Add sparse conversion tests (dense2sparse, sparse2dense) (#873)

Author: aartbik

test/00_sparse/Basic.cu

@@ -38,17 +38,13 @@
 
 using namespace matx;
 
-template <typename T> class SparseTest : public ::testing::Test { };
+template <typename T> class BasicSparseTest : public ::testing::Test { };
 
-template <typename TensorType> class SparseTestsAll : public SparseTest<TensorType> { };
+template <typename T> class BasicSparseTestsAll : public BasicSparseTest<T> { };
 
-// For all operations that run on host and device.
-TYPED_TEST_SUITE(SparseTestsAll, MatXAllTypesAllExecs);
+TYPED_TEST_SUITE(BasicSparseTestsAll, MatXAllTypesAllExecs);
 
-// For all operations that run on device only.
-TYPED_TEST_SUITE(SparseTestsAllCUDA, MatXAllTypesCUDAExec);
-
-TYPED_TEST(SparseTestsAll, MakeZeroCOO) {
+TYPED_TEST(BasicSparseTestsAll, MakeZeroCOO) {
   MATX_ENTER_HANDLER();
   using TestType = cuda::std::tuple_element_t<0, TypeParam>;
   auto A = experimental::make_zero_tensor_coo<TestType, index_t>({17, 33});
@@ -60,10 +56,12 @@ TYPED_TEST(SparseTestsAll, MakeZeroCOO) {
   ASSERT_EQ(A.posSize(1), 0);
   ASSERT_EQ(A.crdSize(0), 0);
   ASSERT_EQ(A.crdSize(1), 0);
+  // Element getter.
+  ASSERT_EQ(A(0, 0), static_cast<TestType>(0)); // not found
   MATX_EXIT_HANDLER();
 }
 
-TYPED_TEST(SparseTestsAll, MakeCOO) {
+TYPED_TEST(BasicSparseTestsAll, MakeCOO) {
   MATX_ENTER_HANDLER();
   using TestType = cuda::std::tuple_element_t<0, TypeParam>;
   auto vals = make_tensor<TestType>({3});
@@ -91,7 +89,7 @@ TYPED_TEST(SparseTestsAll, MakeCOO) {
   MATX_EXIT_HANDLER();
 }
 
-TYPED_TEST(SparseTestsAll, MakeZeroCSR) {
+TYPED_TEST(BasicSparseTestsAll, MakeZeroCSR) {
   MATX_ENTER_HANDLER();
   using TestType = cuda::std::tuple_element_t<0, TypeParam>;
   auto A = experimental::make_zero_tensor_csr<TestType, index_t, index_t>({17, 33});
@@ -103,10 +101,12 @@ TYPED_TEST(SparseTestsAll, MakeZeroCSR) {
   ASSERT_EQ(A.posSize(1), 18);
   ASSERT_EQ(A.crdSize(0), 0);
   ASSERT_EQ(A.crdSize(1), 0);
+  // Element getter.
+  ASSERT_EQ(A(0, 0), static_cast<TestType>(0)); // not found
   MATX_EXIT_HANDLER();
 }
 
-TYPED_TEST(SparseTestsAll, MakeCSR) {
+TYPED_TEST(BasicSparseTestsAll, MakeCSR) {
   MATX_ENTER_HANDLER();
   using TestType = cuda::std::tuple_element_t<0, TypeParam>;
   auto vals = make_tensor<TestType>({3});
@@ -134,7 +134,7 @@ TYPED_TEST(SparseTestsAll, MakeCSR) {
   MATX_EXIT_HANDLER();
 }
 
-TYPED_TEST(SparseTestsAll, MakeZeroCSC) {
+TYPED_TEST(BasicSparseTestsAll, MakeZeroCSC) {
   MATX_ENTER_HANDLER();
   using TestType = cuda::std::tuple_element_t<0, TypeParam>;
   auto A = experimental::make_zero_tensor_csc<TestType, index_t, index_t>({17, 33});
@@ -146,10 +146,12 @@ TYPED_TEST(SparseTestsAll, MakeZeroCSC) {
   ASSERT_EQ(A.posSize(1), 34);
   ASSERT_EQ(A.crdSize(0), 0);
   ASSERT_EQ(A.crdSize(1), 0);
+  // Element getter.
+  ASSERT_EQ(A(0, 0), static_cast<TestType>(0)); // not found
   MATX_EXIT_HANDLER();
 }
 
-TYPED_TEST(SparseTestsAll, MakeCSC) {
+TYPED_TEST(BasicSparseTestsAll, MakeCSC) {
   MATX_ENTER_HANDLER();
   using TestType = cuda::std::tuple_element_t<0, TypeParam>;
   auto vals = make_tensor<TestType>({3});

test/00_sparse/Convert.cu

@@ -0,0 +1,200 @@
+////////////////////////////////////////////////////////////////////////////////
+// BSD 3-Clause License
+//
+// Copyright (c) 2025, NVIDIA Corporation
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+// 1. Redistributions of source code must retain the above copyright notice,
+//    this list of conditions and the following disclaimer.
+//
+// 2. Redistributions in binary form must reproduce the above copyright notice,
+//    this list of conditions and the following disclaimer in the documentation
+//    and/or other materials provided with the distribution.
+//
+// 3. Neither the name of the copyright holder nor the names of its
+//    contributors may be used to endorse or promote products derived from
+//    this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+/////////////////////////////////////////////////////////////////////////////////
+
+#include "assert.h"
+#include "matx.h"
+#include "test_types.h"
+#include "utilities.h"
+#include "gtest/gtest.h"
+
+using namespace matx;
+
+// Helper method
+template <typename T> static auto makeD() {
+  const index_t m = 10;
+  const index_t n = 10;
+  tensor_t<T, 2> D = make_tensor<T>({m, n});
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      D(i, j) = static_cast<T>(0);
+    }
+  }
+  D(0, 1) = static_cast<T>(1);
+  D(4, 4) = static_cast<T>(2);
+  D(9, 1) = static_cast<T>(3);
+  D(9, 9) = static_cast<T>(4);
+  return D;
+}
+
+template <typename T> class ConvertSparseTest : public ::testing::Test { };
+
+template <typename T> class ConvertSparseTestsAll : public ConvertSparseTest<T> { };
+
+TYPED_TEST_SUITE(ConvertSparseTestsAll, MatXFloatNonComplexTypesCUDAExec);
+
+TYPED_TEST(ConvertSparseTestsAll, ConvertCOO) {
+  MATX_ENTER_HANDLER();
+  using TestType = cuda::std::tuple_element_t<0, TypeParam>;
+  using ExecType = cuda::std::tuple_element_t<1, TypeParam>;
+
+  ExecType exec{};
+
+  auto D = makeD<TestType>();
+  const auto m = D.Size(0);
+  const auto n = D.Size(1);
+
+  // Convert dense D to sparse S.
+  auto S = experimental::make_zero_tensor_coo<TestType, index_t>({m, n});
+  (S = dense2sparse(D)).run(exec);
+  ASSERT_EQ(S.Rank(), 2);
+  ASSERT_EQ(S.Size(0), m);
+  ASSERT_EQ(S.Size(1), n);
+  ASSERT_EQ(S.Nse(), 4);
+  ASSERT_EQ(S.posSize(0), 2);
+  ASSERT_EQ(S.posSize(1), 0);
+  ASSERT_EQ(S.crdSize(0), 4);
+  ASSERT_EQ(S.crdSize(1), 4);
+
+  // Getters are expensive, but fully functional!
+  exec.sync();
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      ASSERT_EQ(S(i, j), D(i, j));
+    }
+  }
+
+  // Convert sparse S back to dense D.
+  auto O = make_tensor<TestType>({m, n});
+  (O = sparse2dense(S)).run(exec);
+  
+  // Back to cheap random-access getters only.
+  exec.sync();
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      ASSERT_EQ(O(i, j), D(i, j));
+    }
+  }
+
+  MATX_EXIT_HANDLER();
+}
+
+TYPED_TEST(ConvertSparseTestsAll, ConvertCSR) {
+  MATX_ENTER_HANDLER();
+  using TestType = cuda::std::tuple_element_t<0, TypeParam>;
+  using ExecType = cuda::std::tuple_element_t<1, TypeParam>;
+
+  ExecType exec{};
+
+  auto D = makeD<TestType>();
+  const auto m = D.Size(0);
+  const auto n = D.Size(1);
+
+  // Convert dense D to sparse S.
+  auto S = experimental::make_zero_tensor_csr<TestType, index_t, index_t>({m, n});
+  (S = dense2sparse(D)).run(exec);
+  ASSERT_EQ(S.Rank(), 2);
+  ASSERT_EQ(S.Size(0), m);
+  ASSERT_EQ(S.Size(1), n);
+  ASSERT_EQ(S.Nse(), 4);
+  ASSERT_EQ(S.posSize(0), 0);
+  ASSERT_EQ(S.posSize(1), m + 1);
+  ASSERT_EQ(S.crdSize(0), 0);
+  ASSERT_EQ(S.crdSize(1), 4);
+
+  // Getters are expensive, but fully functional!
+  exec.sync();
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      ASSERT_EQ(S(i, j), D(i, j));
+    }
+  }
+
+  // Convert sparse S back to dense D.
+  auto O = make_tensor<TestType>({m, n});
+  (O = sparse2dense(S)).run(exec);
+  
+  // Back to cheap random-access getters only.
+  exec.sync();
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      ASSERT_EQ(O(i, j), D(i, j));
+    }
+  }
+
+  MATX_EXIT_HANDLER();
+}
+
+TYPED_TEST(ConvertSparseTestsAll, ConvertCSC) {
+  MATX_ENTER_HANDLER();
+  using TestType = cuda::std::tuple_element_t<0, TypeParam>;
+  using ExecType = cuda::std::tuple_element_t<1, TypeParam>;
+
+  ExecType exec{};
+
+  auto D = makeD<TestType>();
+  const auto m = D.Size(0);
+  const auto n = D.Size(1);
+
+  // Convert dense D to sparse S.
+  auto S = experimental::make_zero_tensor_csc<TestType, index_t, index_t>({m, n});
+  (S = dense2sparse(D)).run(exec);
+  ASSERT_EQ(S.Rank(), 2);
+  ASSERT_EQ(S.Size(0), m);
+  ASSERT_EQ(S.Size(1), n);
+  ASSERT_EQ(S.Nse(), 4);
+  ASSERT_EQ(S.posSize(0), 0);
+  ASSERT_EQ(S.posSize(1), n + 1);
+  ASSERT_EQ(S.crdSize(0), 0);
+  ASSERT_EQ(S.crdSize(1), 4);
+
+  // Getters are expensive, but fully functional!
+  exec.sync();
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      ASSERT_EQ(S(i, j), D(i, j));
+    }
+  }
+
+  // Convert sparse S back to dense D.
+  auto O = make_tensor<TestType>({m, n});
+  (O = sparse2dense(S)).run(exec);
+  
+  // Back to cheap random-access getters only.
+  exec.sync();
+  for (index_t i = 0; i < m; i++) {
+    for (index_t j = 0; j < n; j++) {
+      ASSERT_EQ(O(i, j), D(i, j));
+    }
+  }
+
+  MATX_EXIT_HANDLER();
+}

test/CMakeLists.txt

@@ -38,6 +38,7 @@ set (test_sources
     01_radar/ambgfun.cu
     01_radar/dct.cu
     00_sparse/Basic.cu
+    00_sparse/Convert.cu
     main.cu
 )