fail to build _mm_cvtsd_si64

[pythoneer]

/// Convert the lower double-precision (64-bit) floating-point element in a to a 64-bit integer.
#[inline(always)]
#[target_feature = "+sse2"]
#[cfg_attr(test, assert_instr(cvtsd2si))]
pub unsafe fn _mm_cvtsd_si64(a: f64x2) -> i64 {
    cvtsd2si64(a)
}

....

    #[link_name = "llvm.x86.sse2.cvtsd2si64"]
    fn cvtsd2si64(a: f64x2) -> i64;

fails to build with

ExpandIntegerResult #0: t12: i64 = llvm.x86.sse2.cvtsd2si64 TargetConstant:i32<5716>, t10

Do not know how to expand the result of this operator!

UNREACHABLE executed at /checkout/src/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp:1279!

error: Could not compile `stdsimd`.

https://travis-ci.org/rust-lang-nursery/stdsimd/jobs/281644652

EDIT:

looks like _mm_cvttsd_si64 is also failing

/// Convert the lower double-precision (64-bit) floating-point element in `a` to a 64-bit integer
/// with truncation.
#[inline(always)]
#[target_feature = "+sse2"]
#[cfg_attr(test, assert_instr(cvttsd2si))]
pub unsafe fn _mm_cvttsd_si64(a: f64x2) -> i64 {
    cvttsd2si64(a)
}

ExpandIntegerResult #0: t12: i64 = llvm.x86.sse2.cvttsd2si64 TargetConstant:i32<5724>, t10

Do not know how to expand the result of this operator!

UNREACHABLE executed at /checkout/src/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp:1279!

error: Could not compile `stdsimd`.

https://travis-ci.org/rust-lang-nursery/stdsimd/jobs/281653879

[gwenn]

In clang builtins-x86.c file, they are wrapped into a condition:

#ifdef USE_64
  tmp_LLi = __builtin_ia32_cvtsd2si64(tmp_V2d);
  tmp_LLi = __builtin_ia32_cvttsd2si64(tmp_V2d);
#endif

That would explain why only the i686 build fails...

[alexcrichton]

I think these may just need to be excluded on i686? We had to do that for some other bit-related intrinsics as well

[pythoneer]

thx @gwenn ... ok, is there something specific i should do for documenting this?

[alexcrichton]

@pythoneer nah I think it's fine to just exclude it on i686