FMA

Hi,

I use Intel 17.0.4 compiler and Intel Xeon E5 2697 V4 (Broadwell) processor. I know that this processor supports fused multiply add instruction.

For this line of code:

      yy +=  (A[i] * B[i]);

If I convert the C++ code to assembly I can see vfmadd231pd 16(%rdx,%r11,8), %xmm6, %xmm1

However, when I use vYY = _mm256_fmadd_pd (vA, vB, vYY) in the C++ code, the compiler uses add and multiply vector instructions only: 

vmulpd    (%r15,%rsi,8), %ymm4, %ymm5              

vaddpd    %ymm1, %ymm5, %ymm1         

Is there any explanation for this ?

Thanks,