added C= support; modified threading logic in threshold

modules/core/src/parallel.cpp

@@ -42,7 +42,7 @@
 
 #include "precomp.hpp"
 
-#if !defined HAVE_TBB && !defined HAVE_OPENMP && !defined HAVE_GCD && !defined HAVE_CONCURRENCY
+#if !defined HAVE_TBB && !defined HAVE_OPENMP && !defined HAVE_GCD && !defined HAVE_CONCURRENCY && !defined HAVE_CSTRIPES
     #ifdef __APPLE__
         #define HAVE_GCD
     #elif defined _MSC_VER && _MSC_VER >= 1600
@@ -66,7 +66,10 @@
     #else
         #undef HAVE_TBB
     #endif // end TBB version
-#endif // HAVE_CONCURRENCY
+#elif defined HAVE_CSTRIPES
+    #include "C=.h"
+    #undef shared
+#endif
 
 /*
     HAVE_TBB - using TBB
@@ -137,11 +140,22 @@ namespace cv
         for (int i = range.start; i < range.end; ++i)
             body(Range(i, i + 1));
 
-#elif defined (HAVE_GCD)
+#elif defined HAVE_GCD
 
         dispatch_queue_t concurrent_queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
         dispatch_apply_f(range.end - range.start, concurrent_queue, &const_cast<ParallelLoopBody&>(body), block_function);
 
+#elif defined HAVE_CSTRIPES
+
+        parallel()
+        {
+            int offset = range.start;
+            int len = range.end - offset;
+            Range r(offset + CPX_RANGE_START(len), offset + CPX_RANGE_END(len));
+            body(r);
+            barrier();
+        }
+
 #else
 
         body(range);

modules/imgproc/src/thresh.cpp

@@ -730,11 +730,6 @@ double cv::threshold( InputArray _src, OutputArray _dst, double thresh, double m
     _dst.create( src.size(), src.type() );
     Mat dst = _dst.getMat();
 
-    int nStripes = 1;
-#if defined HAVE_TBB && defined ANDROID
-    nStripes = 4;
-#endif
-
     if( src.depth() == CV_8U )
     {
         int ithresh = cvFloor(thresh);
@@ -793,8 +788,10 @@ double cv::threshold( InputArray _src, OutputArray _dst, double thresh, double m
         ;
     else
         CV_Error( CV_StsUnsupportedFormat, "" );
-        
-    parallel_for_(Range(0, nStripes),
+    
+    size_t nStripes = (src.total() + (1<<15)) >> 16;
+    nStripes = MAX(MIN(nStripes, (size_t)4), (size_t)1);
+    parallel_for_(Range(0, (int)nStripes),
                   ThresholdRunner(src, dst, nStripes, thresh, maxval, type));
     return thresh;
 }