61 Commits

Author SHA1 Message Date
Daniel Dunbar
9b21e63251 Block pointer types are not aggregate types.
llvm-svn: 61973
2009-01-09 02:44:18 +00:00
Anders Carlsson
5d985f5f16 Handle typedefs to VLAs (Emit the size expr when we encounter the typedef
llvm-svn: 61290
2008-12-20 21:51:53 +00:00
Anders Carlsson
c20879a6e4 Make sure to generate code for arguments that have a variably modified type.
llvm-svn: 61288
2008-12-20 21:28:43 +00:00
Anders Carlsson
8a01b79274 Change EmitVLASize to take a QualType that must be a variably modified type.
Emit the size even if the declared type is a variably modified type. This lets us handle

void f(int n) {
  int (*a)[n];
  
  printf("size: %d\n", sizeof(*a));
}

llvm-svn: 61285
2008-12-20 20:46:34 +00:00
Anders Carlsson
e388a5bf44 Split up emitting of VLA sizes and getting the size of a VLA.
llvm-svn: 61284
2008-12-20 20:27:15 +00:00
Anders Carlsson
ccbe9200f9 Add map of VLA types and their sizes
llvm-svn: 60939
2008-12-12 07:19:02 +00:00
Anders Carlsson
8628645e94 Change more code over to using the new Expr::Evaluate
llvm-svn: 60324
2008-12-01 02:46:24 +00:00
Chris Lattner
86d7d91366 Rename NamedDecl::getIdentifierName() to ::getNameAsCString() and make it
assert if the name is not an identifier.  Update callers to do the right
thing and avoid this method in unsafe cases.  This also fixes an objc
warning that was missing a space, and migrates a couple more to taking
IdentifierInfo and QualTypes instead of std::strings.

llvm-svn: 59936
2008-11-24 03:54:41 +00:00
Anders Carlsson
4046e65b27 An expression is not foldable if it can't be fully evaluated. Fixes PR3060
llvm-svn: 59887
2008-11-22 22:32:07 +00:00
Douglas Gregor
77324f3854 Introduction the DeclarationName class, as a single, general method of
representing the names of declarations in the C family of
languages. DeclarationName is used in NamedDecl to store the name of
the declaration (naturally), and ObjCMethodDecl is now a NamedDecl.

llvm-svn: 59441
2008-11-17 14:58:09 +00:00
Chris Lattner
67d7b9204c rename Expr::tryEvaluate to Expr::Evaluate.
llvm-svn: 59426
2008-11-16 21:24:15 +00:00
Daniel Dunbar
a612e79be4 Normalize many BasicBlock names.
- Use dotted notation for blocks related to a particular statement
   type.
 - Use .end for landing pads.

No functionality change in NDEBUG mode. :)

llvm-svn: 59210
2008-11-13 01:38:36 +00:00
Daniel Dunbar
f32443cdcd Quick fix for crash in IRgen when we can tryEvaluate a condition to
something that is not an int.

 - Ignore these cases for now, added FIXME that we should also boolize
   them.

llvm-svn: 59184
2008-11-12 22:37:10 +00:00
Daniel Dunbar
bf3c22e51d Handle ?: in EmitBranchOnBoolExpr.
llvm-svn: 59129
2008-11-12 10:30:32 +00:00
Chris Lattner
d95377341b Handle Unary ! in EmitBranchOnBoolExpr, so that we can efficiently
codegen stuff like "if (!(X && Y))"

llvm-svn: 59115
2008-11-12 08:13:36 +00:00
Chris Lattner
cd439299b0 Move EmitBranchOnBoolExpr and ConstantFoldsToSimpleInteger to
CodeGenFunction.cpp.  Change VisitConditionalOperator to use
constant fold instead of codegen'ing a constant conditional.

Change ForStmt to use EmitBranchOnBoolExpr, this shrinks
expr.c very slightly to 40239 lines.

llvm-svn: 59113
2008-11-12 08:04:58 +00:00
Daniel Dunbar
5c7e3935f1 Rework IRgen invariant w.r.t. current insert point.
- EmitStmt is no longer required to finish with a current insertion
   point defined (i.e. it does not need to make dummy
   blocks). Instead, it can clear the insertion point in the builder
   which indicates that the current insertion point is unreachable.
 - CodeGenFunction provides HaveInsertPoint and EnsureInsertPoint
   which respectively test if there is an insert point and ensure an
   insertion point exists (by making a dummy block).
 - Clearly mark functions in CodeGenFunction which can be called with
   no insertion point defined. Currently this is a limited set, and
   EmitStmt simply EnsureInsertPoint()s before emitting subsequent IR.

Remove EmitDummyBlock, which is no longer needed. Clients who haven't
already cleared the insertion point (typically via EmitBranch) can do
so by hand.

Remove isDummyBlock, which has effectively been renamed to
HaveInsertPoint.

The main thrust of this change is that we no longer have create dummy
blocks just to destroy them a short time later in EmitBlock in the
common case that there is no unreachable code following something like
a goto. 

Additionally, this means that we are not using the hokey condition in
isDummyBlock that a block without a name is a dummy block. Guess how
well that works when we never emit block names!

llvm-svn: 59089
2008-11-11 23:11:34 +00:00
Daniel Dunbar
fab3f936dd Emit debug region end in unified return block.
llvm-svn: 59081
2008-11-11 20:59:54 +00:00
Chris Lattner
5b1964b5f9 short circuit && and || when possible. This substantially reduces
the size of the -O0 output on some cases.  For example, on expr.c from
176.gcc, it shrinks the .ll file from 43164 to 42835 lines, and removed
references to two external symbols.

llvm-svn: 59034
2008-11-11 07:41:27 +00:00
Daniel Dunbar
05e629aec6 Remove CodeGenFunction::StartBlock.
- Was confusing and only used in one small part of the code.

llvm-svn: 59020
2008-11-11 04:12:31 +00:00
Daniel Dunbar
75283ff38a Centralize basic block creation in CodeGenFunction::createBasicBlock.
- No functionality change.

llvm-svn: 59017
2008-11-11 02:29:29 +00:00
Anders Carlsson
13abd7e98b Implement lowering of va_arg in clang directly. (This is 32-bit X86 only for now).
llvm-svn: 58681
2008-11-04 05:30:00 +00:00
Daniel Dunbar
5b06d9f202 Don't run the verifier as part of IRgen, this is now down (per module)
in the driver (this means we no longer run the verifier per function, however).

llvm-svn: 58307
2008-10-28 00:36:04 +00:00
Daniel Dunbar
354d278518 Debug info bug fix, function start wasn't getting generated correctly
for Obj-C methods.

llvm-svn: 57769
2008-10-18 18:22:23 +00:00
Daniel Dunbar
b9fd902fdc Change CGDebugInfo::setLocation to just ignore invalid locations. This
simplifies clients.

Also, add assert that RegionStack is empty when the CGDebugInfo is
destroyed.

llvm-svn: 57684
2008-10-17 16:15:48 +00:00
Daniel Dunbar
274ff37583 Simplify code to emit the unified return block and move it to the end
of the function.

llvm-svn: 56752
2008-09-27 07:15:59 +00:00
Daniel Dunbar
d9d1f5f522 Print current function on verification failures.
llvm-svn: 56291
2008-09-17 21:13:22 +00:00
Anders Carlsson
9396a89899 Make sure to store the exception in the catch parameter.
llvm-svn: 56102
2008-09-11 09:15:33 +00:00
Daniel Dunbar
613855c8af Move ABI specific code for functions / calls to CGCall.cpp:
- Factor out EmitFunction{Pro,Epi}log

llvm-svn: 56031
2008-09-09 23:27:19 +00:00
Daniel Dunbar
bc915f4025 Factor CodeGenFunction::StartFunction out of GenerateCode and
StartObjCMethod.

llvm-svn: 56030
2008-09-09 23:14:03 +00:00
Daniel Dunbar
54bb1933b1 Use a unified return block.
- For the time being this means our emitted code is somewhat worse,
   especially for aggregates. This will be fixed.

llvm-svn: 56013
2008-09-09 21:00:17 +00:00
Daniel Dunbar
f2cf6d16e6 Avoid superfluous errors regarding variable-length arrays (casts).
llvm-svn: 55759
2008-09-04 03:43:08 +00:00
Anders Carlsson
2e744e8aa6 Stub out CodeGenFunction::EmitObjCForCollectionStmt.
Add CodeGenFunction::EmitMemSetToZero and make AggExprEmitter::EmitAggregateClear use it.

llvm-svn: 55573
2008-08-30 19:51:14 +00:00
Daniel Dunbar
89654eef3a Objective-C @synthesize support.
- Only supports simple assignment and atomic semantics are ignored.
 - Not quite usable yet because the methods do not actually get added
   to the class metadata.
 - Added ObjCPropertyDecl::getSetterKind (one of Assign, Copy, Retain).
 - Rearrange CodeGenFunction so synthesis can reuse function prolog /
   epilog code.

llvm-svn: 55365
2008-08-26 08:29:31 +00:00
Eli Friedman
3d421e190a Do typechecking and codegen for K&R-style function declarations
correctly.  Not a regression, but made more obvious by my recent fix 
which made function type compatibility checking a bit more strict.

llvm-svn: 55339
2008-08-25 21:31:01 +00:00
Daniel Dunbar
a7c8cf6200 Change WarnUnsupported to ErrorUnsupported (in name and in practice).
- We are beyond the point where this shows up often and when it does
   generating miscompiled files is bad.

llvm-svn: 54836
2008-08-16 00:56:44 +00:00
Daniel Dunbar
56fdb6ae69 More #include cleaning
- Kill unnecessary #includes in .cpp files. This is an automatic
   sweep so some things removed are actually used, but happen to be
   included by a previous header. I tried to get rid of the obvious
   examples and this was the easiest way to trim the #includes in one
   fell swoop.
 - We now return to regularly scheduled development.

llvm-svn: 54632
2008-08-11 06:23:49 +00:00
Daniel Dunbar
6e8aa537f8 More #include cleaning
- Drop {Decl.h,DeclObjC.h,IdentifierTable.h} from Expr.h
 - Moved Sema::getCurMethodDecl() out of line (dependent on
   ObjCMethodDecl via dyn_cast).

llvm-svn: 54629
2008-08-11 05:35:13 +00:00
Daniel Dunbar
ad319a73f2 More #include cleaning
- Remove internal uses of AST.h

llvm-svn: 54628
2008-08-11 05:00:27 +00:00
Daniel Dunbar
d27262f369 Avoid creating dummy block when no indirect gotos are present.
llvm-svn: 54322
2008-08-04 17:24:44 +00:00
Daniel Dunbar
88402ce8d0 Add CodeGen support for indirect goto.
- Follows emission scheme used by llvm-gcc, i.e. invent an id for
   each label whose address is taken and replace each indirect goto by
   a switch to each possible target.
 - Currently we emit a switch for each indirect goto instead of
   merging them as llvm-gcc does.

llvm-svn: 54318
2008-08-04 16:51:22 +00:00
Daniel Dunbar
9c426524f8 Rework codegen emission of globals
- No (intended) functionality change.

 - Primary purpose is to clearly separate (lazy) construction of
   globals that are a forward declaration or tentative definition from
   those that are the final definition.

 - Lazy construction is now encapsulated in
   GetAddrOf{Function,GlobalVar} while final definitions are
   constructed in EmitGlobal{Function,Var}Definition.
 
 - External interface for dealing with globals is now limited to
   EmitGlobal and GetAddrOf{Function,GlobalVar}.

 - Also updated helper functions dealing with statics, annotations,
   and ctors to be private.

llvm-svn: 54179
2008-07-29 23:18:29 +00:00
Chris Lattner
4591fd0aac Fix rdar://6103124, the only dummy blocks are ones without names. Named blocks
are important part of control flow structures.

llvm-svn: 54078
2008-07-25 23:40:10 +00:00
Sanjiv Gupta
1e8b608751 Fixed bugzilla bug# 2489.
This was broken when the GenerateCode function was splitted to use GenerateFunction.

llvm-svn: 53136
2008-07-04 11:04:26 +00:00
Chris Lattner
5696e7badf Change self/_cmd to be instances of ImplicitParamDecl instead of ParmVarDecl.
Patch by David Chisnall!

llvm-svn: 52422
2008-06-17 18:05:57 +00:00
Eli Friedman
327944b343 Basic support for volatile loads and stores. Stores the volatile
qualifier in the lvalue, and changes lvalue loads/stores to honor 
the volatile flag.  Places which need some further attention are marked 
with FIXMEs.

Patch by Cédric Venet.

llvm-svn: 52264
2008-06-13 23:01:12 +00:00
Eli Friedman
f6b091f73a First cut at setting attributes for functions and calls; this puts us
much closer to passing the gcc struct layout tests.

It might be possible to refactor this a bit, but I'm not sure there's 
actually enough common code for that to be useful.

To get the calling convention completely correct, a bit of 
platform-specific code is necessary even for x86-Linux. On x86-Linux, the
alignment of function parameters is extremely strange; as far as I can tell,
it's always 4 except for SSE vectors or structs containing SSE vectors.  I'm 
continuing to investigate this.

llvm-svn: 51839
2008-06-01 15:54:49 +00:00
Anton Korobeynikov
1200aca763 Support for code generation of Objective-C top-level language constructs.
Implemented by David Chisnall!

llvm-svn: 51835
2008-06-01 14:13:53 +00:00
Sanjiv Gupta
98070578bf Generate subprogram debug info with -g.
Also take care of freeing memory at the right places.

llvm-svn: 51553
2008-05-25 05:15:42 +00:00
Eli Friedman
1763075b9c Make debugging information usable. This is barebones, but it makes -g
actually work (instead of crashing llc), and there's enough info emitted 
to get line number information in gdb. This should hopefully be helpful 
for debugging non-working programs.

I got rid of the begin/endregion calls because the implementation wasn't 
working; someone who knows the debugging info a bit better might try to 
add it. I really have no clue how a compiler is supposed to emit them.

This commit shouldn't have any effect without -g.

llvm-svn: 51404
2008-05-22 01:40:10 +00:00