Merge pull request #9717 from alalek:fix_fma

modules/core/src/stat.cpp

@@ -4365,7 +4365,7 @@ float normL2Sqr_(const float* a, const float* b, int n)
     for( ; j <= n - 8; j += 8 )
     {
         __m256 t0 = _mm256_sub_ps(_mm256_loadu_ps(a + j), _mm256_loadu_ps(b + j));
-#ifdef CV_FMA3
+#if CV_FMA3
         d0 = _mm256_fmadd_ps(t0, t0, d0);
 #else
         d0 = _mm256_add_ps(d0, _mm256_mul_ps(t0, t0));

modules/dnn/src/layers/layers_common.simd.hpp

@@ -59,7 +59,7 @@ void fastGEMM( const float* aptr, size_t astep, const float* bptr,
 
 #if !defined(CV_CPU_OPTIMIZATION_DECLARATIONS_ONLY) && CV_AVX
 
-#if !CV_FMA // AVX workaround
+#if !CV_FMA3 // AVX workaround
 #undef _mm256_fmadd_ps
 #define _mm256_fmadd_ps(a, b, c) _mm256_add_ps(c, _mm256_mul_ps(a, b))
 #endif