Chris Lattner
74ed76bb0e
remove a dead argument
...
llvm-svn: 41377
2007-08-24 21:41:10 +00:00
Chris Lattner
9f0ad96b3e
implement codegen for real/imag. TODO: imag of non-complex.
...
llvm-svn: 41376
2007-08-24 21:20:17 +00:00
Chris Lattner
30b5dd0b79
Implement sema support for __real/__imag nodes.
...
llvm-svn: 41375
2007-08-24 21:16:53 +00:00
Ted Kremenek
066dd939c0
Added child_begin/child_end to all subclasses of Stmt in Stmt.h. All
...
concrete subclasses of Stmt are now required to implement
child_begin/child_end.
llvm-svn: 41374
2007-08-24 21:09:09 +00:00
Chris Lattner
3d966d6556
Teach emit-llvm for scalars to properly handle compound assignment
...
operators in all their glory :)
llvm-svn: 41373
2007-08-24 21:00:35 +00:00
Dale Johannesen
7891d8edf0
Use APFloat internally for ConstantFPSDNode.
...
llvm-svn: 41372
2007-08-24 20:59:15 +00:00
Chris Lattner
0b490b0d38
sink clone() down the class hierarchy from CmpInst into ICmpInst/FCmpInst.
...
This eliminates a conditional on that path, and ensures ICmpInst/FCmpInst
both have an out-of-line virtual method to home the class.
llvm-svn: 41371
2007-08-24 20:48:18 +00:00
Ted Kremenek
485b13f397
Added ExprCXX.cpp
...
llvm-svn: 41370
2007-08-24 20:24:16 +00:00
Ted Kremenek
e3a0c14614
Implementation of child_begin/child_end for C++ expressions.
...
llvm-svn: 41369
2007-08-24 20:21:10 +00:00
Ted Kremenek
acc4903b3e
Implemented child_begin/child_end for our current set of C++ expressions.
...
llvm-svn: 41368
2007-08-24 20:20:38 +00:00
Ted Kremenek
23702b6afb
Finished adding child_begin/child_end to all subclasses of Stmt in Expr.h.
...
llvm-svn: 41366
2007-08-24 20:06:47 +00:00
Devang Patel
f5a01bf025
Fix regression that I caused yesterday night while adding logic to select appropriate split condition branch.
...
llvm-svn: 41365
2007-08-24 19:32:26 +00:00
Steve Naroff
be4c4d14c3
Surpress ImplicitCastExprs for compound assignment expressions. For compound assignments,
...
it is o.k. for the LHS and RHS to have different types. Converting the type can cause
errors like the one Chris noticed (below).
This change required a fair number of diffs (since there is a lot of shared code
between single and compound assignments). This makes the API's look a bit uglier,
however I couldn't think of a better way to do it (without duplicating code).
Fix the following (incorrect) error:
int A;
long long B;
void foo() {
A /= B;
}
$ clang ~/scalar.c -emit-llvm
/Users/sabre/scalar.c:6:5: error: expression is not assignable
A /= B;
~ ^
Now it works properly...
[dylan:~/llvm/tools/clang] admin% cat compound.c
int A;
long long B;
void foo() {
A /= B;
}
[dylan:~/llvm/tools/clang] admin% ../../Debug/bin/clang compound.c -parse-ast-dump
Read top-level variable decl: 'A'
Read top-level variable decl: 'B'
void foo()
(CompoundStmt 0x2605c40
(BinaryOperator 0x2605c20 'int' '/=' ComputeTy='long long'
(DeclRefExpr 0x2605be0 'int' Decl='A' 0x2605a80)
(DeclRefExpr 0x2605c00 'long long' Decl='B' 0x2605ab0)))
llvm-svn: 41364
2007-08-24 19:07:16 +00:00
Ted Kremenek
85e92ec6ff
Began implementing "child iterator" interface for Stmts and Exprs. Each
...
subclass of Stmt will implement child_begin() and child_end(), which will
be used to iterate over all the children (subexpressions/substatements) of
a Stmt object. This will provide for easy traversal over the AST, which
is useful for a variety of purposes.
None of the interfaces to subclasses of Stmt will be changed (other than
adding the child_begin and child_end methods).
The only caveat is that the implementation of subclasses of Stmt will require
colocating all substatements (subexpressions) in an array. This is because
we define child_iterator as Stmt**. All accessor methods to subexpressions
will need to be altered to reflect this new implementation.
This patch includes the typedefs for child_iterator, as well the implementation
for child_begin/child_end for the primary expressions and some postfix
expressions.
llvm-svn: 41363
2007-08-24 18:13:47 +00:00
Steve Naroff
9d13917cff
Make sure we get extension diagnostics for GCC's complex extensions.
...
Now we emit the following when -pedantic-errors is enabled...
[dylan:~/llvm/tools/clang] admin% ../../Debug/bin/clang complex.c -pedantic-errors
complex.c:4:3: error: ISO C does not support '++'/'--' on complex integer types
++x;
^ ~
complex.c:9:7: error: ISO C does not support '~' for complex conjugation
X = ~Y;
^
complex.c:10:7: error: ISO C does not support '~' for complex conjugation
x = ~y;
^
llvm-svn: 41362
2007-08-24 17:20:07 +00:00
Chris Lattner
100198f3f7
print the computation type for compound assignment operators in dumps.
...
llvm-svn: 41361
2007-08-24 16:24:49 +00:00
Chris Lattner
a124f69c52
Disable EH generation until PPC works 100%.
...
llvm-svn: 41360
2007-08-24 16:00:15 +00:00
Chris Lattner
51883acec1
add a note
...
llvm-svn: 41359
2007-08-24 15:17:59 +00:00
Devang Patel
4bc9298f2a
It is not safe to execute split condition's true branch first all the time. If split
...
condition predicate is GT or GE then execute false branch first.
llvm-svn: 41358
2007-08-24 06:17:19 +00:00
Devang Patel
4be56a5d12
Reject ICMP_NE as index split condition.
...
llvm-svn: 41357
2007-08-24 06:02:25 +00:00
Devang Patel
5e46fac6de
Tightenup loop filter.
...
llvm-svn: 41356
2007-08-24 05:36:56 +00:00
Chris Lattner
2da04b3322
completely refactor codegen of scalar expressions out into its own CGExprScalar.cpp file.
...
This patch temporarily breaks compound assignment operators, but greatly simplifies many
things.
llvm-svn: 41355
2007-08-24 05:35:26 +00:00
Devang Patel
504dc0aaed
Remove incomplete cost analysis.
...
llvm-svn: 41354
2007-08-24 05:21:13 +00:00
Dale Johannesen
918c33c67a
Revised per review feedback from previous patch.
...
llvm-svn: 41353
2007-08-24 05:08:11 +00:00
Chris Lattner
3da18eb22a
silence some warnings in an optimized build.
...
llvm-svn: 41352
2007-08-24 03:02:34 +00:00
Chris Lattner
11e0de5193
rename two files.
...
llvm-svn: 41350
2007-08-24 02:22:53 +00:00
Chris Lattner
db0d60d5b5
implement codegen of builtin_choose_expr for complex.
...
llvm-svn: 41349
2007-08-24 02:18:47 +00:00
Dale Johannesen
a719a60609
Change internal representation of ConstantFP to use APFloat.
...
Interface to rest of the compiler unchanged, as yet.
llvm-svn: 41348
2007-08-24 00:56:33 +00:00
Chris Lattner
8ad56707ea
stub out complex -> bool conversion.
...
llvm-svn: 41347
2007-08-24 00:01:20 +00:00
Chris Lattner
afd455c0cf
make this harder
...
llvm-svn: 41346
2007-08-23 23:49:47 +00:00
Chris Lattner
0b984995b7
implement complex subtraction
...
llvm-svn: 41345
2007-08-23 23:46:33 +00:00
Chris Lattner
08b15df766
implement passing of complex and aggregates through call args.
...
llvm-svn: 41344
2007-08-23 23:43:33 +00:00
Chris Lattner
247ef955bc
in the truncation case, make sure to propagate the sign correctly, this
...
fixes an assertion on:
void f (int z) { switch (z) { case ~0ULL: case -1: return; } }
testcase from Neil.
llvm-svn: 41343
2007-08-23 22:08:35 +00:00
Steve Naroff
8ddb23a6c5
Support '~' for complex conjugation. This is a GCC extension.
...
This following now compiles without error...
_Complex unsigned X, Y;
_Complex double x, y;
void test2(int c) {
X = ~Y;
x = ~y;
}
llvm-svn: 41341
2007-08-23 22:06:40 +00:00
Chris Lattner
7b939cf606
fix a bug where we would type 0ULL as unsigned instead of unsigned long long
...
llvm-svn: 41340
2007-08-23 21:58:08 +00:00
Chris Lattner
e5b60445e7
Pretty print as:
...
"case sizeof x:"
instead of:
"case sizeofx:"
llvm-svn: 41339
2007-08-23 21:46:40 +00:00
Chris Lattner
9568880c3e
sizeof(x) doesn't require x to be an i-c-e for sizeof to be an i-c-e. Thanks to Neil for pointing this out.
...
llvm-svn: 41338
2007-08-23 21:42:50 +00:00
Ted Kremenek
9aae513318
No functionality change. Moved visitor methods for CFGBuilder out-of-line
...
from the class declaration. This enables a nice view of what visitor methods
have been implemented.
llvm-svn: 41337
2007-08-23 21:42:29 +00:00
Chris Lattner
1aaab49a3e
add support for codegen of calls returning complex.
...
llvm-svn: 41336
2007-08-23 21:38:16 +00:00
Steve Naroff
9e1e551c01
Remove a FIXME, allowing ++/-- on Complex types (a GCC extension).
...
Now, the following test case succeeds...
_Complex double x, y;
void test2(int c) {
++x;
}
llvm-svn: 41335
2007-08-23 21:37:33 +00:00
Ted Kremenek
04cca64f8c
moved CFGBuilder into an anonymous namespace
...
llvm-svn: 41334
2007-08-23 21:26:19 +00:00
Steve Naroff
4e1f80d255
- Cleanup "hack" comment and fix typos.
...
- Use getLang().ObjC2 when appropriate.
llvm-svn: 41333
2007-08-23 19:56:30 +00:00
Ted Kremenek
879d8e1f77
Added support for switch, default, and case statements in source-level CFGs.
...
llvm-svn: 41332
2007-08-23 18:43:24 +00:00
Chris Lattner
67998451c7
finish off switch case overlap checking, adding support for
...
verifying case ranges.
llvm-svn: 41331
2007-08-23 18:29:20 +00:00
Steve Naroff
7c34817902
Add helper functions Token::isObjCAtKeyword() and Token::getObjCKeywordID().
...
Convert all clients to the new cleaner, more robust API.
llvm-svn: 41330
2007-08-23 18:16:40 +00:00
Ted Kremenek
13fdfe99cd
Refactored "getSubStmt" and "SubStmt" from the CaseStmt and DefaultStmt
...
class to their common parent class SwitchCase.
llvm-svn: 41329
2007-08-23 18:04:38 +00:00
Chris Lattner
f81460f99c
detect and diagnose empty case ranges:
...
switch.c:16:8: warning: empty case range specified
case 100 ... 99: ; // expected-warning {{empty case range}}
^~~~~~~~~~
llvm-svn: 41328
2007-08-23 17:48:14 +00:00
Ted Kremenek
9ee3bcf81c
For gotos, breaks, and continues where we cannot find a target successor
...
block (because we are creating a CFG from an incomplete AST) we now
(gracefully) have a block ending with such statements not have any successors
instead of firing an assertion.
llvm-svn: 41327
2007-08-23 17:29:58 +00:00
Duncan Sands
a0f12c5493
Fix this testcase: there are two matches for
...
llvm.cttz.i64 because of the declaration of
the intrinsic. Also, emit-llvm is automatic
and doesn't need to be specified.
llvm-svn: 41326
2007-08-23 17:22:50 +00:00
Ted Kremenek
9443f5b1e0
Added support for do..while loops in CFG construction.
...
llvm-svn: 41325
2007-08-23 17:15:32 +00:00