potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-30 15:53:03 +04:00
Code Packages Releases Wiki Activity

core(intrin): drop hasSIMD128 checks

Browse Source 
- use compile-time checks instead (`#if CV_SIMD128`)
- runtime checks are useless
This commit is contained in:
Alexander Alekhin
2019-06-08 19:14:35 +00:00
parent 254f88f805
commit 1e9ad5476d
14 changed files with 27 additions and 166 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/imgproc/src/imgwarp.cpp
+3 -23
View File
@@ -446,8 +446,7 @@ struct RemapVec_8u
{
int cn = _src.channels(), x = 0, sstep = (int)_src.step;
if( (cn != 1 && cn != 3 && cn != 4) || !hasSIMD128() ||
sstep > 0x8000 )
if( (cn != 1 && cn != 3 && cn != 4) || sstep > 0x8000 )
return 0;
const uchar *S0 = _src.ptr(), *S1 = _src.ptr(1);
@@ -1098,9 +1097,6 @@ public:
int brows0 = std::min(128, dst->rows), map_depth = m1->depth();
int bcols0 = std::min(buf_size/brows0, dst->cols);
brows0 = std::min(buf_size/bcols0, dst->rows);
#if CV_SIMD128
bool useSIMD = hasSIMD128();
#endif
Mat _bufxy(brows0, bcols0, CV_16SC2), _bufa;
if( !nnfunc )
@@ -1147,7 +1143,6 @@ public:
x1 = 0;
#if CV_SIMD128
if( useSIMD )
{
int span = v_float32x4::nlanes;
for( ; x1 <= bcols - span * 2; x1 += span * 2 )
@@ -1189,7 +1184,6 @@ public:
x1 = 0;
#if CV_SIMD128
if (useSIMD)
{
v_uint16x8 v_scale = v_setall_u16(INTER_TAB_SIZE2 - 1);
int span = v_uint16x8::nlanes;
@@ -1207,7 +1201,6 @@ public:
x1 = 0;
#if CV_SIMD128
if( useSIMD )
{
v_float32x4 v_scale = v_setall_f32((float)INTER_TAB_SIZE);
v_int32x4 v_scale2 = v_setall_s32(INTER_TAB_SIZE - 1);
@@ -1245,7 +1238,6 @@ public:
x1 = 0;
#if CV_SIMD128
if( useSIMD )
{
v_float32x4 v_scale = v_setall_f32((float)INTER_TAB_SIZE);
v_int32x4 v_scale2 = v_setall_s32(INTER_TAB_SIZE - 1), v_scale3 = v_setall_s32(INTER_TAB_SIZE);
@@ -1898,9 +1890,6 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
size.height = 1;
}
#if CV_SIMD128
bool useSIMD = hasSIMD128();
#endif
#if CV_TRY_SSE4_1
bool useSSE4_1 = CV_CPU_HAS_SUPPORT_SSE4_1;
#endif
@@ -1931,7 +1920,6 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
#endif
{
#if CV_SIMD128
if( useSIMD )
{
int span = v_int16x8::nlanes;
for( ; x <= size.width - span; x += span )
@@ -1961,7 +1949,6 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
#endif
{
#if CV_SIMD128
if( useSIMD )
{
v_float32x4 v_scale = v_setall_f32((float)INTER_TAB_SIZE);
v_int32x4 v_mask = v_setall_s32(INTER_TAB_SIZE - 1);
@@ -2002,10 +1989,11 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
{
#if CV_SIMD128
int span = v_float32x4::nlanes;
if( useSIMD )
{
for( ; x <= (size.width << 1) - span * 2; x += span * 2 )
v_store(dst1 + x, v_pack(v_round(v_load(src1f + x)),
v_round(v_load(src1f + x + span))));
}
#endif
for( ; x < size.width; x++ )
{
@@ -2022,7 +2010,6 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
#endif
{
#if CV_SIMD128
if( useSIMD )
{
v_float32x4 v_scale = v_setall_f32((float)INTER_TAB_SIZE);
v_int32x4 v_mask = v_setall_s32(INTER_TAB_SIZE - 1);
@@ -2063,7 +2050,6 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
else if( m1type == CV_16SC2 && dstm1type == CV_32FC1 )
{
#if CV_SIMD128
if( useSIMD )
{
v_uint16x8 v_mask2 = v_setall_u16(INTER_TAB_SIZE2-1);
v_uint32x4 v_zero = v_setzero_u32(), v_mask = v_setall_u32(INTER_TAB_SIZE-1);
@@ -2113,7 +2099,6 @@ void cv::convertMaps( InputArray _map1, InputArray _map2,
else if( m1type == CV_16SC2 && dstm1type == CV_32FC2 )
{
#if CV_SIMD128
if( useSIMD )
{
v_int16x8 v_mask2 = v_setall_s16(INTER_TAB_SIZE2-1);
v_int32x4 v_zero = v_setzero_s32(), v_mask = v_setall_s32(INTER_TAB_SIZE-1);
@@ -2189,9 +2174,6 @@ public:
#if CV_TRY_AVX2
bool useAVX2 = CV_CPU_HAS_SUPPORT_AVX2;
#endif
#if CV_SIMD128
bool useSIMD = hasSIMD128();
#endif
#if CV_TRY_SSE4_1
bool useSSE4_1 = CV_CPU_HAS_SUPPORT_SSE4_1;
#endif
@@ -2226,7 +2208,6 @@ public:
#endif
{
#if CV_SIMD128
if( useSIMD )
{
v_int32x4 v_X0 = v_setall_s32(X0), v_Y0 = v_setall_s32(Y0);
int span = v_uint16x8::nlanes;
@@ -2260,7 +2241,6 @@ public:
x1 = opt_AVX2::warpAffineBlockline(adelta + x, bdelta + x, xy, alpha, X0, Y0, bw);
#endif
#if CV_SIMD128
if( useSIMD )
{
v_int32x4 v__X0 = v_setall_s32(X0), v__Y0 = v_setall_s32(Y0);
v_int32x4 v_mask = v_setall_s32(INTER_TAB_SIZE - 1);