an invoke instead of after the invoke (in its block), which is invalid.

llvm-svn: 67139

llvm-svn: 67138

llvm-svn: 67137

inverted).

llvm-svn: 67135

it is not APInt clean, but even when it is it needs to be evaluated carefully
to determine whether it is actually profitable.

This fixes a crash on PR3806

llvm-svn: 67134

llvm-svn: 67133

Some architectures (like x86) don't require it.
This fixes bug 3779.

llvm-svn: 67132

-ccc-no-clang, -ccc-no-clang-cxx, -ccc-no-clang-cpp, -fsyntax-only).

llvm-svn: 67131

llvm-svn: 67130

llvm-svn: 67129

no functionality change.

llvm-svn: 67128

discovered as part of PR3817

llvm-svn: 67127

a type.  For example, if we have a typedef of a typeof of 
int, we strip all the say down to int.  This allows us to
simplify the getAs* methods, and is the first step towards
fixing PR3817

llvm-svn: 67126

driver has no corresponding option.

llvm-svn: 67125

into gcc:: namespace.

llvm-svn: 67120

we resolve the semantics of this function when GC is enabled.

llvm-svn: 67116

llvm-svn: 67112

llvm-svn: 67110

llvm-svn: 67109

llvm-svn: 67108

llvm-svn: 67107

host; the toolchain may differ based on command line arguments.

llvm-svn: 67106

llvm-svn: 67105

cleaner visitor framework.

Added a visitor for declarations, which is quite similar to the
visitor for statatements.

llvm-svn: 67104

 - Use for exceptional buffer conditions in raw_ostream:write to shave
   off a cycle or two.

 - Please rename if you have a better one.

llvm-svn: 67103

to pass to ToolChains, which may need Driver specific information (for
example, to form search paths).

llvm-svn: 67102

llvm-svn: 67101

to see if this is producing the expected code or not, I'm
not sure what the test was intended to check.

llvm-svn: 67099

llvm-svn: 67096

llvm-svn: 67094

size by the array amount as an i32 value instead of promoting from
i32 to i64 then doing the multiply.  Not doing this broke wrap-around
assumptions that the optimizers (validly) made.  The ultimate real
fix for this is to introduce i64 version of alloca and remove mallocinst.

This fixes PR3829

llvm-svn: 67093

quite as great as it sounds, because, while we can refer to the
enumerator values outside the template, e.g.,

  adder<long, 3, 4>::value

we can't yet refer to them with dependent names, so no Fibonacci
(yet). 

InstantiateClassTemplateSpecialization is getting messy; next commit
will put it into a less-ugly state.

llvm-svn: 67092

 - Also, normalize arch names a tad and stub out getToolChain
   implementations.

llvm-svn: 67091

llvm-svn: 67090

llvm-svn: 67089

temporary file, user provided name, derived name).

llvm-svn: 67088

llvm-svn: 67087

 - Slightly strange, but the idea is that the ArgList data structure
   is primarily a list of arguments; we want to allow clients to still
   add argument strings to an ArgList to avoid worrying about string
   lifetimes (or unnecessary string copying).

llvm-svn: 67086

llvm-svn: 67085

Revert inadvertent mis-fix of fneg.

llvm-svn: 67084