Compilation Bug: the cuda code generated from gelu example fails to compile without -std=c++11

[BurtonQin]

The example gelu CUDA does not work on my machine.
The reason is the generated CUDA code

l_0_3 = float_4{
erf(l_0_3.i_0),erf(l_0_3.i_1),erf(l_0_3.i_2),erf(l_0_3.i_3),};

can only be compiled with C++11 and later.

System Info:
Ubuntu 20.04
nvcc Build cuda_11.8.r11.8/compiler.31833905_0
gcc (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0

cargo run --example gelu --features cuda

generates the following error message:

thread 'main' panicked at crates/cubecl-cuda/src/compute/server.rs:327:17:
[Compilation Error] 
    default_program(36): error: type name is not allowed
    default_program(36): error: expected a ";"
    default_program(38): error: type name is not allowed
    default_program(38): error: expected a ";"
    default_program(40): error: type name is not allowed
    default_program(40): error: expected a ";"
    default_program(42): error: type name is not allowed
    default_program(42): error: expected a ";"
    default_program(44): error: type name is not allowed
    default_program(44): error: expected a ";"
    default_program(27): warning #177-D: variable "rank_2" was declared but never referenced
    default_program(30): warning #550-D: variable "l_0_2" was set but never used
    10 errors detected in the compilation of "default_program".
[Source]  
typedef unsigned char uint8;
typedef unsigned short uint16;
typedef unsigned int uint;
typedef unsigned long long int uint64;
typedef long long int int64;

struct __align__(16) float_4 {
    float i_0;
    float i_1;
    float i_2;
    float i_3;
};


extern "C" __global__ void kernel(
float_4 input_0[],float_4 output_0[],uint info[]
) {

    int3 absoluteIdx = make_int3(
        blockIdx.x * blockDim.x + threadIdx.x,
        blockIdx.y * blockDim.y + threadIdx.y,
        blockIdx.z * blockDim.z + threadIdx.z
    );

    uint idxGlobal = (absoluteIdx.z * gridDim.x * blockDim.x * gridDim.y * blockDim.y) + (absoluteIdx.y * gridDim.x * blockDim.x) + absoluteIdx.x;
uint rank = info[0];
uint rank_2 = rank * 2;
uint l_0_0;
bool l_0_1;
float_4 l_0_2;
float_4 l_0_3;
l_0_0 = info[(2 * 2 * info[0]) + 1] / 4;
l_0_1 = idxGlobal < l_0_0;
if (l_0_1) {
l_0_2 = input_0[idxGlobal];
l_0_3 = float_4{
l_0_2.i_0 / float(1.4142135), l_0_2.i_1 / float(1.4142135), l_0_2.i_2 / float(1.4142135), l_0_2.i_3 / float(1.4142135), };
l_0_3 = float_4{
erf(l_0_3.i_0),erf(l_0_3.i_1),erf(l_0_3.i_2),erf(l_0_3.i_3),};
l_0_3 = float_4{
l_0_3.i_0 + float(1.0), l_0_3.i_1 + float(1.0), l_0_3.i_2 + float(1.0), l_0_3.i_3 + float(1.0), };
l_0_3 = float_4{
l_0_2.i_0 * l_0_3.i_0, l_0_2.i_1 * l_0_3.i_1, l_0_2.i_2 * l_0_3.i_2, l_0_2.i_3 * l_0_3.i_3, };
l_0_3 = float_4{
l_0_3.i_0 / float(2.0), l_0_3.i_1 / float(2.0), l_0_3.i_2 / float(2.0), l_0_3.i_3 / float(2.0), };
output_0[idxGlobal] = l_0_3;
}

}
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

A possible fix is to append "-std=c++11" or "-std=c++14" to options at

cubecl/crates/cubecl-cuda/src/compute/server.rs

Line 312 in 370e5cf

let options = &[arch.as_str(), include_option.as_str()];

let options = &[arch.as_str(), include_option.as_str(), "-std=c++14"];