On Windows x86_64, class TypeInfo_Struct contains 2 additional fields
(m_arg1/m_arg2) which are used for the X86_64 System V ABI varargs
implementation. They are not present on any other os.
This commit changes an assertion which did not check for the os and therefore
broke on 64bit non-86 systems like ppc64.
Referencing alloca'd memory in tail calls is invalid IR. This was
not caught by the verifier, but produced misoptimizations due to
wrong alias analysis results.
This also allows us to enable the assert in IrType::IrType.
Unfortunately, this is mostly a "peace of mind" commit, there
doesn't seem to have been a bug actually caused by the transitory
duplicate IrTypePointer/IrTypeStruct instances.
The remaining xyz2llvm static methods are not exactly pretty,
they should probably just be folded into get.
I chose to fix the problem this way because it increases uniformity
between 'this' and normal explicit parameters. Another possibility
would be to just change the type determinatin code in
DtoCreateNestedContextType to not expect the value to be already
present, because it doesn't need it when isVthisPtr is true anyway.
GitHub: Fixes#217.
Just executing toConstElem speculatively in
AssocArrayLiteralExp::toElem probably isn't the best idea
anyway, I would not be surprised if there are other similar bugs.
GitHub: Fixes#248.
Adding pragma llvm_inline_ir.
Improved the error messages.
Append "ret void" when the return type is void
Improved the error message
in case when the string passed as llvm inline ir isn't valid llvm
assembly language.
LLVM 3.2 fix.
Add attribute AlwaysInline inside DtoInlineIRFunction.
Always generate a body for llvm_inline_ir
Also, always make llvm_inline_ir functions linkonce_odr. Because
the body is always generated when a module uses a llvm_inline_ir
function, the fact that the linker removes the function shouldn't
cause problems.
-disable-simplify-drtcalls was only intended to disable our
custom druntime call simplication pass for debugging purposes.
The new flag controls C runtime library calls optimizations as
performed by the LLVM pass, just as the corresponding LLVM tool
flags do.
Previously, set_param_attrs would overwrite any pre-existing
attributes, which is problematic, as per-function attributes are
also stored in a slot in that attribute list. This for example
lead to "noinline" being dropped for functions with inline asm.
