Pw transform reduce (#8)

* univariate pairwise transform reduce init version added

* initial checkin of bivariate_pairwise_transformReduce

* dividing by 3 elements in vector for piecwiese transform reduce, to bring on rounding error.

* tweak test tolerance

.gitignore

@@ -55,3 +55,15 @@
 /out/build/x64-Debug
 /lattice/x64
 /dancingAVX512/x64
+/cmake-build-debug
+/curveExample/Intel® VTune™ Profiler Results/curveExample
+/curveExample/My Advisor Results - curveExample
+/curveExample/x64/clang-cl/curveExample.tlog
+/curveExample/x64/clang-cl
+/.idea
+/curveExample/x64
+/scratch/x64/Debug
+/scratch/x64
+/Vectorisation/cmake-build-debug/Testing/Temporary
+/Vectorisation/cmake-build-debug
+/lattice/cmake-build-debug

VectorTest/TestAccumulator.cpp

@@ -18,6 +18,7 @@
 
 
 #include <numeric>
+#include <algorithm>
 
 
 void setEpsilonTransformAccumulate(long double& val)
@@ -472,3 +473,112 @@ TEST(TestAccumulator, simple_pairwise)
  evalAccumulateUR_X_pairwise(159, 159);
 
 }
+
+
+
+////////////////////////////////////////////////////
+
+void evalTransformAccumulateUR_X_pairwise(int startLen, int endLen)
+{
+ Numeric testEpsilon;
+ setEpsilonAccumulate(testEpsilon);
+
+
+ for (int SZ = startLen; SZ <= endLen; SZ++)
+ {
+ std::vector<Numeric> v(SZ, asNumber(0.0));
+ for (int i = 0; i < SZ; i++)
+ {
+ v[i] -= asNumber(500.0 - i);
+ }
+
+ VecXX test(v);
+ auto Sum = [](auto lhs, auto rhs) { return lhs + rhs; };
+ auto SQR = [](auto lhs) { return lhs * lhs; };
+ Numeric resSTL = std::inner_product(v.begin(), v.end(), v.begin(), 0.0);
+ Numeric resAcc1 = ApplyTransformAccumulate2UR_X_pairwise(test, SQR,Sum);
+ testEpsilon *= (abs(resSTL) + abs(resAcc1));
+ EXPECT_NEAR(resSTL, resAcc1, testEpsilon);
+ }
+
+}
+
+TEST(TestAccumulator, simple_pairwise_transformReduce)
+{
+ EXPECT_EQ(1, 1);
+ EXPECT_TRUE(true);
+
+ evalTransformAccumulateUR_X_pairwise(32, 32);
+ evalTransformAccumulateUR_X_pairwise(64, 64);
+ evalTransformAccumulateUR_X_pairwise(128, 128);
+ evalTransformAccumulateUR_X_pairwise(256, 256);
+ evalTransformAccumulateUR_X_pairwise(512, 512);
+ evalTransformAccumulateUR_X_pairwise(1024, 1024);
+
+ evalTransformAccumulateUR_X_pairwise(1, 31);
+
+ evalTransformAccumulateUR_X_pairwise(32, 80);
+
+ evalTransformAccumulateUR_X_pairwise(957, 1043);
+
+
+ evalTransformAccumulateUR_X_pairwise(159, 159);
+
+}
+
+
+
+void evalTransformAccumulateUR_X_bivariate_pairwise(int startLen, int endLen)
+{
+ Numeric testEpsilon;
+ setEpsilonAccumulate(testEpsilon);
+
+ //rounding errors in stl 
+
+
+ for (int SZ = startLen; SZ <= endLen; SZ++)
+ {
+ std::vector<Numeric> v(SZ, asNumber(0.0));
+
+ auto v2 = v;
+ for (int i = 0; i < SZ; i++)
+ {
+ v[i] -= asNumber(500.0 - i) / 3.0;
+ v2[i] = 2 * v[i];
+ }
+
+ VecXX test(v);
+ VecXX test2 = asNumber(2.0) * test;
+
+ auto Sum = [](auto lhs, auto rhs) { return lhs + rhs; };
+ auto mult = [](auto lhs,auto rhs) { return lhs * rhs; };
+ Numeric resSTL = std::inner_product(v.begin(), v.end(), v2.begin(), 0.0);
+ Numeric resAcc1 = ApplyTransformAccumulate2UR_X_pairwise(test, test2, mult, Sum);
+
+ testEpsilon *= (abs(resSTL) + abs(resAcc1));
+ EXPECT_NEAR(resSTL, resAcc1, testEpsilon);
+ }
+
+}
+
+TEST(TestAccumulator, bivariate_pairwise_transformReduce)
+{
+ EXPECT_EQ(1, 1);
+ EXPECT_TRUE(true);
+
+ evalTransformAccumulateUR_X_bivariate_pairwise(32, 32);
+ evalTransformAccumulateUR_X_bivariate_pairwise(64, 64);
+ evalTransformAccumulateUR_X_bivariate_pairwise(128, 128);
+ evalTransformAccumulateUR_X_bivariate_pairwise(256, 256);
+ evalTransformAccumulateUR_X_bivariate_pairwise(512, 512);
+ evalTransformAccumulateUR_X_bivariate_pairwise(1024, 1024);
+
+ evalTransformAccumulateUR_X_bivariate_pairwise(1, 31);
+ evalTransformAccumulateUR_X_bivariate_pairwise(32, 80);
+
+ evalTransformAccumulateUR_X_bivariate_pairwise(957, 1043);
+ evalTransformAccumulateUR_X_bivariate_pairwise(159, 159);
+
+}
+
+