Added the XSWAP, XSCAL and XCOPY level-1 routines

10 files changed, 707 insertions, 0 deletions

diff --git a/test/correctness/routines/level1/xcopy.cc b/test/correctness/routines/level1/xcopy.cc
new file mode 100644
index 00000000..8a06a722
--- /dev/null
+++ b/test/correctness/routines/level1/xcopy.cc
@@ -0,0 +1,32 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements the tests for the Xcopy routine.
+//
+// =================================================================================================
+
+#include "correctness/testblas.h"
+#include "routines/level1/xcopy.h"
+
+// =================================================================================================
+
+// Shortcuts to the clblast namespace
+using float2 = clblast::float2;
+using double2 = clblast::double2;
+
+// Main function (not within the clblast namespace)
+int main(int argc, char *argv[]) {
+  clblast::RunTests<clblast::TestXcopy<float>, float, float>(argc, argv, false, "SCOPY");
+  clblast::RunTests<clblast::TestXcopy<double>, double, double>(argc, argv, true, "DCOPY");
+  clblast::RunTests<clblast::TestXcopy<float2>, float2, float2>(argc, argv, true, "CCOPY");
+  clblast::RunTests<clblast::TestXcopy<double2>, double2, double2>(argc, argv, true, "ZCOPY");
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/correctness/routines/level1/xscal.cc b/test/correctness/routines/level1/xscal.cc
new file mode 100644
index 00000000..ceb1b7cd
--- /dev/null
+++ b/test/correctness/routines/level1/xscal.cc
@@ -0,0 +1,32 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements the tests for the Xscal routine.
+//
+// =================================================================================================
+
+#include "correctness/testblas.h"
+#include "routines/level1/xscal.h"
+
+// =================================================================================================
+
+// Shortcuts to the clblast namespace
+using float2 = clblast::float2;
+using double2 = clblast::double2;
+
+// Main function (not within the clblast namespace)
+int main(int argc, char *argv[]) {
+  clblast::RunTests<clblast::TestXscal<float>, float, float>(argc, argv, false, "SSCAL");
+  clblast::RunTests<clblast::TestXscal<double>, double, double>(argc, argv, true, "DSCAL");
+  clblast::RunTests<clblast::TestXscal<float2>, float2, float2>(argc, argv, true, "CSCAL");
+  clblast::RunTests<clblast::TestXscal<double2>, double2, double2>(argc, argv, true, "ZSCAL");
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/correctness/routines/level1/xswap.cc b/test/correctness/routines/level1/xswap.cc
new file mode 100644
index 00000000..140ccf24
--- /dev/null
+++ b/test/correctness/routines/level1/xswap.cc
@@ -0,0 +1,32 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements the tests for the Xswap routine.
+//
+// =================================================================================================
+
+#include "correctness/testblas.h"
+#include "routines/level1/xswap.h"
+
+// =================================================================================================
+
+// Shortcuts to the clblast namespace
+using float2 = clblast::float2;
+using double2 = clblast::double2;
+
+// Main function (not within the clblast namespace)
+int main(int argc, char *argv[]) {
+  clblast::RunTests<clblast::TestXswap<float>, float, float>(argc, argv, false, "SSWAP");
+  clblast::RunTests<clblast::TestXswap<double>, double, double>(argc, argv, true, "DSWAP");
+  clblast::RunTests<clblast::TestXswap<float2>, float2, float2>(argc, argv, true, "CSWAP");
+  clblast::RunTests<clblast::TestXswap<double2>, double2, double2>(argc, argv, true, "ZSWAP");
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/performance/routines/level1/xcopy.cc b/test/performance/routines/level1/xcopy.cc
new file mode 100644
index 00000000..70b6b348
--- /dev/null
+++ b/test/performance/routines/level1/xcopy.cc
@@ -0,0 +1,40 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements the Xcopy command-line interface performance tester.
+//
+// =================================================================================================
+
+#include "performance/client.h"
+#include "routines/level1/xcopy.h"
+
+// =================================================================================================
+
+// Shortcuts to the clblast namespace
+using float2 = clblast::float2;
+using double2 = clblast::double2;
+
+// Main function (not within the clblast namespace)
+int main(int argc, char *argv[]) {
+  switch(clblast::GetPrecision(argc, argv)) {
+    case clblast::Precision::kHalf:
+      throw std::runtime_error("Unsupported precision mode");
+    case clblast::Precision::kSingle:
+      clblast::RunClient<clblast::TestXcopy<float>, float, float>(argc, argv); break;
+    case clblast::Precision::kDouble:
+      clblast::RunClient<clblast::TestXcopy<double>, double, double>(argc, argv); break;
+    case clblast::Precision::kComplexSingle:
+      clblast::RunClient<clblast::TestXcopy<float2>, float2, float2>(argc, argv); break;
+    case clblast::Precision::kComplexDouble:
+      clblast::RunClient<clblast::TestXcopy<double2>, double2, double2>(argc, argv); break;
+  }
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/performance/routines/level1/xscal.cc b/test/performance/routines/level1/xscal.cc
new file mode 100644
index 00000000..3963ba3a
--- /dev/null
+++ b/test/performance/routines/level1/xscal.cc
@@ -0,0 +1,40 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements the Xscal command-line interface performance tester.
+//
+// =================================================================================================
+
+#include "performance/client.h"
+#include "routines/level1/xscal.h"
+
+// =================================================================================================
+
+// Shortcuts to the clblast namespace
+using float2 = clblast::float2;
+using double2 = clblast::double2;
+
+// Main function (not within the clblast namespace)
+int main(int argc, char *argv[]) {
+  switch(clblast::GetPrecision(argc, argv)) {
+    case clblast::Precision::kHalf:
+      throw std::runtime_error("Unsupported precision mode");
+    case clblast::Precision::kSingle:
+      clblast::RunClient<clblast::TestXscal<float>, float, float>(argc, argv); break;
+    case clblast::Precision::kDouble:
+      clblast::RunClient<clblast::TestXscal<double>, double, double>(argc, argv); break;
+    case clblast::Precision::kComplexSingle:
+      clblast::RunClient<clblast::TestXscal<float2>, float2, float2>(argc, argv); break;
+    case clblast::Precision::kComplexDouble:
+      clblast::RunClient<clblast::TestXscal<double2>, double2, double2>(argc, argv); break;
+  }
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/performance/routines/level1/xswap.cc b/test/performance/routines/level1/xswap.cc
new file mode 100644
index 00000000..94f271ee
--- /dev/null
+++ b/test/performance/routines/level1/xswap.cc
@@ -0,0 +1,40 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements the Xswap command-line interface performance tester.
+//
+// =================================================================================================
+
+#include "performance/client.h"
+#include "routines/level1/xswap.h"
+
+// =================================================================================================
+
+// Shortcuts to the clblast namespace
+using float2 = clblast::float2;
+using double2 = clblast::double2;
+
+// Main function (not within the clblast namespace)
+int main(int argc, char *argv[]) {
+  switch(clblast::GetPrecision(argc, argv)) {
+    case clblast::Precision::kHalf:
+      throw std::runtime_error("Unsupported precision mode");
+    case clblast::Precision::kSingle:
+      clblast::RunClient<clblast::TestXswap<float>, float, float>(argc, argv); break;
+    case clblast::Precision::kDouble:
+      clblast::RunClient<clblast::TestXswap<double>, double, double>(argc, argv); break;
+    case clblast::Precision::kComplexSingle:
+      clblast::RunClient<clblast::TestXswap<float2>, float2, float2>(argc, argv); break;
+    case clblast::Precision::kComplexDouble:
+      clblast::RunClient<clblast::TestXswap<double2>, double2, double2>(argc, argv); break;
+  }
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/routines/level1/xcopy.h b/test/routines/level1/xcopy.h
new file mode 100644
index 00000000..8d324d88
--- /dev/null
+++ b/test/routines/level1/xcopy.h
@@ -0,0 +1,117 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements a class with static methods to describe the Xcopy routine. Examples of
+// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These
+// static methods are used by the correctness tester and the performance tester.
+//
+// =================================================================================================
+
+#ifndef CLBLAST_TEST_ROUTINES_XCOPY_H_
+#define CLBLAST_TEST_ROUTINES_XCOPY_H_
+
+#include <vector>
+#include <string>
+
+#include "wrapper_clblas.h"
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class TestXcopy {
+ public:
+
+  // The BLAS level: 1, 2, or 3
+  static size_t BLASLevel() { return 1; }
+
+  // The list of arguments relevant for this routine
+  static std::vector<std::string> GetOptions() {
+    return {kArgN,
+            kArgXInc, kArgYInc,
+            kArgXOffset, kArgYOffset};
+  }
+
+  // Describes how to obtain the sizes of the buffers
+  static size_t GetSizeX(const Arguments<T> &args) {
+    return args.n * args.x_inc + args.x_offset;
+  }
+  static size_t GetSizeY(const Arguments<T> &args) {
+    return args.n * args.y_inc + args.y_offset;
+  }
+
+  // Describes how to set the sizes of all the buffers
+  static void SetSizes(Arguments<T> &args) {
+    args.x_size = GetSizeX(args);
+    args.y_size = GetSizeY(args);
+  }
+
+  // Describes what the default values of the leading dimensions of the matrices are
+  static size_t DefaultLDA(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine
+
+  // Describes which transpose options are relevant for this routine
+  using Transposes = std::vector<Transpose>;
+  static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine
+  static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine
+
+  // Describes how to run the CLBlast routine
+  static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = Copy<T>(args.n,
+                          buffers.x_vec(), args.x_offset, args.x_inc,
+                          buffers.y_vec(), args.y_offset, args.y_inc,
+                          &queue_plain, &event);
+    clWaitForEvents(1, &event);
+    return status;
+  }
+
+  // Describes how to run the clBLAS routine (for correctness/performance comparison)
+  static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = clblasXcopy<T>(args.n,
+                                 buffers.x_vec(), args.x_offset, args.x_inc,
+                                 buffers.y_vec(), args.y_offset, args.y_inc,
+                                 1, &queue_plain, 0, nullptr, &event);
+    clWaitForEvents(1, &event);
+    return static_cast<StatusCode>(status);
+  }
+
+  // Describes how to download the results of the computation (more importantly: which buffer)
+  static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+    std::vector<T> result(args.y_size, static_cast<T>(0));
+    buffers.y_vec.Read(queue, args.y_size, result);
+    return result;
+  }
+
+  // Describes how to compute the indices of the result buffer
+  static size_t ResultID1(const Arguments<T> &args) { return args.n; }
+  static size_t ResultID2(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t) {
+    return id1*args.y_inc + args.y_offset;
+  }
+
+  // Describes how to compute performance metrics
+  static size_t GetFlops(const Arguments<T> &args) {
+    return 1 * args.n;
+  }
+  static size_t GetBytes(const Arguments<T> &args) {
+    return (2 * args.n) * sizeof(T);
+  }
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_TEST_ROUTINES_XCOPY_H_
+#endif
diff --git a/test/routines/level1/xscal.h b/test/routines/level1/xscal.h
new file mode 100644
index 00000000..d990afcc
--- /dev/null
+++ b/test/routines/level1/xscal.h
@@ -0,0 +1,112 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements a class with static methods to describe the Xscal routine. Examples of
+// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These
+// static methods are used by the correctness tester and the performance tester.
+//
+// =================================================================================================
+
+#ifndef CLBLAST_TEST_ROUTINES_XSCAL_H_
+#define CLBLAST_TEST_ROUTINES_XSCAL_H_
+
+#include <vector>
+#include <string>
+
+#include "wrapper_clblas.h"
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class TestXscal {
+ public:
+
+  // The BLAS level: 1, 2, or 3
+  static size_t BLASLevel() { return 1; }
+
+  // The list of arguments relevant for this routine
+  static std::vector<std::string> GetOptions() {
+    return {kArgN,
+            kArgXInc,
+            kArgXOffset,
+            kArgAlpha};
+  }
+
+  // Describes how to obtain the sizes of the buffers
+  static size_t GetSizeX(const Arguments<T> &args) {
+    return args.n * args.x_inc + args.x_offset;
+  }
+
+  // Describes how to set the sizes of all the buffers
+  static void SetSizes(Arguments<T> &args) {
+    args.x_size = GetSizeX(args);
+  }
+
+  // Describes what the default values of the leading dimensions of the matrices are
+  static size_t DefaultLDA(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine
+
+  // Describes which transpose options are relevant for this routine
+  using Transposes = std::vector<Transpose>;
+  static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine
+  static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine
+
+  // Describes how to run the CLBlast routine
+  static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = Scal(args.n, args.alpha,
+                       buffers.x_vec(), args.x_offset, args.x_inc,
+                       &queue_plain, &event);
+    clWaitForEvents(1, &event);
+    return status;
+  }
+
+  // Describes how to run the clBLAS routine (for correctness/performance comparison)
+  static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = clblasXscal(args.n, args.alpha,
+                              buffers.x_vec(), args.x_offset, args.x_inc,
+                              1, &queue_plain, 0, nullptr, &event);
+    clWaitForEvents(1, &event);
+    return static_cast<StatusCode>(status);
+  }
+
+  // Describes how to download the results of the computation (more importantly: which buffer)
+  static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+    std::vector<T> result(args.x_size, static_cast<T>(0));
+    buffers.x_vec.Read(queue, args.x_size, result);
+    return result;
+  }
+
+  // Describes how to compute the indices of the result buffer
+  static size_t ResultID1(const Arguments<T> &args) { return args.n; }
+  static size_t ResultID2(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t) {
+    return id1*args.x_inc + args.x_offset;
+  }
+
+  // Describes how to compute performance metrics
+  static size_t GetFlops(const Arguments<T> &args) {
+    return args.n;
+  }
+  static size_t GetBytes(const Arguments<T> &args) {
+    return (2 * args.n) * sizeof(T);
+  }
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_TEST_ROUTINES_XSCAL_H_
+#endif
diff --git a/test/routines/level1/xswap.h b/test/routines/level1/xswap.h
new file mode 100644
index 00000000..2096a2c3
--- /dev/null
+++ b/test/routines/level1/xswap.h
@@ -0,0 +1,118 @@
+
+// =================================================================================================
+// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This
+// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max-
+// width of 100 characters per line.
+//
+// Author(s):
+//   Cedric Nugteren <www.cedricnugteren.nl>
+//
+// This file implements a class with static methods to describe the Xswap routine. Examples of
+// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These
+// static methods are used by the correctness tester and the performance tester.
+//
+// =================================================================================================
+
+#ifndef CLBLAST_TEST_ROUTINES_XSWAP_H_
+#define CLBLAST_TEST_ROUTINES_XSWAP_H_
+
+#include <vector>
+#include <string>
+
+#include "wrapper_clblas.h"
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class TestXswap {
+ public:
+
+  // The BLAS level: 1, 2, or 3
+  static size_t BLASLevel() { return 1; }
+
+  // The list of arguments relevant for this routine
+  static std::vector<std::string> GetOptions() {
+    return {kArgN,
+            kArgXInc, kArgYInc,
+            kArgXOffset, kArgYOffset};
+  }
+
+  // Describes how to obtain the sizes of the buffers
+  static size_t GetSizeX(const Arguments<T> &args) {
+    return args.n * args.x_inc + args.x_offset;
+  }
+  static size_t GetSizeY(const Arguments<T> &args) {
+    return args.n * args.y_inc + args.y_offset;
+  }
+
+  // Describes how to set the sizes of all the buffers
+  static void SetSizes(Arguments<T> &args) {
+    args.x_size = GetSizeX(args);
+    args.y_size = GetSizeY(args);
+  }
+
+  // Describes what the default values of the leading dimensions of the matrices are
+  static size_t DefaultLDA(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine
+  static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine
+
+  // Describes which transpose options are relevant for this routine
+  using Transposes = std::vector<Transpose>;
+  static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine
+  static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine
+
+  // Describes how to run the CLBlast routine
+  static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = Swap<T>(args.n,
+                          buffers.x_vec(), args.x_offset, args.x_inc,
+                          buffers.y_vec(), args.y_offset, args.y_inc,
+                          &queue_plain, &event);
+    clWaitForEvents(1, &event);
+    return status;
+  }
+
+  // Describes how to run the clBLAS routine (for correctness/performance comparison)
+  static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = clblasXswap<T>(args.n,
+                                 buffers.x_vec(), args.x_offset, args.x_inc,
+                                 buffers.y_vec(), args.y_offset, args.y_inc,
+                                 1, &queue_plain, 0, nullptr, &event);
+    clWaitForEvents(1, &event);
+    return static_cast<StatusCode>(status);
+  }
+
+  // Describes how to download the results of the computation (more importantly: which buffer)
+  static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+    std::vector<T> result(args.x_size + args.y_size, static_cast<T>(0));
+    buffers.x_vec.Read(queue, args.x_size, &result[0]);
+    buffers.y_vec.Read(queue, args.y_size, &result[args.x_size]);
+    return result;
+  }
+
+  // Describes how to compute the indices of the result buffer
+  static size_t ResultID1(const Arguments<T> &args) { return args.n; }
+  static size_t ResultID2(const Arguments<T> &) { return 2; } // x_vec and y_vec
+  static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t id2) {
+    return (id2 == 0) ? id1*args.x_inc + args.x_offset : id1*args.y_inc + args.y_offset;
+  }
+
+  // Describes how to compute performance metrics
+  static size_t GetFlops(const Arguments<T> &args) {
+    return args.n;
+  }
+  static size_t GetBytes(const Arguments<T> &args) {
+    return (2 * args.n) * sizeof(T);
+  }
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_TEST_ROUTINES_XSWAP_H_
+#endif
diff --git a/test/wrapper_clblas.h b/test/wrapper_clblas.h
index 86810fa2..e17fae76 100644
--- a/test/wrapper_clblas.h
+++ b/test/wrapper_clblas.h
@@ -23,6 +23,150 @@ namespace clblast {
 // =================================================================================================
 // BLAS level-1 (vector-vector) routines
 
+// Calls {clblasSswap, clblasDswap, clblasCswap, clblasZswap} with the arguments forwarded.
+template <typename T> clblasStatus clblasXswap(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events);
+template <> clblasStatus clblasXswap<float>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasSswap(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+template <> clblasStatus clblasXswap<double>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasDswap(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+template <> clblasStatus clblasXswap<float2>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasCswap(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+template <> clblasStatus clblasXswap<double2>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasZswap(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+
+// Calls {clblasSscal, clblasDscal, clblasCscal, clblasZscal} with the arguments forwarded.
+clblasStatus clblasXscal(
+  size_t n, float alpha,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasSscal(n, alpha,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+clblasStatus clblasXscal(
+  size_t n, double alpha,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasDscal(n, alpha,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+clblasStatus clblasXscal(
+  size_t n, float2 alpha,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
+    return clblasCscal(n, cl_alpha,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+clblasStatus clblasXscal(
+  size_t n, double2 alpha,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
+    return clblasZscal(n, cl_alpha,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+
+// Calls {clblasScopy, clblasDcopy, clblasCcopy, clblasZcopy} with the arguments forwarded.
+template <typename T> clblasStatus clblasXcopy(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events);
+template <> clblasStatus clblasXcopy<float>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasScopy(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+template <> clblasStatus clblasXcopy<double>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasDcopy(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+template <> clblasStatus clblasXcopy<float2>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasCcopy(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+template <> clblasStatus clblasXcopy<double2>(
+  size_t n,
+  const cl_mem x_vec, size_t x_offset, size_t x_inc,
+  const cl_mem y_vec, size_t y_offset, size_t y_inc,
+  cl_uint num_queues, cl_command_queue *queues,
+  cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
+    return clblasZcopy(n,
+                       x_vec, x_offset, static_cast<int>(x_inc),
+                       y_vec, y_offset, static_cast<int>(y_inc),
+                       num_queues, queues, num_wait_events, wait_events, events);
+}
+
 // Calls {clblasSaxpy, clblasDaxpy, clblasCaxpy, clblasZaxpy} with the arguments forwarded.
 clblasStatus clblasXaxpy(
   size_t n, float alpha,