static inline void PreShiftW32toW16Neon(int32_t* inre,
int32_t* inim,
int16_t* outre,
int16_t* outim,
int32_t sh) {
int k;
int32x4_t sh32x4 = vdupq_n_s32(sh);
for (k = 0; k < FRAMESAMPLES/2; k += 16) {
int32x4x4_t inre32x4x4 = vld4q_s32(inre);
int32x4x4_t inim32x4x4 = vld4q_s32(inim);
inre += 16;
inim += 16;
inre32x4x4.val[0] = vrshlq_s32(inre32x4x4.val[0], sh32x4);
inre32x4x4.val[1] = vrshlq_s32(inre32x4x4.val[1], sh32x4);
inre32x4x4.val[2] = vrshlq_s32(inre32x4x4.val[2], sh32x4);
inre32x4x4.val[3] = vrshlq_s32(inre32x4x4.val[3], sh32x4);
inim32x4x4.val[0] = vrshlq_s32(inim32x4x4.val[0], sh32x4);
inim32x4x4.val[1] = vrshlq_s32(inim32x4x4.val[1], sh32x4);
inim32x4x4.val[2] = vrshlq_s32(inim32x4x4.val[2], sh32x4);
inim32x4x4.val[3] = vrshlq_s32(inim32x4x4.val[3], sh32x4);
int16x4x4_t outre16x4x4;
int16x4x4_t outim16x4x4;
outre16x4x4.val[0] = vmovn_s32(inre32x4x4.val[0]);
outre16x4x4.val[1] = vmovn_s32(inre32x4x4.val[1]);
outre16x4x4.val[2] = vmovn_s32(inre32x4x4.val[2]);
outre16x4x4.val[3] = vmovn_s32(inre32x4x4.val[3]);
outim16x4x4.val[0] = vmovn_s32(inim32x4x4.val[0]);
outim16x4x4.val[1] = vmovn_s32(inim32x4x4.val[1]);
outim16x4x4.val[2] = vmovn_s32(inim32x4x4.val[2]);
outim16x4x4.val[3] = vmovn_s32(inim32x4x4.val[3]);
vst4_s16(outre, outre16x4x4);
vst4_s16(outim, outim16x4x4);
outre += 16;
outim += 16;
}
}
inline void vst4(s16 * ptr, const int16x4x4_t & v) { return vst4_s16(ptr, v); }
