CUDA host/device calling

[hrehfeld]

glm\glm\detail\func_integer.inl(202): warning : calling a __host__ function from a __host__ __device__ function is not allowed

[Groovounet]

What version?

On master, line 202 is in the middle of nowhere.

[hrehfeld]

I pulled yesterday:

ext\glm\glm\detail>git rev-parse HEAD
e8fbcf76ddcee6862165a37b602df4aabbba5f63

GLM_FUNC_QUALIFIER uint uaddCarry(uint const & x, uint const & y, uint & Carry)
{
    uint64 const Value64(static_cast<uint64>(x) + static_cast<uint64>(y));
    uint64 const Max32(static_cast<uint64>(std::numeric_limits<uint>::max())); //here
    Carry = Value64 > Max32 ? 1 : 0;
    return static_cast<uint32>(Value64 % (Max32 + static_cast<uint64>(1)));
}

[Groovounet]

It looks like that CUDA doesn't fully support std::numeric_limits. :/

[ChRis6]

on 0.9.6.0 I'm getting:
glm/detail/func_integer.inl(206): warning: calling a constexpr host function("max") from a host device function("uaddCarry") is not allowed

and it gives seg fault.It was fine on 0.9.5.4

[Groovounet]

Ah, I have been handwavy on the previous fix. This second commit should resolve this issue.

This fix is available for in GLM 0.9.6 branch for GLM 0.9.6.1.

Thanks for contributing,
Christophe

[hrehfeld]

With #499

 glm\glm\gtc\constants.inl(40): warning : calling a __host__ function from a __host__ __device__ function is not allowed

from this code:

if (idealAngle <= ::glm::epsilon<float>())

and yes, these device annotations are really annoying in CUDA, sorry :/