static INLINE void idct4x4_16_kernel_bd10(const int32x4_t cospis,
int32x4_t *const a0,
int32x4_t *const a1,
int32x4_t *const a2,
int32x4_t *const a3) {
int32x4_t b0, b1, b2, b3;
transpose_s32_4x4(a0, a1, a2, a3);
b0 = vaddq_s32(*a0, *a2);
b1 = vsubq_s32(*a0, *a2);
b0 = vmulq_lane_s32(b0, vget_high_s32(cospis), 0);
b1 = vmulq_lane_s32(b1, vget_high_s32(cospis), 0);
b2 = vmulq_lane_s32(*a1, vget_high_s32(cospis), 1);
b3 = vmulq_lane_s32(*a1, vget_low_s32(cospis), 1);
b2 = vmlsq_lane_s32(b2, *a3, vget_low_s32(cospis), 1);
b3 = vmlaq_lane_s32(b3, *a3, vget_high_s32(cospis), 1);
b0 = vrshrq_n_s32(b0, DCT_CONST_BITS);
b1 = vrshrq_n_s32(b1, DCT_CONST_BITS);
b2 = vrshrq_n_s32(b2, DCT_CONST_BITS);
b3 = vrshrq_n_s32(b3, DCT_CONST_BITS);
*a0 = vaddq_s32(b0, b3);
*a1 = vaddq_s32(b1, b2);
*a2 = vsubq_s32(b1, b2);
*a3 = vsubq_s32(b0, b3);
}
void test_vmlsQ_lanes32 (void)
{
int32x4_t out_int32x4_t;
int32x4_t arg0_int32x4_t;
int32x4_t arg1_int32x4_t;
int32x2_t arg2_int32x2_t;
out_int32x4_t = vmlsq_lane_s32 (arg0_int32x4_t, arg1_int32x4_t, arg2_int32x2_t, 1);
}
static INLINE void iadst_butterfly_lane_1_0_bd10_neon(const int32x4_t in0,
const int32x4_t in1,
const int32x2_t c,
int32x4_t *const s0,
int32x4_t *const s1) {
const int32x4_t t0 = vmulq_lane_s32(in0, c, 1);
const int32x4_t t1 = vmulq_lane_s32(in0, c, 0);
*s0 = vmlaq_lane_s32(t0, in1, c, 0);
*s1 = vmlsq_lane_s32(t1, in1, c, 1);
}
int32x4_t test_vmlsq_lane_s32(int32x4_t a, int32x4_t b, int32x2_t v) {
// CHECK: test_vmlsq_lane_s32
return vmlsq_lane_s32(a, b, v, 1);
// CHECK: mls {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, {{v[0-9]+}}.s[1]
}