llvm-svn: 190748
Before:
template <template <typename>
class Fooooooo, template <typename>
class Baaaaaaar>
struct C {};
After:
template <template <typename> class Fooooooo,
template <typename> class Baaaaaaar>
struct C {};
llvm-svn: 190747
Summary:
We should treat a non-dependent template specialization like it wasn't
templated at all.
Reviewers: rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1554
llvm-svn: 190743
and a mach kernel in all the pages of the core file. If it finds
a user-process dyld binary, assume this is a user process that had
a copy of the mach kernel in memory when it crashed (e.g. lldb doing
kernel debugging) even though we found the kernel binary first.
Also, change the error messages about sections extending past the end
of the file to be warnings and make the messages sound less severe.
Most user process core files have one section that isn't included in
the file and there's no reason to worry people about that.
<rdar://problem/14473235>
llvm-svn: 190741
This is necessary when running two scan-build processes in parallel. The
directory naming scheme is now:
yyyy-MM-dd-HHmmss-PID-N
2013-09-13-174210-123-1
where "PID" is the scan-build process ID, and "N" is a sequential counter
(not likely to be needed now that seconds are mangled in, but just in case).
PR17196, using a suggested fix from Greg Czajkowski!
llvm-svn: 190735
-S : Specifies a command file which will get sourced after the ~/.lldbinit but before file arguments are processed
-O : Specifies a single (one-line) command that will get ditto
and
-s : Specifies a command file which will get sourced after `pwd`/.lldbinit
-o : Specifies a command file which ditto
I also changed it so that by default these sourced commands will print their command result, but there's a
-q option to change that if you wish.
llvm-svn: 190734
This was somewhat tricky because ~PrettyStackTraceEntry() may run after
llvm_shutdown() has been called. This is rare and only happens for a common idiom
used in the main() functions of command-line tools. This works around the idiom by
skipping the stack clean-up if the PrettyStackTraceHead ManagedStatic is not
constructed (i.e. llvm_shutdown() has been called).
llvm-svn: 190730
1. existing breakpoints weren't being re-resolved after the sections of a library were loaded (ie. through dlopen).
2. loaded sections weren't being removed after a shared library had been unloaded.
llvm-svn: 190727
Summary:
This fixes PR17145 and avoids unknown pragma warnings.
This change does not attempt to map MSVC warning numbers to clang
warning flags. Perhaps in the future we will implement a mapping for
some common subset of Microsoft warnings, but for now we don't.
Reviewers: rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1652
llvm-svn: 190726
There was a bug that if a section has an alignment requirement and there are
multiple symbols at offset 0 in the section, only the last atom at offset 0
would be aligned properly. That bug would move only the last symbol to an
alignment boundary, leaving other symbols unaligned, although they should be at
the same location. That caused a mysterious SEGV error of the resultant
executable.
With this patch, we manage all symbols at the same location properly, rather
than keeping the last one.
llvm-svn: 190724
Alignment(1) does not mean that the atom should be aligned on a 1 byte
boundary but on a 2^1 boundary. So, atoms without any specific alignment
requirements should have Alignment(0).
llvm-svn: 190723
Peter Collingbourne