CreateEnumType doesn't participate in caching so the descriptor for the enum
gets recomputed for every reference of an element of an enum, only to get
discarded when it gets turned into an MDNode.
No functionality change except performance.
llvm-svn: 158832
Regunit live ranges are computed on demand, so when mi-sched calls
handleMove, some regunits may not have live ranges yet.
That makes updating them easier: Just skip the non-existing ranges. They
will be computed correctly from the rescheduled machine code when they
are needed.
llvm-svn: 158831
to get it working again, two changes were needed:
- I had to give up on glob-based file expansion. This just isn't well
supported by CMake, and until we convince upstream there of its value,
it's not worth dealing with the pain.
- Add the common library as otherwise even ASan won't build.
This now builds again, although the "correctness" of it is a touch
debatable. ;] Specifically, there is no merging of the common runtime
library with the asan runtime library into a single archive file. I'm
not really sure what the best technique is for that, and it may be
influenced by the ongoing discussion about how best to link runtime
libraries.
Note of course that this is still very much WIP. It doesn't entirely
work yet, and remains disabled by the LLVM projects/CMakeLists.txt until
it is in a working state.
llvm-svn: 158811
That commit added a new library just to hold the RawCommentList. I've
started a discussion on the commit thread about whether that is really
meritted -- it certainly doesn't seem necessary at this stage.
However, the immediate problem is that the AST library has a hard
dependency on the Comment library, but the dependencies were set up
completely backward. In addition to the layering violation, this had an
unfortunate effect if scattering the Comments library dependency
throughout the build system, but inconsistently so -- several parts of
the CMake dependencies were missing and only showed up due to transitive
deps or the fact that the target wasn't being built by tho bots.
It turns out that the Comments library can't (currently) be a well
formed layer *below* the AST library either, as it has an API that
accepts an ASTContext. That parameter is currently unused, so maybe that
was a mistake?
Anyways, it really seems like this is logically part of the AST --
that's the whole point of the ASTContext providing access to it as far
as I can tell -- so I've merged it into the AST library to solve the
immediate layering violation problems and remove some of the churn from
our library dependencies.
llvm-svn: 158807
There is a pretty staggering amount of this in LLVM's header files, this
is not all of the instances I'm afraid. These include all of the
functions that (in my build) are used by a non-static inline (or
external) function. Specifically, these issues were caught by the new
'-Winternal-linkage-in-inline' warning.
I'll try to just clean up the remainder of the clearly redundant "static
inline" cases on functions (not methods!) defined within headers if
I can do so in a reliable way.
There were even several cases of a missing 'inline' altogether, or my
personal favorite "static bool inline". Go figure. ;]
llvm-svn: 158800
I'll admit I'm not entirely satisfied with this change, but it seemed
the cleanest option. Other suggestions quite welcome
The issue is that the traits specializations have static methods which
return the typedef'ed PHI_iterator type. In both the IR and MI layers
this is typedef'ed to a custom iterator class defined in an anonymous
namespace giving the types and the functions returning them internal
linkage. However, because the traits specialization is defined in the
'llvm' namespace (where it has to be, specialized template lives there),
and is in turn used in the templated implementation of the SSAUpdater.
This led to the linkage conflict that Clang now warns about.
The simplest solution to me was just to define the PHI_iterator as
a nested class inside the trait specialization. That way it still
doesn't get scoped widely, it can't be accidentally reused somewhere,
etc. This is a little gross just because nested class definitions are
a little gross, but the alternatives seem more ad-hoc.
llvm-svn: 158799
The TEST_F macros actually declare *subclasses* of the test fixtures.
Even if they didn't we don't want them to declare external functions.
The entire unit test, including both the fixture class and the fixture
test cases should be wrapped in the anonymous namespace.
This issue was caught by the new '-Winternal-linkage-in-inline' warning.
llvm-svn: 158798
../tools/clang/lib/Serialization/ASTReader.cpp:6316:9: warning: default label in switch which covers all enumeration values [-Wcovered-switch-default]
Also fix the indentation here to match the coding conventions.
llvm-svn: 158794
target Objective-C runtime down to the frontend: break this
down into a single target runtime kind and version, and compute
all the relevant information from that. This makes it
relatively painless to add support for new runtimes to the
compiler. Make the new -cc1 flag, -fobjc-runtime=blah-x.y.z,
available at the driver level as a better and more general
alternative to -fgnu-runtime and -fnext-runtime. This new
concept of an Objective-C runtime also encompasses what we
were previously separating out as the "Objective-C ABI", so
fragile vs. non-fragile runtimes are now really modelled as
different kinds of runtime, paving the way for better overall
differentiation.
As a sort of special case, continue to accept the -cc1 flag
-fobjc-runtime-has-weak, as a sop to PLCompatibilityWeak.
I won't go so far as to say "no functionality change", even
ignoring the new driver flag, but subtle changes in driver
semantics are almost certainly not intended.
llvm-svn: 158793
Per Anna's comment, this is a better way to handle "to-do list"-type failures.
This way we'll know if any of the features get fixed; in an XFAIL file, /all/
the cases have to be fixed before lit would tell us anything.
llvm-svn: 158791
-stable-loops enables a new algorithm for generating the Loop
forest. It differs from the original algorithm in a few respects:
- Not determined by use-list order.
- Initially guarantees RPO order of block and subloops.
- Linear in the number of CFG edges.
- Nonrecursive.
I didn't want to change the LoopInfo API yet, so the block lists are
still inclusive. This seems strange to me, and it means that building
LoopInfo is not strictly linear, but it may not be a problem in
practice. At least the block lists start out in RPO order now. In the
future we may add an attribute or wrapper analysis that allows other
passes to assume RPO order.
The primary motivation of this work was not to optimize LoopInfo, but
to allow reproducing performance issues by decomposing the compilation
stages. I'm often unable to do this with the current LoopInfo, because
the loop tree order determines Loop pass order. Serializing the IR
tends to invert the order, which reverses the optimization order. This
makes it nearly impossible to debug interdependent loop optimizations
such as LSR.
I also believe this will provide more stable performance results across time.
llvm-svn: 158790
The implementation only needs inclusion from LoopInfo.cpp and
MachineLoopInfo.cpp. Clients of the interface should only include the
interface. This makes the interface readable and speeds up rebuilds
after modifying the implementation.
llvm-svn: 158787
The default global placement new just returns the pointer it is given.
Note that other custom 'new' implementations with placement args are not
guaranteed to do this.
In addition, we need to invalidate placement args, since they may be updated by
the allocator function. (Also, right now we don't properly handle the
constructor inside a CXXNewExpr, so we need to invalidate the placement args
just so that callers know something changed!)
This invalidation is not perfect because CallOrObjCMessage doesn't support
CXXNewExpr, and all of our invalidation callbacks expect that if there's no
CallOrObjCMessage, the invalidation is happening manually (e.g. by a direct
assignment) and shouldn't affect checker-specific metadata (like malloc state);
hence the malloc test case in new-fail.cpp. But region values are now
properly invalidated, at least.
The long-term solution to this problem is to rework CallOrObjCMessage into
something more general, rather than the morass of branches it is today.
<rdar://problem/11679031>
llvm-svn: 158784
[NSNumber numberWithDouble:cppb];
warning: converting to boxing syntax requires a cast
to something like:
[NSNumber numberWithDouble:cppb];
warning: converting to boxing syntax requires casting 'bool' to 'double'
This is way better to fully understand the warning.
rdar://11705106
llvm-svn: 158783
Benjamin Kramer