From: Alexey Spizhevoy <no@email>
Date: Mon, 31 Jan 2011 10:42:33 +0000 (+0000)
Subject: added buffered version of gpu::integral function and updated performance test (it... 
X-Git-Tag: accepted/2.0/20130307.220821~3586
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=7a29d96cf41ed4e0b14670e4e650a3d13caa2c85;p=profile%2Fivi%2Fopencv.git

added buffered version of gpu::integral function and updated performance test (it still works too slow)
---

diff --git a/modules/gpu/include/opencv2/gpu/gpu.hpp b/modules/gpu/include/opencv2/gpu/gpu.hpp
index 967b813..b719417 100644
--- a/modules/gpu/include/opencv2/gpu/gpu.hpp
+++ b/modules/gpu/include/opencv2/gpu/gpu.hpp
@@ -650,6 +650,9 @@ namespace cv
         //! supports only CV_8UC1 source type
         CV_EXPORTS void integral(const GpuMat& src, GpuMat& sum);
 
+        //! buffered version
+        CV_EXPORTS void integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer);
+
         //! computes the integral image and integral for the squared image
         //! sum will have CV_32S type, sqsum - CV32F type
         //! supports only CV_8UC1 source type
diff --git a/modules/gpu/src/imgproc_gpu.cpp b/modules/gpu/src/imgproc_gpu.cpp
index 4c24684..35aa3d5 100644
--- a/modules/gpu/src/imgproc_gpu.cpp
+++ b/modules/gpu/src/imgproc_gpu.cpp
@@ -61,6 +61,7 @@ void cv::gpu::warpAffine(const GpuMat&, GpuMat&, const Mat&, Size, int) { throw_
 void cv::gpu::warpPerspective(const GpuMat&, GpuMat&, const Mat&, Size, int) { throw_nogpu(); }
 void cv::gpu::rotate(const GpuMat&, GpuMat&, Size, double, double, double, int) { throw_nogpu(); }
 void cv::gpu::integral(const GpuMat&, GpuMat&) { throw_nogpu(); }
+void cv::gpu::integralBuffered(const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
 void cv::gpu::integral(const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
 void cv::gpu::sqrIntegral(const GpuMat&, GpuMat&) { throw_nogpu(); }
 void cv::gpu::columnSum(const GpuMat&, GpuMat&) { throw_nogpu(); }
@@ -546,6 +547,12 @@ void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, d
 
 void cv::gpu::integral(const GpuMat& src, GpuMat& sum)
 {
+    GpuMat buffer;
+    integralBuffered(src, sum, buffer);
+}
+
+void cv::gpu::integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer)
+{
     CV_Assert(src.type() == CV_8UC1);
 
     sum.create(src.rows + 1, src.cols + 1, CV_32S);
@@ -555,10 +562,8 @@ void cv::gpu::integral(const GpuMat& src, GpuMat& sum)
     roiSize.height = src.rows;
 
     NppSt32u bufSize;
-
     nppSafeCall( nppiStIntegralGetSize_8u32u(roiSize, &bufSize) );
-
-    GpuMat buffer(1, bufSize, CV_8UC1);
+    ensureSizeIsEnough(1, bufSize, CV_8UC1, buffer);
 
     nppSafeCall( nppiStIntegral_8u32u_C1R(const_cast<NppSt8u*>(src.ptr<NppSt8u>()), src.step, 
         sum.ptr<NppSt32u>(), sum.step, roiSize, buffer.ptr<NppSt8u>(), bufSize) );
diff --git a/samples/gpu/performance/tests.cpp b/samples/gpu/performance/tests.cpp
index eea14c0..469c84e 100644
--- a/samples/gpu/performance/tests.cpp
+++ b/samples/gpu/performance/tests.cpp
@@ -170,24 +170,26 @@ TEST(cornerHarris)
 TEST(integral)
 {
     Mat src, sum;
-    gpu::GpuMat d_src, d_sum;
+    gpu::GpuMat d_src, d_sum, d_buf;
 
-    for (int size = 1000; size <= 8000; size *= 2)
+    int size = 4000;
+
+    gen(src, size, size, CV_8U, 0, 256);
+    sum.create(size + 1, size + 1, CV_32S);
+
+    d_src = src;
+    d_sum.create(size + 1, size + 1, CV_32S);
+
+    for (int i = 0; i < 5; ++i)
     {
         SUBTEST << "size " << size << ", 8U";
 
-        gen(src, size, size, CV_8U, 0, 256);
-        sum.create(size + 1, size + 1, CV_32S);
-
         CPU_ON;
         integral(src, sum);
         CPU_OFF;
 
-        d_src = src;
-        d_sum.create(size + 1, size + 1, CV_32S);
-
         GPU_ON;
-        gpu::integral(d_src, d_sum);
+        gpu::integralBuffered(d_src, d_sum, d_buf);
         GPU_OFF;
     }
 }