C++ (Cpp) _mm256_max_epi32 Examples

Example #1

File: TransLut_avx2.cpp Project: EleonoreMizo/fmtconv

void	TransLut_FindIndexAvx2 <TransLut::MapperLin>::find_index (const TransLut::FloatIntMix val_arr [8], __m256i &index, __m256 &frac)
{
	assert (val_arr != 0);

	const __m256   scale     = _mm256_set1_ps (1 << LINLUT_RES_L2);
	const __m256i  offset    =
		_mm256_set1_epi32 (-LINLUT_MIN_F * (1 << LINLUT_RES_L2));
	const __m256i  val_min   = _mm256_setzero_si256 ();
	const __m256i  val_max   = _mm256_set1_epi32 (LINLUT_SIZE_F - 2);

	const __m256   v         =
		_mm256_load_ps (reinterpret_cast <const float *> (val_arr));
	const __m256   val_scl   = _mm256_mul_ps (v, scale);
	const __m256i  index_raw = _mm256_cvtps_epi32 (val_scl);
	__m256i        index_tmp = _mm256_add_epi32 (index_raw, offset);
	index_tmp = _mm256_min_epi32 (index_tmp, val_max);
	index     = _mm256_max_epi32 (index_tmp, val_min);
	frac      = _mm256_sub_ps (val_scl, _mm256_cvtepi32_ps (index_raw));
}

Example #2

File: av1_highbd_quantize_avx2.c Project: jfiguinha/Regards

static INLINE void quantize(const __m256i *qp, __m256i *c,
                            const int16_t *iscan_ptr, int log_scale,
                            tran_low_t *qcoeff, tran_low_t *dqcoeff,
                            __m256i *eob) {
  const __m256i abs_coeff = _mm256_abs_epi32(*c);
  __m256i q = _mm256_add_epi32(abs_coeff, qp[0]);

  __m256i q_lo = _mm256_mul_epi32(q, qp[1]);
  __m256i q_hi = _mm256_srli_epi64(q, 32);
  const __m256i qp_hi = _mm256_srli_epi64(qp[1], 32);
  q_hi = _mm256_mul_epi32(q_hi, qp_hi);
  q_lo = _mm256_srli_epi64(q_lo, 16 - log_scale);
  q_hi = _mm256_srli_epi64(q_hi, 16 - log_scale);
  q_hi = _mm256_slli_epi64(q_hi, 32);
  q = _mm256_or_si256(q_lo, q_hi);
  const __m256i abs_s = _mm256_slli_epi32(abs_coeff, 1 + log_scale);
  const __m256i mask = _mm256_cmpgt_epi32(qp[2], abs_s);
  q = _mm256_andnot_si256(mask, q);

  __m256i dq = _mm256_mullo_epi32(q, qp[2]);
  dq = _mm256_srai_epi32(dq, log_scale);
  q = _mm256_sign_epi32(q, *c);
  dq = _mm256_sign_epi32(dq, *c);

  _mm256_storeu_si256((__m256i *)qcoeff, q);
  _mm256_storeu_si256((__m256i *)dqcoeff, dq);

  const __m128i isc = _mm_loadu_si128((const __m128i *)iscan_ptr);
  const __m128i zr = _mm_setzero_si128();
  const __m128i lo = _mm_unpacklo_epi16(isc, zr);
  const __m128i hi = _mm_unpackhi_epi16(isc, zr);
  const __m256i iscan =
      _mm256_insertf128_si256(_mm256_castsi128_si256(lo), hi, 1);

  const __m256i zero = _mm256_setzero_si256();
  const __m256i zc = _mm256_cmpeq_epi32(dq, zero);
  const __m256i nz = _mm256_cmpeq_epi32(zc, zero);
  __m256i cur_eob = _mm256_sub_epi32(iscan, nz);
  cur_eob = _mm256_and_si256(cur_eob, nz);
  *eob = _mm256_max_epi32(cur_eob, *eob);
}

Example #3

File: avx2-builtins.c Project: mgranberry/clang

__m256i test_mm256_max_epi32(__m256i a, __m256i b) {
  // CHECK: @llvm.x86.avx2.pmaxs.d
  return _mm256_max_epi32(a, b);
}

Example #4

File: avx2-builtins.c Project: autodesk-forks/llvm-clang

__m256i test_mm256_max_epi32(__m256i a, __m256i b) {
  // CHECK-LABEL: test_mm256_max_epi32
  // CHECK:       [[CMP:%.*]] = icmp sgt <8 x i32> [[X:%.*]], [[Y:%.*]]
  // CHECK-NEXT:  select <8 x i1> [[CMP]], <8 x i32> [[X]], <8 x i32> [[Y]]
  return _mm256_max_epi32(a, b);
}

Example #5

File: sw_striped_avx2_256_32.c Project: dikaiosune/parasail-sys

static inline int32_t _mm256_hmax_epi32_rpl(__m256i a) {
    a = _mm256_max_epi32(a, _mm256_permute2x128_si256(a, a, _MM_SHUFFLE(0,0,0,0)));
    a = _mm256_max_epi32(a, _mm256_slli_si256(a, 8));
    a = _mm256_max_epi32(a, _mm256_slli_si256(a, 4));
    return _mm256_extract_epi32_rpl(a, 7);
}

Example #6

File: avx2-vpmaxsd-1.c Project: 0day-ci/gcc

void extern
avx2_test (void)
{
  x = _mm256_max_epi32 (x, x);
}