A High-Performance Parallel FDTD Method Enhanced by Using SSE Instruction Set

Algorithm 1

for(i = 0; i <= nx; i ++) {

 vHi_Coeff = _mm_load1_ps(& 𝐻_𝑖_Coeff);

 //load single float value to vector

 for(j = 0; j <= ny; j++) {

  vHj_Coeff = _mm_load1_ps(& 𝐻_𝑗_Coeff);

  //load single float value to vector

 𝑣𝐸_𝑧= ( _---𝑚128 *) 𝐸_𝑧[i][j];

 𝑣𝐻_𝑦= ( _---𝑚128 *) 𝐻_𝑦[i][j];

 𝑣𝐻_𝑥= ( _---𝑚128 *) 𝐻_𝑥[i][j];

 𝑣𝐻_𝑦_minus = ( _---𝑚128 *) 𝐻_𝑦[i-1][j];

 𝑣𝐻_𝑥_minus = ( _---𝑚128 *) 𝐻_𝑥[i][j-1];

 for(k = 0, vk = 0; k < nz; k += 4, vk ++) {

  vEk_Coeff = _mm_load1_ps(&Ek_Coeff);

  xmm0 = _mm_sub_ps( 𝑣𝐻_𝑥[vk], vHx_minus [vk]);

  xmm0 = _mm_mul_ps( 𝑣𝐻_𝑗_Coeff, xmm0);

  xmm1 = _mm_sub_ps( 𝑣𝐻_𝑦[vk], vHy_minus [vk] );

  xmm1 = _mm_mul_ps( 𝑣𝐻_𝑖_Coeff, xmm1 );

  xmm0 = _mm_sub_ps(xmm1, xmm0 );

  xmm1 = _mm_mul_ps(vEk [vk], vEk_Coeff);

  vEk [vk] = _mm_add_ps(xmm1, xmm0);

}}}