• State Accepted Answer
  • Locked Locked
  • Replies 3 replies
  • Subscribers 21 subscribers
  • Views 5544 views
  • Users 0 members are here
Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Cast uint32x4_t to uint8x8x2_t

DorJo

Is there a possibility to reinterpret cast an uint32x4_t into a uint8x8x2_t using intrinsics?

 

Sincerely

Parents
  • +1

    DorJo,

    As far as I am aware, there are no reinterprets that yield an NxMx2 vector. The following code does what I believe you are trying to achieve by going via a pair of uint8x8_t, and with ARMCC and ARMCLANG produces no intermediate moves, i.e. the output is just "VLD1.32 {d16,d17},[r1]", "VST2.8 {d16,d17},[r0]".

    #include <arm_neon.h>

void func(uint8_t *dst, const uint32_t *src)
{

  uint32x4_t in;
  uint8x8x2_t out;

  in = vld1q_u32(src);

  out.val[0] = vget_low_u8(vreinterpretq_u8_u32(in));
  out.val[1] = vget_high_u8(vreinterpretq_u8_u32(in));

  vst2_u8(dst, out);
}

    Best regards

    Simon.

Reply
  • +1

    DorJo,

    As far as I am aware, there are no reinterprets that yield an NxMx2 vector. The following code does what I believe you are trying to achieve by going via a pair of uint8x8_t, and with ARMCC and ARMCLANG produces no intermediate moves, i.e. the output is just "VLD1.32 {d16,d17},[r1]", "VST2.8 {d16,d17},[r0]".

    #include <arm_neon.h>

void func(uint8_t *dst, const uint32_t *src)
{

  uint32x4_t in;
  uint8x8x2_t out;

  in = vld1q_u32(src);

  out.val[0] = vget_low_u8(vreinterpretq_u8_u32(in));
  out.val[1] = vget_high_u8(vreinterpretq_u8_u32(in));

  vst2_u8(dst, out);
}

    Best regards

    Simon.

Children