Merge branch 'level1_routines' into development

10 files changed, 610 insertions, 5 deletions

diff --git a/test/correctness/routines/level1/xamax.cc b/test/correctness/routines/level1/xamax.cc
new file mode 100644
index 00000000..ade09e7a
--- /dev/null
+++ b/test/correctness/routines/level1/xamax.cc
@@ -0,0 +1,28 @@
+
+// =================================================================================================
+// 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>
+//
+// =================================================================================================
+
+#include "correctness/testblas.h"
+#include "routines/level1/xamax.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::TestXamax<float>, float, float>(argc, argv, false, "iSAMAX");
+  clblast::RunTests<clblast::TestXamax<double>, double, double>(argc, argv, true, "iDAMAX");
+  clblast::RunTests<clblast::TestXamax<float2>, float2, float2>(argc, argv, true, "iCAMAX");
+  clblast::RunTests<clblast::TestXamax<double2>, double2, double2>(argc, argv, true, "iZAMAX");
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/correctness/routines/level1/xasum.cc b/test/correctness/routines/level1/xasum.cc
new file mode 100644
index 00000000..5ec20596
--- /dev/null
+++ b/test/correctness/routines/level1/xasum.cc
@@ -0,0 +1,28 @@
+
+// =================================================================================================
+// 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>
+//
+// =================================================================================================
+
+#include "correctness/testblas.h"
+#include "routines/level1/xasum.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::TestXasum<float>, float, float>(argc, argv, false, "SASUM");
+  clblast::RunTests<clblast::TestXasum<double>, double, double>(argc, argv, true, "DASUM");
+  clblast::RunTests<clblast::TestXasum<float2>, float2, float2>(argc, argv, true, "ScASUM");
+  clblast::RunTests<clblast::TestXasum<double2>, double2, double2>(argc, argv, true, "DzASUM");
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/correctness/testblas.h b/test/correctness/testblas.h
index 8181aaf6..13be921a 100644
--- a/test/correctness/testblas.h
+++ b/test/correctness/testblas.h
@@ -153,6 +153,8 @@ void RunTests(int argc, char *argv[], const bool silent, const std::string &name
   auto ap_offsets = std::vector<size_t>{args.ap_offset};
   auto dot_offsets = std::vector<size_t>{args.dot_offset};
   auto nrm2_offsets = std::vector<size_t>{args.nrm2_offset};
+  auto asum_offsets = std::vector<size_t>{args.asum_offset};
+  auto imax_offsets = std::vector<size_t>{args.imax_offset};
   auto alphas = std::vector<U>{args.alpha};
   auto betas = std::vector<U>{args.beta};
   auto x_sizes = std::vector<size_t>{args.x_size};
@@ -193,6 +195,8 @@ void RunTests(int argc, char *argv[], const bool silent, const std::string &name
     if (option == kArgAPOffset) { ap_offsets = tester.kOffsets; }
     if (option == kArgDotOffset) { dot_offsets = tester.kOffsets; }
     if (option == kArgNrm2Offset) { nrm2_offsets = tester.kOffsets; }
+    if (option == kArgAsumOffset) { asum_offsets = tester.kOffsets; }
+    if (option == kArgImaxOffset) { imax_offsets = tester.kOffsets; }
     if (option == kArgAlpha) { alphas = tester.kAlphaValues; }
     if (option == kArgBeta) { betas = tester.kBetaValues; }
 
@@ -233,10 +237,14 @@ void RunTests(int argc, char *argv[], const bool silent, const std::string &name
                                             for (auto &ap_offset: ap_offsets) { r_args.ap_offset = ap_offset;
                                               for (auto &dot_offset: dot_offsets) { r_args.dot_offset = dot_offset;
                                                 for (auto &nrm2_offset: nrm2_offsets) { r_args.nrm2_offset = nrm2_offset;
-                                                  for (auto &alpha: alphas) { r_args.alpha = alpha;
-                                                    for (auto &beta: betas) { r_args.beta = beta;
-                                                      C::SetSizes(r_args);
-                                                      regular_test_vector.push_back(r_args);
+                                                  for (auto &asum_offset: asum_offsets) { r_args.asum_offset = asum_offset;
+                                                    for (auto &imax_offset: imax_offsets) { r_args.imax_offset = imax_offset;
+                                                      for (auto &alpha: alphas) { r_args.alpha = alpha;
+                                                        for (auto &beta: betas) { r_args.beta = beta;
+                                                          C::SetSizes(r_args);
+                                                          regular_test_vector.push_back(r_args);
+                                                        }
+                                                      }
                                                     }
                                                   }
                                                 }
diff --git a/test/performance/client.cc b/test/performance/client.cc
index 56ab8c8d..9aaf1e4e 100644
--- a/test/performance/client.cc
+++ b/test/performance/client.cc
@@ -80,8 +80,11 @@ Arguments<U> Client<T,U>::ParseArguments(int argc, char *argv[], const GetMetric
     if (o == kArgCOffset)  { args.c_offset = GetArgument(argc, argv, help, kArgCOffset, size_t{0}); }
     if (o == kArgAPOffset) { args.ap_offset= GetArgument(argc, argv, help, kArgAPOffset, size_t{0}); }
 
-    // Dot arguments
+    // Scalar result arguments
     if (o == kArgDotOffset)  { args.dot_offset = GetArgument(argc, argv, help, kArgDotOffset, size_t{0}); }
+    if (o == kArgNrm2Offset)  { args.nrm2_offset = GetArgument(argc, argv, help, kArgNrm2Offset, size_t{0}); }
+    if (o == kArgAsumOffset)  { args.asum_offset = GetArgument(argc, argv, help, kArgAsumOffset, size_t{0}); }
+    if (o == kArgImaxOffset)  { args.imax_offset = GetArgument(argc, argv, help, kArgImaxOffset, size_t{0}); }
 
     // Scalar values 
     if (o == kArgAlpha) { args.alpha = GetArgument(argc, argv, help, kArgAlpha, GetScalar<U>()); }
@@ -292,6 +295,8 @@ void Client<T,U>::PrintTableRow(const Arguments<U>& args,
     else if (o == kArgAPOffset) { integers.push_back(args.ap_offset); }
     else if (o == kArgDotOffset) {integers.push_back(args.dot_offset); }
     else if (o == kArgNrm2Offset){integers.push_back(args.nrm2_offset); }
+    else if (o == kArgAsumOffset){integers.push_back(args.asum_offset); }
+    else if (o == kArgImaxOffset){integers.push_back(args.imax_offset); }
   }
   auto strings = std::vector<std::string>{};
   for (auto &o: options_) {
diff --git a/test/performance/routines/level1/xamax.cc b/test/performance/routines/level1/xamax.cc
new file mode 100644
index 00000000..85caa483
--- /dev/null
+++ b/test/performance/routines/level1/xamax.cc
@@ -0,0 +1,35 @@
+
+// =================================================================================================
+// 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>
+//
+// =================================================================================================
+
+#include "performance/client.h"
+#include "routines/level1/xamax.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, clblast::Precision::kSingle)) {
+    case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode");
+    case clblast::Precision::kSingle:
+      clblast::RunClient<clblast::TestXamax<float>, float, float>(argc, argv); break;
+    case clblast::Precision::kDouble:
+      clblast::RunClient<clblast::TestXamax<double>, double, double>(argc, argv); break;
+    case clblast::Precision::kComplexSingle:
+      clblast::RunClient<clblast::TestXamax<float2>, float2, float2>(argc, argv); break;
+    case clblast::Precision::kComplexDouble:
+      clblast::RunClient<clblast::TestXamax<double2>, double2, double2>(argc, argv); break;
+  }
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/performance/routines/level1/xasum.cc b/test/performance/routines/level1/xasum.cc
new file mode 100644
index 00000000..2680966e
--- /dev/null
+++ b/test/performance/routines/level1/xasum.cc
@@ -0,0 +1,35 @@
+
+// =================================================================================================
+// 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>
+//
+// =================================================================================================
+
+#include "performance/client.h"
+#include "routines/level1/xasum.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, clblast::Precision::kSingle)) {
+    case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode");
+    case clblast::Precision::kSingle:
+      clblast::RunClient<clblast::TestXasum<float>, float, float>(argc, argv); break;
+    case clblast::Precision::kDouble:
+      clblast::RunClient<clblast::TestXasum<double>, double, double>(argc, argv); break;
+    case clblast::Precision::kComplexSingle:
+      clblast::RunClient<clblast::TestXasum<float2>, float2, float2>(argc, argv); break;
+    case clblast::Precision::kComplexDouble:
+      clblast::RunClient<clblast::TestXasum<double2>, double2, double2>(argc, argv); break;
+  }
+  return 0;
+}
+
+// =================================================================================================
diff --git a/test/routines/level1/xamax.h b/test/routines/level1/xamax.h
new file mode 100644
index 00000000..7b404dc3
--- /dev/null
+++ b/test/routines/level1/xamax.h
@@ -0,0 +1,139 @@
+
+// =================================================================================================
+// 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 Xamax 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_XAMAX_H_
+#define CLBLAST_TEST_ROUTINES_XAMAX_H_
+
+#include <vector>
+#include <string>
+
+#ifdef CLBLAST_REF_CLBLAS
+  #include "wrapper_clblas.h"
+#endif
+#ifdef CLBLAST_REF_CBLAS
+  #include "wrapper_cblas.h"
+#endif
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class TestXamax {
+ 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, kArgImaxOffset};
+  }
+
+  // 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 GetSizeImax(const Arguments<T> &args) {
+    return 1 + args.imax_offset;
+  }
+
+  // Describes how to set the sizes of all the buffers
+  static void SetSizes(Arguments<T> &args) {
+    args.x_size = GetSizeX(args);
+    args.scalar_size = GetSizeImax(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, Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = Amax<T>(args.n,
+                          buffers.scalar(), args.imax_offset,
+                          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)
+  #ifdef CLBLAST_REF_CLBLAS
+    static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+      auto queue_plain = queue();
+      auto event = cl_event{};
+      auto status = clblasXamax<T>(args.n,
+                                   buffers.scalar(), args.imax_offset,
+                                   buffers.x_vec(), args.x_offset, args.x_inc,
+                                   1, &queue_plain, 0, nullptr, &event);
+      clWaitForEvents(1, &event);
+      return static_cast<StatusCode>(status);
+    }
+  #endif
+
+  // Describes how to run the CPU BLAS routine (for correctness/performance comparison)
+  #ifdef CLBLAST_REF_CBLAS
+    static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+      std::vector<T> scalar_cpu(args.scalar_size, static_cast<T>(0));
+      std::vector<T> x_vec_cpu(args.x_size, static_cast<T>(0));
+      buffers.scalar.Read(queue, args.scalar_size, scalar_cpu);
+      buffers.x_vec.Read(queue, args.x_size, x_vec_cpu);
+      cblasXamax(args.n,
+                 scalar_cpu, args.imax_offset,
+                 x_vec_cpu, args.x_offset, args.x_inc);
+      buffers.scalar.Write(queue, args.scalar_size, scalar_cpu);
+      return StatusCode::kSuccess;
+    }
+  #endif
+
+  // 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.scalar_size, static_cast<T>(0));
+    buffers.scalar.Read(queue, args.scalar_size, result);
+    return result;
+  }
+
+  // Describes how to compute the indices of the result buffer
+  static size_t ResultID1(const Arguments<T> &) { return 1; } // N/A for this routine
+  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, const size_t) {
+    return args.imax_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 ((args.n) + 1) * sizeof(T);
+  }
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_TEST_ROUTINES_XAMAX_H_
+#endif
diff --git a/test/routines/level1/xasum.h b/test/routines/level1/xasum.h
new file mode 100644
index 00000000..6eae3c83
--- /dev/null
+++ b/test/routines/level1/xasum.h
@@ -0,0 +1,139 @@
+
+// =================================================================================================
+// 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 Xasum 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_XASUM_H_
+#define CLBLAST_TEST_ROUTINES_XASUM_H_
+
+#include <vector>
+#include <string>
+
+#ifdef CLBLAST_REF_CLBLAS
+  #include "wrapper_clblas.h"
+#endif
+#ifdef CLBLAST_REF_CBLAS
+  #include "wrapper_cblas.h"
+#endif
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class TestXasum {
+ 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, kArgAsumOffset};
+  }
+
+  // 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 GetSizeAsum(const Arguments<T> &args) {
+    return 1 + args.asum_offset;
+  }
+
+  // Describes how to set the sizes of all the buffers
+  static void SetSizes(Arguments<T> &args) {
+    args.x_size = GetSizeX(args);
+    args.scalar_size = GetSizeAsum(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, Buffers<T> &buffers, Queue &queue) {
+    auto queue_plain = queue();
+    auto event = cl_event{};
+    auto status = Asum<T>(args.n,
+                          buffers.scalar(), args.asum_offset,
+                          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)
+  #ifdef CLBLAST_REF_CLBLAS
+    static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+      auto queue_plain = queue();
+      auto event = cl_event{};
+      auto status = clblasXasum<T>(args.n,
+                                   buffers.scalar(), args.asum_offset,
+                                   buffers.x_vec(), args.x_offset, args.x_inc,
+                                   1, &queue_plain, 0, nullptr, &event);
+      clWaitForEvents(1, &event);
+      return static_cast<StatusCode>(status);
+    }
+  #endif
+
+  // Describes how to run the CPU BLAS routine (for correctness/performance comparison)
+  #ifdef CLBLAST_REF_CBLAS
+    static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+      std::vector<T> scalar_cpu(args.scalar_size, static_cast<T>(0));
+      std::vector<T> x_vec_cpu(args.x_size, static_cast<T>(0));
+      buffers.scalar.Read(queue, args.scalar_size, scalar_cpu);
+      buffers.x_vec.Read(queue, args.x_size, x_vec_cpu);
+      cblasXasum(args.n,
+                 scalar_cpu, args.asum_offset,
+                 x_vec_cpu, args.x_offset, args.x_inc);
+      buffers.scalar.Write(queue, args.scalar_size, scalar_cpu);
+      return StatusCode::kSuccess;
+    }
+  #endif
+
+  // 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.scalar_size, static_cast<T>(0));
+    buffers.scalar.Read(queue, args.scalar_size, result);
+    return result;
+  }
+
+  // Describes how to compute the indices of the result buffer
+  static size_t ResultID1(const Arguments<T> &) { return 1; } // N/A for this routine
+  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, const size_t) {
+    return args.asum_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 ((args.n) + 1) * sizeof(T);
+  }
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_TEST_ROUTINES_XASUM_H_
+#endif
diff --git a/test/wrapper_cblas.h b/test/wrapper_cblas.h
index dec272b0..994b48b1 100644
--- a/test/wrapper_cblas.h
+++ b/test/wrapper_cblas.h
@@ -345,6 +345,58 @@ void cblasXnrm2(const size_t n,
                                          reinterpret_cast<const double*>(&x_buffer[x_offset]), static_cast<int>(x_inc));
 }
 
+// Forwards the Netlib BLAS calls for SASUM/DASUM/ScASUM/DzASUM
+void cblasXasum(const size_t n,
+                std::vector<float>& asum_buffer, const size_t asum_offset,
+                const std::vector<float>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  asum_buffer[asum_offset] = cblas_sasum(n,
+                                         &x_buffer[x_offset], static_cast<int>(x_inc));
+}
+void cblasXasum(const size_t n,
+                std::vector<double>& asum_buffer, const size_t asum_offset,
+                const std::vector<double>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  asum_buffer[asum_offset] = cblas_dasum(n,
+                                         &x_buffer[x_offset], static_cast<int>(x_inc));
+}
+void cblasXasum(const size_t n,
+                std::vector<float2>& asum_buffer, const size_t asum_offset,
+                const std::vector<float2>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  asum_buffer[asum_offset] = cblas_scasum(n,
+                                         reinterpret_cast<const float*>(&x_buffer[x_offset]), static_cast<int>(x_inc));
+}
+void cblasXasum(const size_t n,
+                std::vector<double2>& asum_buffer, const size_t asum_offset,
+                const std::vector<double2>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  asum_buffer[asum_offset] = cblas_dzasum(n,
+                                         reinterpret_cast<const double*>(&x_buffer[x_offset]), static_cast<int>(x_inc));
+}
+
+// Forwards the Netlib BLAS calls for iSAMAX/iDAMAX/iCAMAX/iZAMAX
+void cblasXamax(const size_t n,
+                std::vector<float>& imax_buffer, const size_t imax_offset,
+                const std::vector<float>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  imax_buffer[imax_offset] = cblas_isamax(n,
+                                         &x_buffer[x_offset], static_cast<int>(x_inc));
+}
+void cblasXamax(const size_t n,
+                std::vector<double>& imax_buffer, const size_t imax_offset,
+                const std::vector<double>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  imax_buffer[imax_offset] = cblas_idamax(n,
+                                         &x_buffer[x_offset], static_cast<int>(x_inc));
+}
+void cblasXamax(const size_t n,
+                std::vector<float2>& imax_buffer, const size_t imax_offset,
+                const std::vector<float2>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  imax_buffer[imax_offset] = cblas_icamax(n,
+                                         reinterpret_cast<const float*>(&x_buffer[x_offset]), static_cast<int>(x_inc));
+}
+void cblasXamax(const size_t n,
+                std::vector<double2>& imax_buffer, const size_t imax_offset,
+                const std::vector<double2>& x_buffer, const size_t x_offset, const size_t x_inc) {
+  imax_buffer[imax_offset] = cblas_izamax(n,
+                                         reinterpret_cast<const double*>(&x_buffer[x_offset]), static_cast<int>(x_inc));
+}
+
 // =================================================================================================
 // BLAS level-2 (matrix-vector) routines
 // =================================================================================================
diff --git a/test/wrapper_clblas.h b/test/wrapper_clblas.h
index 89b708b8..a44466c6 100644
--- a/test/wrapper_clblas.h
+++ b/test/wrapper_clblas.h
@@ -558,6 +558,142 @@ clblasStatus clblasXnrm2<double2>(const size_t n,
                      num_queues, queues, num_wait_events, wait_events, events);
 }
 
+// Forwards the clBLAS calls for SASUM/DASUM/ScASUM/DzASUM
+template <typename T>
+clblasStatus clblasXasum(const size_t n,
+                         cl_mem asum_buffer, const size_t asum_offset,
+                         const cl_mem x_buffer, const size_t x_offset, const 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);
+template <>
+clblasStatus clblasXasum<float>(const size_t n,
+                                cl_mem asum_buffer, const size_t asum_offset,
+                                const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<float>(context, n);
+  return clblasSasum(n,
+                     asum_buffer, asum_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+template <>
+clblasStatus clblasXasum<double>(const size_t n,
+                                 cl_mem asum_buffer, const size_t asum_offset,
+                                 const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<double>(context, n);
+  return clblasDasum(n,
+                     asum_buffer, asum_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+template <>
+clblasStatus clblasXasum<float2>(const size_t n,
+                                 cl_mem asum_buffer, const size_t asum_offset,
+                                 const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<float2>(context, n);
+  return clblasScasum(n,
+                     asum_buffer, asum_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+template <>
+clblasStatus clblasXasum<double2>(const size_t n,
+                                  cl_mem asum_buffer, const size_t asum_offset,
+                                  const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<double2>(context, n);
+  return clblasDzasum(n,
+                     asum_buffer, asum_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+
+// Forwards the clBLAS calls for iSAMAX/iDAMAX/iCAMAX/iZAMAX
+template <typename T>
+clblasStatus clblasXamax(const size_t n,
+                         cl_mem imax_buffer, const size_t imax_offset,
+                         const cl_mem x_buffer, const size_t x_offset, const 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);
+template <>
+clblasStatus clblasXamax<float>(const size_t n,
+                                cl_mem imax_buffer, const size_t imax_offset,
+                                const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<float>(context, 2*n);
+  return clblasiSamax(n,
+                     imax_buffer, imax_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+template <>
+clblasStatus clblasXamax<double>(const size_t n,
+                                 cl_mem imax_buffer, const size_t imax_offset,
+                                 const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<double>(context, 2*n);
+  return clblasiDamax(n,
+                     imax_buffer, imax_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+template <>
+clblasStatus clblasXamax<float2>(const size_t n,
+                                 cl_mem imax_buffer, const size_t imax_offset,
+                                 const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<float2>(context, 2*n);
+  return clblasiCamax(n,
+                     imax_buffer, imax_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+template <>
+clblasStatus clblasXamax<double2>(const size_t n,
+                                  cl_mem imax_buffer, const size_t imax_offset,
+                                  const cl_mem x_buffer, const size_t x_offset, const 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 queue = Queue(queues[0]);
+  auto context = queue.GetContext();
+  auto scratch_buffer = Buffer<double2>(context, 2*n);
+  return clblasiZamax(n,
+                     imax_buffer, imax_offset,
+                     x_buffer, x_offset, static_cast<int>(x_inc),
+                     scratch_buffer(),
+                     num_queues, queues, num_wait_events, wait_events, events);
+}
+
 // =================================================================================================
 // BLAS level-2 (matrix-vector) routines
 // =================================================================================================