Code gen of decomposed quant

kernel_cpp_0 = async_compile.cpp('''
#include <ATen/record_function.h>
#include "/tmp/torchinductor_root/dm/cdmaihqxwe73zkb3he2zizktpq5uujetg2db26c3r4lgsmlx3b4c.h"
extern "C" void kernel(const float* __restrict__ in_ptr0,
                       const float* __restrict__ in_ptr1,
                       const long* __restrict__ in_ptr2,
                       unsigned char* __restrict__ out_ptr0)
{
    RECORD_FUNCTION("graph_1_kernel_cpp_0", c10::ArrayRef<c10::IValue>({}));
    #pragma omp parallel num_threads(28)
    {
        {
            #pragma omp for 
            for(long i0=0; i0<17461248; i0+=1)
            {
                auto tmp2 = in_ptr0[i0];
                auto tmp3 = in_ptr1[0];
                auto tmp6 = in_ptr2[0];
                auto tmp0 = static_cast<float>(0);
                auto tmp1 = static_cast<float>(127);
                auto tmp4 = tmp2 / tmp3;
                auto tmp5 = std::nearbyint(tmp4);
                auto tmp7 = static_cast<float>(tmp6);
                auto tmp8 = tmp5 + tmp7;
                auto tmp9 = (tmp8 != tmp8) ? tmp8 : std::min(tmp1, tmp8);
                auto tmp10 = (tmp9 != tmp9) ? tmp9 : std::max(tmp0, tmp9);
                auto tmp11 = static_cast<unsigned char>(tmp10);
                out_ptr0[i0] = tmp11;
            }
        }
    }
}
''')