void kernel_1_16_simple(double* a, double* b, double* c, size_t size);
void kernel_1_16_fuseaware(double* a, double* b, double* c, size_t size);
void kernel_8_1_simple(double* a, double* b, double* c, size_t size);
void kernel_8_1_fuseaware(double* a, double* b, double* c, size_t size);
void kernel_1_8_vo(double* a, double* b, double* c, size_t size);