static uint8x16_t xtime(uint8x16_t x)
{
uint8x16_t y = vshlq_n_u8(x,1);
x = vshrq_n_u8(x,7);
uint8x16_t n27 = vmovq_n_u8(0x1b);
x = vmulq_u8(x,n27);
x = veorq_u8(x,y);
return x;
}
void ar_vmul_u8_neon(uint8_t* res,
const uint8_t* a,
const uint8_t* b,
uint32_t n)
{
uint8x16_t a_loaded;
uint8x16_t b_loaded;
uint8x16_t res_loaded;
for (uint32_t i = 0; i < n; i += 16) {
a_loaded = vld1q_u8(&(a[i]));
b_loaded = vld1q_u8(&(b[i]));
res_loaded = vmulq_u8(a_loaded, b_loaded);
vst1q_u8(&(res[i]),res_loaded);
}
}
/* u8x16 mv mul */
void mw_neon_mv_mul_u8x16(unsigned char * A, int Row, int T, unsigned char * B, unsigned char * C)
{
int i = 0;
int k = 0;
uint8x16_t neon_b, neon_c;
uint8x16_t neon_a0, neon_a1, neon_a2, neon_a3, neon_a4, neon_a5, neon_a6, neon_a7;
uint8x16_t neon_a8, neon_a9, neon_a10, neon_a11, neon_a12, neon_a13, neon_a14, neon_a15;
uint8x16_t neon_b0, neon_b1, neon_b2, neon_b3, neon_b4, neon_b5, neon_b6, neon_b7;
uint8x16_t neon_b8, neon_b9, neon_b10, neon_b11, neon_b12, neon_b13, neon_b14, neon_b15;
for (i = 0; i < Row; i+=16)
{
neon_c = vmovq_n_u8(0);
for (k = 0; k < T; k+=16)
{
int j = k * T + i;
neon_a0 = vld1q_u8(A + j);
j+=Row;
neon_a1 = vld1q_u8(A + j);
j+=Row;
neon_a2 = vld1q_u8(A + j);
j+=Row;
neon_a3 = vld1q_u8(A + j);
j+=Row;
neon_a4 = vld1q_u8(A + j);
j+=Row;
neon_a5 = vld1q_u8(A + j);
j+=Row;
neon_a6 = vld1q_u8(A + j);
j+=Row;
neon_a7 = vld1q_u8(A + j);
j+=Row;
neon_a8 = vld1q_u8(A + j);
j+=Row;
neon_a9 = vld1q_u8(A + j);
j+=Row;
neon_a10 = vld1q_u8(A + j);
j+=Row;
neon_a11 = vld1q_u8(A + j);
j+=Row;
neon_a12 = vld1q_u8(A + j);
j+=Row;
neon_a13 = vld1q_u8(A + j);
j+=Row;
neon_a14 = vld1q_u8(A + j);
j+=Row;
neon_a15 = vld1q_u8(A + j);
neon_b = vld1q_u8(B + k);
neon_b0 = vdupq_n_u8(vgetq_lane_u8(neon_b, 0));
neon_b1 = vdupq_n_u8(vgetq_lane_u8(neon_b, 1));
neon_b2 = vdupq_n_u8(vgetq_lane_u8(neon_b, 2));
neon_b3 = vdupq_n_u8(vgetq_lane_u8(neon_b, 3));
neon_b4 = vdupq_n_u8(vgetq_lane_u8(neon_b, 4));
neon_b5 = vdupq_n_u8(vgetq_lane_u8(neon_b, 5));
neon_b6 = vdupq_n_u8(vgetq_lane_u8(neon_b, 6));
neon_b7 = vdupq_n_u8(vgetq_lane_u8(neon_b, 7));
neon_b8 = vdupq_n_u8(vgetq_lane_u8(neon_b, 8));
neon_b9 = vdupq_n_u8(vgetq_lane_u8(neon_b, 9));
neon_b10 = vdupq_n_u8(vgetq_lane_u8(neon_b, 10));
neon_b11 = vdupq_n_u8(vgetq_lane_u8(neon_b, 11));
neon_b12 = vdupq_n_u8(vgetq_lane_u8(neon_b, 12));
neon_b13 = vdupq_n_u8(vgetq_lane_u8(neon_b, 13));
neon_b14 = vdupq_n_u8(vgetq_lane_u8(neon_b, 14));
neon_b15 = vdupq_n_u8(vgetq_lane_u8(neon_b, 15));
neon_c = vaddq_u8(vmulq_u8(neon_a0, neon_b0), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a1, neon_b1), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a2, neon_b2), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a3, neon_b3), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a4, neon_b4), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a5, neon_b5), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a6, neon_b6), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a7, neon_b7), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a8, neon_b8), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a9, neon_b9), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a10, neon_b10), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a11, neon_b11), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a12, neon_b12), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a13, neon_b13), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a14, neon_b14), neon_c);
neon_c = vaddq_u8(vmulq_u8(neon_a15, neon_b15), neon_c);
}
vst1q_u8(C + i, neon_c);
}
}
