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);
}
static INLINE void iadst_half_butterfly_bd10_neon(int32x4_t *const x,
const int32x2_t c) {
const int32x4_t sum = vaddq_s32(x[0], x[1]);
const int32x4_t sub = vsubq_s32(x[0], x[1]);
x[0] = vmulq_lane_s32(sum, c, 0);
x[1] = vmulq_lane_s32(sub, c, 0);
x[0] = vrshrq_n_s32(x[0], DCT_CONST_BITS);
x[1] = vrshrq_n_s32(x[1], DCT_CONST_BITS);
}
void test_vRshrQ_ns32 (void)
{
int32x4_t out_int32x4_t;
int32x4_t arg0_int32x4_t;
out_int32x4_t = vrshrq_n_s32 (arg0_int32x4_t, 1);
}
static INLINE int32x4_t
sub_dct_const_round_shift_low_8_bd10(const int32x4_t in0, const int32x4_t in1) {
const int32x4_t sub = vsubq_s32(in0, in1);
return vrshrq_n_s32(sub, DCT_CONST_BITS);
}
int32x4_t test_vrshrq_n_s32(int32x4_t in) {
// CHECK-LABEL: @test_vrshrq_n_s32
// CHECK: call <4 x i32> @llvm.arm64.neon.srshl.v4i32(<4 x i32> %in, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>)
return vrshrq_n_s32(in, 1);
}
