[LLVMbugs] [Bug 21398] New: Homegeneous aggregates with C++ base classes passed incorrectly on PPC64, ARM, and Aarch64

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=21398

            Bug ID: 21398
           Summary: Homegeneous aggregates with C++ base classes passed
                    incorrectly on PPC64, ARM, and Aarch64
           Product: clang
           Version: unspecified
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: LLVM Codegen
          Assignee: unassignedclangbugs at nondot.org
          Reporter: rnk at google.com
                CC: david.majnemer at gmail.com, james.molloy at arm.com,
                    llvmbugs at cs.uiuc.edu, t.p.northover at gmail.com
    Classification: Unclassified

Consider:

struct Base {
  int x;
};
struct HVA : Base {
  double y, z;
};
void f(HVA x);
extern HVA h;
void g() {
  f(h);
}

On ARM, Clang will pass this in r0, d0, and d1, but gcc will pass in r0-r5.

AArch64 uses the same code and PPC64 looks like it is duplicated from the same
routine, so I'm sure they have similar issues.

Back on 3.4 when we used to do record flattening, we would completely screw
this up and would only pass the two doubles. Currently we just misclassify, but
we pass all the data.

IMO we should refactor the homogeneous aggregate calculation code to accept an
architecture. There's more complexity in walking Clang types than there is in
the target-specific logic.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20141027/582fdf37/attachment.html>