[UBSan][MIPS] Adding support for MIPS64

[slthakur]

Summary

No MIPS64 architecture dependant code.
Patch dependant on http://reviews.llvm.org/D6402

[kcc]

Maybe we should fix this in clang?

[samsonov]

+1. We should fix really fix this macro definition in Clang.

[dsanders]

Sagar has shown me the error he gets and (if it's as easy to fix as it looks) I'd rather fix the 'Cannot select ... i64,glue = addc ...' error by adding the necessary Pat's than remove the SIZEOF_INT128 macro.

In MipsInstrInfo.td there's this code to support 64-bit ints on MIPS32:

def : MipsPat<(subc GPR32:$lhs, GPR32:$rhs),
               (SUBu GPR32:$lhs, GPR32:$rhs)>;
let AdditionalPredicates = [NotDSP] in {
   def : MipsPat<(addc GPR32:$lhs, GPR32:$rhs),
                 (ADDu GPR32:$lhs, GPR32:$rhs)>;
   def : MipsPat<(addc  GPR32:$src, immSExt16:$imm),
                 (ADDiu GPR32:$src, imm:$imm)>;
}

We don't seem to have a GPR64 version of these three Pat's. I think it might be as simple as copying this into Mips64InstrInfo.td and changing the new copy from GPR32, ADD, and SUB, to GPR64, DADD, and DSUB respectively but I haven't had chance to try this change yet.

For the sake of completeness, the test case he showed me was:

int main ()
{
  __int128 a, b;
  a = 10;
  b = 20;
  a = a + b;
  return 0;
}

and the full error is:

fatal error: error in backend: Cannot select: 0x489e630: i64,glue = addc 0x489e510, 0x489bf98 [ORD=10] [ID=22]
    0x489e510: i64,ch = load 0x489cda8, 0x489cc88, 0x489c028<LD8[%a+8]> [ORD=8] [ID=20]
      0x489cc88: i64 = or 0x489c1d8, 0x489c9b8 [ORD=6] [ID=12]
        0x489c1d8: i64 = FrameIndex<1> [ID=4]
        0x489c9b8: i64 = Constant<8> [ID=10]
      0x489c028: i64 = undef [ID=3]
    0x489bf98: i64 = Constant<20> [ID=7]

[slthakur]

Patch for removing SIZEOF_INT128 in clang has been submitted [1].
No MIPS64 architecture dependant code required for UBSan 64-bit

[1] http://reviews.llvm.org/D6402

[samsonov]

LGTM. Please submit if all test cases are really passing.

[slthakur]

Hi @samsonov,

Test case TypeCheck/misaligned.cpp uses a 48-bit address in one of its checks.
To make this test case pass for mips64 I need to use a 40-bit address.
Should I submit this change along with this patch or separately ?

[samsonov]

Let's submit it in a separate change.

[slthakur]

Hi @samsonov,

In TypeCheck/misaligned.cpp can I mark CHECK-WILD as expected failure for mips64 and misp64el just like arm has done?
Or is there any way to separate CHECK_WILD for mips64 architecture?

[samsonov]

I don't think you can disable only CHECK-WILD under given architecture. So you'd have to either XFAIL the whole test, or factor CHECK-WILD part of it into a separate test case.

[sdkie]

closed by rL224239

cmake/config-ix.cmake

 # LSan common files should be available on all architectures supported
 # by other sanitizers (even if they build into dummy object files).
 filter_available_targets(LSAN_COMMON_SUPPORTED_ARCH
   ${SANITIZER_COMMON_SUPPORTED_ARCH})
 filter_available_targets(MSAN_SUPPORTED_ARCH x86_64 mips64 mips64el)
 filter_available_targets(PROFILE_SUPPORTED_ARCH x86_64 i386 i686 arm mips mips64
   mipsel mips64el aarch64 powerpc64 powerpc64le)
 filter_available_targets(TSAN_SUPPORTED_ARCH x86_64)
-filter_available_targets(UBSAN_SUPPORTED_ARCH x86_64 i386 i686 arm aarch64 mips mipsel)
+filter_available_targets(UBSAN_SUPPORTED_ARCH x86_64 i386 i686 arm aarch64 mips mipsel mips64 mips64el)
 
 if(ANDROID)
   set(OS_NAME "Android")
 else()
   set(OS_NAME "${CMAKE_SYSTEM_NAME}")
 endif()
 
 if (SANITIZER_COMMON_SUPPORTED_ARCH AND NOT LLVM_USE_SANITIZER AND