Rewrite SIMD code by using new Universal Intrinsic API.

pull/23885/head
Liutong HAN 1 year ago
parent c982be3924
commit d17507052e
  1. 66
      modules/core/src/mean.simd.hpp

@ -24,7 +24,7 @@ struct SumSqr_SIMD
}
};
#if CV_SIMD
#if CV_SIMD || CV_SIMD_SCALABLE
template <>
struct SumSqr_SIMD<uchar, int, int>
@ -39,37 +39,37 @@ struct SumSqr_SIMD<uchar, int, int>
v_int32 v_sum = vx_setzero_s32();
v_int32 v_sqsum = vx_setzero_s32();
const int len0 = len & -v_uint8::nlanes;
const int len0 = len & -VTraits<v_uint8>::vlanes();
while(x < len0)
{
const int len_tmp = min(x + 256*v_uint16::nlanes, len0);
const int len_tmp = min(x + 256*VTraits<v_uint16>::vlanes(), len0);
v_uint16 v_sum16 = vx_setzero_u16();
for ( ; x < len_tmp; x += v_uint8::nlanes)
for ( ; x < len_tmp; x += VTraits<v_uint8>::vlanes())
{
v_uint16 v_src0 = vx_load_expand(src0 + x);
v_uint16 v_src1 = vx_load_expand(src0 + x + v_uint16::nlanes);
v_sum16 += v_src0 + v_src1;
v_uint16 v_src1 = vx_load_expand(src0 + x + VTraits<v_uint16>::vlanes());
v_sum16 = v_add(v_sum16, v_add(v_src0, v_src1));
v_int16 v_tmp0, v_tmp1;
v_zip(v_reinterpret_as_s16(v_src0), v_reinterpret_as_s16(v_src1), v_tmp0, v_tmp1);
v_sqsum += v_dotprod(v_tmp0, v_tmp0) + v_dotprod(v_tmp1, v_tmp1);
v_sqsum = v_add(v_sqsum, v_add(v_dotprod(v_tmp0, v_tmp0), v_dotprod(v_tmp1, v_tmp1)));
}
v_uint32 v_half0, v_half1;
v_expand(v_sum16, v_half0, v_half1);
v_sum += v_reinterpret_as_s32(v_half0 + v_half1);
v_sum = v_add(v_sum, v_reinterpret_as_s32(v_add(v_half0, v_half1)));
}
if (x <= len - v_uint16::nlanes)
if (x <= len - VTraits<v_uint16>::vlanes())
{
v_uint16 v_src = vx_load_expand(src0 + x);
v_uint16 v_half = v_combine_high(v_src, v_src);
v_uint32 v_tmp0, v_tmp1;
v_expand(v_src + v_half, v_tmp0, v_tmp1);
v_sum += v_reinterpret_as_s32(v_tmp0);
v_expand(v_add(v_src, v_half), v_tmp0, v_tmp1);
v_sum = v_add(v_sum, v_reinterpret_as_s32(v_tmp0));
v_int16 v_tmp2, v_tmp3;
v_zip(v_reinterpret_as_s16(v_src), v_reinterpret_as_s16(v_half), v_tmp2, v_tmp3);
v_sqsum += v_dotprod(v_tmp2, v_tmp2);
x += v_uint16::nlanes;
v_sqsum = v_add(v_sqsum, v_dotprod(v_tmp2, v_tmp2));
x += VTraits<v_uint16>::vlanes();
}
if (cn == 1)
@ -79,13 +79,13 @@ struct SumSqr_SIMD<uchar, int, int>
}
else
{
int CV_DECL_ALIGNED(CV_SIMD_WIDTH) ar[2 * v_int32::nlanes];
int CV_DECL_ALIGNED(CV_SIMD_WIDTH) ar[2 * VTraits<v_int32>::max_nlanes];
v_store(ar, v_sum);
v_store(ar + v_int32::nlanes, v_sqsum);
for (int i = 0; i < v_int32::nlanes; ++i)
v_store(ar + VTraits<v_int32>::vlanes(), v_sqsum);
for (int i = 0; i < VTraits<v_int32>::vlanes(); ++i)
{
sum[i % cn] += ar[i];
sqsum[i % cn] += ar[v_int32::nlanes + i];
sqsum[i % cn] += ar[VTraits<v_int32>::vlanes() + i];
}
}
v_cleanup();
@ -106,37 +106,37 @@ struct SumSqr_SIMD<schar, int, int>
v_int32 v_sum = vx_setzero_s32();
v_int32 v_sqsum = vx_setzero_s32();
const int len0 = len & -v_int8::nlanes;
const int len0 = len & -VTraits<v_int8>::vlanes();
while (x < len0)
{
const int len_tmp = min(x + 256 * v_int16::nlanes, len0);
const int len_tmp = min(x + 256 * VTraits<v_int16>::vlanes(), len0);
v_int16 v_sum16 = vx_setzero_s16();
for (; x < len_tmp; x += v_int8::nlanes)
for (; x < len_tmp; x += VTraits<v_int8>::vlanes())
{
v_int16 v_src0 = vx_load_expand(src0 + x);
v_int16 v_src1 = vx_load_expand(src0 + x + v_int16::nlanes);
v_sum16 += v_src0 + v_src1;
v_int16 v_src1 = vx_load_expand(src0 + x + VTraits<v_int16>::vlanes());
v_sum16 = v_add(v_sum16, v_add(v_src0, v_src1));
v_int16 v_tmp0, v_tmp1;
v_zip(v_src0, v_src1, v_tmp0, v_tmp1);
v_sqsum += v_dotprod(v_tmp0, v_tmp0) + v_dotprod(v_tmp1, v_tmp1);
v_sqsum = v_add(v_sqsum, v_add(v_dotprod(v_tmp0, v_tmp0), v_dotprod(v_tmp1, v_tmp1)));
}
v_int32 v_half0, v_half1;
v_expand(v_sum16, v_half0, v_half1);
v_sum += v_half0 + v_half1;
v_sum = v_add(v_sum, v_add(v_half0, v_half1));
}
if (x <= len - v_int16::nlanes)
if (x <= len - VTraits<v_int16>::vlanes())
{
v_int16 v_src = vx_load_expand(src0 + x);
v_int16 v_half = v_combine_high(v_src, v_src);
v_int32 v_tmp0, v_tmp1;
v_expand(v_src + v_half, v_tmp0, v_tmp1);
v_sum += v_tmp0;
v_expand(v_add(v_src, v_half), v_tmp0, v_tmp1);
v_sum = v_add(v_sum, v_tmp0);
v_int16 v_tmp2, v_tmp3;
v_zip(v_src, v_half, v_tmp2, v_tmp3);
v_sqsum += v_dotprod(v_tmp2, v_tmp2);
x += v_int16::nlanes;
v_sqsum = v_add(v_sqsum, v_dotprod(v_tmp2, v_tmp2));
x += VTraits<v_int16>::vlanes();
}
if (cn == 1)
@ -146,13 +146,13 @@ struct SumSqr_SIMD<schar, int, int>
}
else
{
int CV_DECL_ALIGNED(CV_SIMD_WIDTH) ar[2 * v_int32::nlanes];
int CV_DECL_ALIGNED(CV_SIMD_WIDTH) ar[2 * VTraits<v_int32>::max_nlanes];
v_store(ar, v_sum);
v_store(ar + v_int32::nlanes, v_sqsum);
for (int i = 0; i < v_int32::nlanes; ++i)
v_store(ar + VTraits<v_int32>::vlanes(), v_sqsum);
for (int i = 0; i < VTraits<v_int32>::vlanes(); ++i)
{
sum[i % cn] += ar[i];
sqsum[i % cn] += ar[v_int32::nlanes + i];
sqsum[i % cn] += ar[VTraits<v_int32>::vlanes() + i];
}
}
v_cleanup();

Loading…
Cancel
Save