1. Eliminate the costly live interval "swapping".
2. Change ValueNumberInfo container from SmallVector to std::vector. The former
performs slowly when the vector size is very large.
llvm-svn: 41536
t.c:1:12: warning: ISO C restricts enumerator values to range of 'int' (180388626432 is too large)
enum e {A, B = 42LL << 32, C = -4, D = 12456 };
^
llvm-svn: 41530
For example, the following code was resulting in spurious warnings. This was the result of
Sema::GetTypeForDeclarator() synthesizing a type to hand back to the caller (in this case,
"int []", instead of "struct s[]", which is invalid).
struct s;
struct s* t (struct s z[]) { // expected-error {{array has incomplete element type}}
return z;
}
Strategy: Flag the error in Declarator/DeclaratorChunk. This info is later stored in
the ParmVarDecl. If the decl is referenced, Sema::ParseIdentifierExpr() will check if
the type is invalid. If so, it quietly returns "true", without instantiating a DeclRefExpr.
This seems to work nicely. If Chris is happy with the approach, I will generalize this to
all VarDecls.
llvm-svn: 41521
+ Added the creation of an empty Entry block at the end of CFG
construction if the Entry block in the CFG contains multiple
predecessors (which can happen with labels and do loops).
+ Fixed bug in the creation of an empty Exit block with functions where not
all paths end in a return statement (but some do). Basic blocks with
return statements were jumping to a (sometimes) non-empty block.
+ FinishBlock no longer checks for labels at the beginning of a basic
block before reversing the statements in the block. This is because
the recursion invariants of the builder methods have been cleaned up,
and blocks are only passed to FinishBlock at most once.
+ Modified handling of "if", "for", "while", "do", and "switch" to allow
condition expressions that can span multiple basic blocks. This allows
such conditions to contain short-circuit expressions (which span multiple
blocks in the CFG).
llvm-svn: 41508
implicit casts from T to T& at the topmost part of the return-value expression.
This checking may be needed within EvalAddr later on. We'll wait until
test cases show this kind of logic is necessary (as more C++ features are
implemented in clang).
llvm-svn: 41493
gcc exception handling: if an exception unwinds through
an invoke, then execution must branch to the invoke's
unwind target. We previously tried to enforce this by
appending a cleanup action to every selector, however
this does not always work correctly due to an optimization
in the C++ unwinding runtime: if only cleanups would be
run while unwinding an exception, then the program just
terminates without actually executing the cleanups, as
invoke semantics would require. I was hoping this
wouldn't be a problem, but in fact it turns out to be the
cause of all the remaining failures in the LLVM testsuite
(these also fail with -enable-correct-eh-support, so turning
on -enable-eh didn't make things worse!). Instead we need
to append a full-blown catch-all to the end of each
selector. The correct way of doing this depends on the
personality function, i.e. it is language dependent, so
can only be done by gcc. Thus this patch which generalizes
the eh.selector intrinsic so that it can handle all possible
kinds of action table entries (before it didn't accomodate
cleanups): now 0 indicates a cleanup, and filters have to be
specified using the number of type infos plus one rather than
the number of type infos. Related gcc patches will cause
Ada to pass a cleanup (0) to force the selector to always
fire, while C++ will use a C++ catch-all (null).
llvm-svn: 41484
Coping 100MB array (after a warmup) shows that glibc 2.6.1 implementation on
x86-64 (core 2) is 30% faster (from 0.270917s to 0.188079s)
llvm-svn: 41479
Chris Lattner