156 Commits

Author SHA1 Message Date
Anders Carlsson
233e272775 More CharUnits conversion.
llvm-svn: 117875
2010-10-31 21:54:55 +00:00
Anders Carlsson
725190ff3f More CharUnits conversion.
llvm-svn: 117874
2010-10-31 21:39:24 +00:00
Anders Carlsson
f8f756d1d5 Use CharUnits in the EmptyClassOffsets map.
llvm-svn: 117873
2010-10-31 21:22:43 +00:00
Anders Carlsson
22f5720c27 Baby steps towards using only CharUnits for base class offsets in ASTRecordLayout. Start by storing the offsets in CharUnits in the ASTRecordLayout object.
llvm-svn: 117869
2010-10-31 21:01:46 +00:00
Benjamin Kramer
2fc373e155 Move classes into anonymous namespaces.
llvm-svn: 117104
2010-10-22 16:33:16 +00:00
Argyrios Kyrtzidis
6843141d39 Store in PCH the key function of C++ class to avoid deserializing the complete declaration context in order to compute it.
Progress for rdar://7260160.

llvm-svn: 116508
2010-10-14 20:14:38 +00:00
Argyrios Kyrtzidis
8c64bbe032 Embrace C++ ABI 5.2.6 and consider that template instantiations don't have key functions (same as GCC).
llvm-svn: 116391
2010-10-13 02:39:41 +00:00
Argyrios Kyrtzidis
ca0d0cd3b9 Implement -Wpadded and -Wpacked.
-Wpadded warns when undesired padding is introduced in a struct. (rdar://7469556)
-Wpacked warns if a struct is given the packed attribute, but the packed attribute has no effect
  on the layout or the size of the struct. Such structs may be mis-aligned for little benefit.

The warnings are emitted at the point where layout is calculated, that is at RecordLayoutBuilder.
To avoid calculating the layouts of all structs regardless of whether they are needed or not,
I let the layouts be lazily constructed when needed. This has the disadvantage that the above warnings
will be emitted only when they are used for IR gen, and not e.g with -fsyntax-only:

$ cat t.c
struct S {
  char c;
  int i;
};
void f(struct S* s) {}

$ clang -fsyntax-only -Wpadded t.c
$ clang -c -Wpadded t.c -o t.o
t.c:3:7: warning: padding struct 'struct S' with 3 bytes to align 'i' [-Wpadded]
  int i;
      ^
1 warning generated.

This is a good tradeoff between providing the warnings and not calculating layouts for all
structs in case the user has enabled a couple of rarely used warnings.

llvm-svn: 114544
2010-09-22 14:32:24 +00:00
Argyrios Kyrtzidis
3c6cd17c8a Use a smart pointer instead of delete.
llvm-svn: 112005
2010-08-25 00:32:14 +00:00
Chandler Carruth
76c0cb493f Add a virtual destructor to the base of another class hierarchy with virtual
methods.

llvm-svn: 111804
2010-08-23 08:37:21 +00:00
Charles Davis
6bcb07ad71 Add some enum goodness as requested by Chris. Now instead of storing the
active C++ ABI as a raw string, we store it as an enum. This should improve
performance somewhat.

And yes, this time, I started from a clean build directory, and
all the tests passed. :)

llvm-svn: 111507
2010-08-19 02:18:14 +00:00
John McCall
1ed0182413 Restore the build
llvm-svn: 111498
2010-08-19 01:21:57 +00:00
Charles Davis
c2c576a9e7 Add a special RecordLayoutBuilder for the Microsoft C++ ABI.
All it does right now is add space for two vtable pointers instead of one
when a class has both virtual methods and virtual bases. This is a requirement
of the Microsoft ABI, since it has separate vtables for methods and bases. Other
stuff will come up over time, but we'll cross those bridges when we get to
them.

llvm-svn: 111493
2010-08-19 00:55:19 +00:00
Alexis Hunt
dcfba7b35b Generate Attr subclasses with TableGen.
Now all classes derived from Attr are generated from TableGen.
Additionally, Attr* is no longer its own linked list; SmallVectors or
Attr* are used. The accompanying LLVM commit contains the updates to
TableGen necessary for this.

Some other notes about newly-generated attribute classes:

 - The constructor arguments are a SourceLocation and a Context&,
   followed by the attributes arguments in the order that they were
   defined in Attr.td

 - Every argument in Attr.td has an appropriate accessor named getFoo,
   and there are sometimes a few extra ones (such as to get the length
   of a variadic argument).

Additionally, specific_attr_iterator has been introduced, which will
iterate over an AttrVec, but only over attributes of a certain type. It
can be accessed through either Decl::specific_attr_begin/end or
the global functions of the same name.

llvm-svn: 111455
2010-08-18 23:23:40 +00:00
Daniel Dunbar
56df97761d StringRef'ication of lots stuff, patch by Peter Davies!
llvm-svn: 111314
2010-08-17 22:39:59 +00:00
Argyrios Kyrtzidis
499e6e49e7 Remove ATTRIBUTE_UNUSED from the common pattern of disallowing copying.
llvm-svn: 111101
2010-08-15 10:17:39 +00:00
Argyrios Kyrtzidis
0a33cb62a6 Add ATTRIBUTE_UNUSED to methods not supposed to be used.
llvm-svn: 111087
2010-08-15 01:15:27 +00:00
Daniel Dunbar
f35e76552f Sema: Fix a subtle i64 -> i32 truncation which broke layout of large structures
with bit-fields.

llvm-svn: 107185
2010-06-29 18:34:35 +00:00
Jeffrey Yasskin
bb219e01a6 Delete assert in ComputeKeyFunction. The function runs fine without it, since
there's an explicit guard on isPolymorphic, and virtual bases don't affect the
key function calculation.  This allows people to call
ASTContext::getKeyFunction on arbitrary classes.

llvm-svn: 107143
2010-06-29 15:27:35 +00:00
Anders Carlsson
cc59cc5d80 Do the same short-circuit optimization when laying out bases.
llvm-svn: 105920
2010-06-13 18:00:18 +00:00
Anders Carlsson
ae111dc821 Implement part of the EmptySubobjectMap optimization described in PR6998. We still need to do this for bases.
llvm-svn: 105919
2010-06-13 17:49:16 +00:00
Anders Carlsson
446872069f Correctly handle fields with virtual bases containing empty subobjects.
llvm-svn: 105628
2010-06-08 19:09:24 +00:00
Anders Carlsson
45c1d2898e When checking whether we can place a base subobject at an offset, we don't need to go past the highest offset that's known to contain an empty base subobject.
llvm-svn: 105611
2010-06-08 16:20:35 +00:00
Anders Carlsson
cc5de097a6 Minor cleanups to the empty subobject map.
llvm-svn: 105608
2010-06-08 15:56:03 +00:00
Anders Carlsson
7498b32117 And now for the best part: Removing the old code.
llvm-svn: 105162
2010-05-30 06:56:46 +00:00
Anders Carlsson
b1fcdd063c Turn on the new empty base subobject tracking code. It's a bit faster than the previous code. However, it still has quadratic performance, something which I intend to fix shortly in a subsequent patch.
llvm-svn: 105161
2010-05-30 06:52:33 +00:00
Anders Carlsson
a7774a685d Cleanup.
llvm-svn: 105114
2010-05-29 21:10:24 +00:00
Anders Carlsson
a60b86a3fa Make EmptySubobjectMap::CanPlaceBaseAtOffset take a BaseSubobjectInfo as well.
llvm-svn: 105113
2010-05-29 20:49:49 +00:00
Anders Carlsson
d7f3fcf248 Change RecordLayoutBuilder::LayoutBase to take a BaseSubobjectInfo. No functionality change.
llvm-svn: 105112
2010-05-29 20:47:33 +00:00
Anders Carlsson
6b0d914ad0 Rework the way virtual primary bases are added when laying out classes. Instead of doing it as a separate step, we now use the BaseSubobjectInfo and use it when laying out the bases. This fixes a bug where we would either not add a primary virtual base at all, or add it at the wrong offset.
llvm-svn: 105110
2010-05-29 19:44:50 +00:00
Anders Carlsson
d6ff5d7787 Change LayoutVirtualBase to also take a BaseSubobjectInfo.
llvm-svn: 105104
2010-05-29 17:48:36 +00:00
Anders Carlsson
bb0e6787b0 Change RecordLayoutBuilder::LayoutNonVirtualBase to take a BaseSubobjectInfo. No functionality change.
llvm-svn: 105103
2010-05-29 17:42:25 +00:00
Anders Carlsson
e3c24c7033 Move computing the base subobject info for a class into the RecordLayoutBuilder because we're going to need it to be able to correctly add offsets for primary virtual bases.
llvm-svn: 105102
2010-05-29 17:35:14 +00:00
Anders Carlsson
a7f3cdb0f2 Rename BaseInfo to BaseSubobjectInfo.
llvm-svn: 105007
2010-05-28 21:24:37 +00:00
Anders Carlsson
056818f8a2 Move BaseInfo outside of the EmptySubobject class.
llvm-svn: 105002
2010-05-28 21:13:31 +00:00
Anders Carlsson
db319761e7 More work on the empty subobject map. This code is not yet used.
llvm-svn: 104861
2010-05-27 18:20:57 +00:00
Daniel Dunbar
6da1098b4b Sema: Add initial support for '#pragma options align=mac68k'.
- Docs are fairly sketchy, if someone wants to pore through gcc to look for
   holes I'd appreciate any failing test cases!

llvm-svn: 104809
2010-05-27 05:45:51 +00:00
Anders Carlsson
439edd1aea More work on the new empty subobject map.
llvm-svn: 104808
2010-05-27 05:41:06 +00:00
Daniel Dunbar
592a85c137 Strip trailing whitespace.
llvm-svn: 104801
2010-05-27 02:25:46 +00:00
Daniel Dunbar
401304462a AST: Rename PragmaPackAttr to MaxFieldAlignmentAttr, which is more accurate.
llvm-svn: 104795
2010-05-27 01:12:46 +00:00
Anders Carlsson
c121b4e59f More work on the empty subobjects map.
llvm-svn: 104787
2010-05-27 00:07:01 +00:00
Anders Carlsson
c5ca1f7799 Move ComputeEmptySubobjectSizes to EmptySubobjectMap.
llvm-svn: 104702
2010-05-26 15:54:25 +00:00
Anders Carlsson
f58de11b61 Stub out the EmptySubobjectsMap class.
llvm-svn: 104701
2010-05-26 15:32:58 +00:00
Anders Carlsson
ce7f77729a Inline the RecordLayoutBuilder constructor.
llvm-svn: 104700
2010-05-26 15:20:08 +00:00
Anders Carlsson
c28a6c9251 Add a Layout overload that takes a CXXRecordDecl.
llvm-svn: 104695
2010-05-26 15:10:00 +00:00
Benjamin Kramer
c7656cdf83 Move RecordLayoutBuilder into an anonymous namespace.
llvm-svn: 104693
2010-05-26 09:58:31 +00:00
Anders Carlsson
c222620adf Rename ASTRecordLayoutBuilder to RecordLayoutBuilder.
llvm-svn: 104688
2010-05-26 05:58:59 +00:00
Anders Carlsson
35a36eb78d Move the ASTRecordLayoutBuilder class declaration into the .cpp file.
llvm-svn: 104686
2010-05-26 05:41:04 +00:00
Anders Carlsson
b48414f4cb Fix thinko and remove another unused function.
llvm-svn: 104683
2010-05-26 05:31:23 +00:00
Anders Carlsson
1d1e4cf04d Fix build.
llvm-svn: 104682
2010-05-26 05:25:15 +00:00