/* Transpose bits within bytes. */
int64_t bshuf_trans_bit_byte_AVX(void* in, void* out, const size_t size,
const size_t elem_size) {
size_t ii, kk;
char* in_b = (char*) in;
char* out_b = (char*) out;
int32_t* out_i32;
size_t nbyte = elem_size * size;
int64_t count;
__m256i ymm;
int32_t bt;
for (ii = 0; ii + 31 < nbyte; ii += 32) {
ymm = _mm256_loadu_si256((__m256i *) &in_b[ii]);
for (kk = 0; kk < 8; kk++) {
bt = _mm256_movemask_epi8(ymm);
ymm = _mm256_slli_epi16(ymm, 1);
out_i32 = (int32_t*) &out_b[((7 - kk) * nbyte + ii) / 8];
*out_i32 = bt;
}
}
count = bshuf_trans_bit_byte_remainder(in, out, size, elem_size,
nbyte - nbyte % 32);
return count;
}
/* Transpose bits within bytes. */
int64_t bshuf_trans_bit_byte_sse2(void* in, void* out, const size_t size,
const size_t elem_size) {
char* in_b = (char*) in;
char* out_b = (char*) out;
uint16_t* out_ui16;
int64_t count;
size_t nbyte = elem_size * size;
__m128i xmm;
int32_t bt;
size_t ii, kk;
CHECK_MULT_EIGHT(nbyte);
for (ii = 0; ii + 15 < nbyte; ii += 16) {
xmm = _mm_loadu_si128((__m128i *) &in_b[ii]);
for (kk = 0; kk < 8; kk++) {
bt = _mm_movemask_epi8(xmm);
xmm = _mm_slli_epi16(xmm, 1);
out_ui16 = (uint16_t*) &out_b[((7 - kk) * nbyte + ii) / 8];
*out_ui16 = bt;
}
}
count = bshuf_trans_bit_byte_remainder(in, out, size, elem_size,
nbyte - nbyte % 16);
return count;
}
/* Transpose bits within bytes. */
int64_t bshuf_trans_bit_byte_scal(void* in, void* out, const size_t size,
const size_t elem_size) {
return bshuf_trans_bit_byte_remainder(in, out, size, elem_size, 0);
}
