void test_vmov_ns16 (void)
{
int16x4_t out_int16x4_t;
int16_t arg0_int16_t;
out_int16x4_t = vmov_n_s16 (arg0_int16_t);
}
void matMult(int16_t mat1[], int16_t mat2[], int32_t prod[matrix_size][matrix_size])
{
int output_size = 2 * matrix_size;
int l,k;
int16x4_t data1;
int32x4_t mac_output[output_size/4];
int32x4_t MAC_addvalue[output_size/4];
int16x4_t constant_value;
unsigned int index_input = 0;
unsigned int transfer_index = 0 ;
int32_t *pres_ver;
/* Allocate output */
pres_ver = malloc(output_size * output_size * sizeof(int32_t));
for(l = 0 ; l < matrix_size/4; l++)
{
MAC_addvalue[l] = vmovq_n_s32(0);
}
/* Perform the multiplication */
for(l = 0; l < matrix_size*matrix_size; l++)
{
constant_value = vmov_n_s16 (mat1[l]);
for(k = 0 ; k < matrix_size/4 ; k++)
{
data1 = vld1_s16 (&mat2[index_input]);
MAC4 (&MAC_addvalue[k], &constant_value, &data1,&mac_output[k]);
MAC_addvalue[k] = mac_output[k];
index_input +=4;
}
index_input+=output_size-matrix_size;
if ((l + 1) % matrix_size == 0 )
{
index_input = 0;
for(k = 0 ; k < matrix_size/4 ; k++)
{
vst1q_s32(&pres_ver[transfer_index],MAC_addvalue[k]);
transfer_index +=4;
}
transfer_index += output_size-matrix_size;
for(k = 0 ; k < matrix_size/4; k++)
{
MAC_addvalue[k] = vmovq_n_s32(0);
}
}
}
}
int16x4_t test_vmov_n_s16(int16_t v1) {
// CHECK: test_vmov_n_s16
return vmov_n_s16(v1);
// CHECK: dup {{v[0-9]+}}.4h, {{w[0-9]+}}
}
