Fix for SSE2 intrinsics problem in the part of saturation arithmetic processing durin...
authoramatyuko <andrey.matyukov@intel.com>
Tue, 31 Jul 2018 18:54:15 +0000 (21:54 +0300)
committeramatyuko <andrey.matyukov@intel.com>
Wed, 1 Aug 2018 13:04:08 +0000 (16:04 +0300)
commit3ea2586a5a1749877e4fc865d29bdf31c3bff956
treef461f818d33a7a65d6a4df5c4f2ba400e5ba5bfe
parent9433784e470f8df997ada9f8ecbecc94b5d3438f
Fix for SSE2 intrinsics problem in the part of saturation arithmetic processing during 32s->16u packed conversion -
for some big negative values less than -INT_MAX+32767 the sign of the numbers is lost due to overflow that leads to
incorrect saturation to MAX value, instead of zero.
The issue is not reproduced with CV_ENABLED_INTRINSICS=OFF
modules/core/include/opencv2/core/hal/intrin_sse.hpp
modules/core/src/convert.cpp