Remove multiplication with inv

code/matrix_multiplication.h

@@ -132,7 +132,6 @@ void matrixMultiplyAVX2b(const std::complex<float>* a,
   __m256i a_4_ind = _mm256_set_epi32(7, 7, 7, 7, 3, 3, 3, 3);
   __m256i b_4_ind = _mm256_set_epi32(6, 7, 4, 5, 6, 7, 4, 5);
 
-  __m256 inv = _mm256_set_ps(1., -1., 1., -1., 1., -1., 1., -1.);
   __m256 a_1 = _mm256_permutevar8x32_ps(a_m, a_1_ind);
   __m256 b_1 = _mm256_permutevar8x32_ps(b_m, b_1_ind);
 
@@ -140,16 +139,14 @@ void matrixMultiplyAVX2b(const std::complex<float>* a,
   __m256 b_2 = _mm256_permutevar8x32_ps(b_m, b_2_ind);
 
   __m256 a_3 = _mm256_permutevar8x32_ps(a_m, a_3_ind);
-  __m256 a_3_inv = _mm256_mul_ps(a_3, inv);
   __m256 b_3 = _mm256_permutevar8x32_ps(b_m, b_3_ind);
 
   __m256 a_4 = _mm256_permutevar8x32_ps(a_m, a_4_ind);
-  __m256 a_4_inv = _mm256_mul_ps(a_4, inv);
   __m256 b_4 = _mm256_permutevar8x32_ps(b_m, b_4_ind);
 
   __m256 c_m = _mm256_mul_ps(a_1, b_1);
-  c_m = _mm256_fmadd_ps(a_2, b_2, c_m);
-  c_m = _mm256_fmadd_ps(a_3_inv, b_3, c_m);
-  c_m = _mm256_fmadd_ps(a_4_inv, b_4, c_m);
+  c_m = _mm256_add_ps(_mm256_mul_ps(a_2, b_2), c_m);
+  c_m = _mm256_addsub_ps(c_m, _mm256_mul_ps(a_3, b_3));
+  c_m = _mm256_addsub_ps(c_m, _mm256_mul_ps(a_4, b_4));
   _mm256_store_ps(c_ptr, c_m);
   }
\ No newline at end of file