Added preliminary support for xHER2 and xSYR2 routines

5 files changed, 169 insertions, 3 deletions

diff --git a/src/routines/level2/xher2.cc b/src/routines/level2/xher2.cc
new file mode 100644
index 00000000..63144f77
--- /dev/null
+++ b/src/routines/level2/xher2.cc
@@ -0,0 +1,114 @@
+
+// =================================================================================================
+// 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 Xher2 class (see the header for information about the class).
+//
+// =================================================================================================
+
+#include "internal/routines/level2/xher2.h"
+
+#include <string>
+
+namespace clblast {
+// =================================================================================================
+
+// Specific implementations to get the memory-type based on a template argument
+template <> const Precision Xher2<float>::precision_ = Precision::kSingle;
+template <> const Precision Xher2<double>::precision_ = Precision::kDouble;
+template <> const Precision Xher2<float2>::precision_ = Precision::kComplexSingle;
+template <> const Precision Xher2<double2>::precision_ = Precision::kComplexDouble;
+
+// =================================================================================================
+
+// Constructor: forwards to base class constructor
+template <typename T>
+Xher2<T>::Xher2(Queue &queue, Event &event, const std::string &name):
+    Routine<T>(queue, event, name, {"Xger"}, precision_) {
+  source_string_ =
+    #include "../../kernels/level2/level2.opencl"
+    #include "../../kernels/level2/xher2.opencl"
+  ;
+}
+
+// =================================================================================================
+
+// The main routine
+template <typename T>
+StatusCode Xher2<T>::DoHer2(const Layout layout, const Triangle triangle,
+                            const size_t n,
+                            const T alpha,
+                            const Buffer<T> &x_buffer, const size_t x_offset, const size_t x_inc,
+                            const Buffer<T> &y_buffer, const size_t y_offset, const size_t y_inc,
+                            const Buffer<T> &a_buffer, const size_t a_offset, const size_t a_ld,
+                            const bool packed) {
+
+  // Makes sure the dimensions are larger than zero
+  if (n == 0) { return StatusCode::kInvalidDimension; }
+
+  // The data is either in the upper or lower triangle
+  const auto is_upper = ((triangle == Triangle::kUpper && layout != Layout::kRowMajor) ||
+                         (triangle == Triangle::kLower && layout == Layout::kRowMajor));
+  const auto is_rowmajor = (layout == Layout::kRowMajor);
+
+  // Tests the matrix and the vectors for validity
+  auto status = StatusCode::kSuccess;
+  if (packed) { status = TestMatrixAP(n, a_buffer, a_offset, sizeof(T)); }
+  else { status = TestMatrixA(n, n, a_buffer, a_offset, a_ld, sizeof(T)); }
+  if (ErrorIn(status)) { return status; }
+  status = TestVectorX(n, x_buffer, x_offset, x_inc, sizeof(T));
+  if (ErrorIn(status)) { return status; }
+  status = TestVectorY(n, y_buffer, y_offset, y_inc, sizeof(T));
+  if (ErrorIn(status)) { return status; }
+
+  // Retrieves the Xgemv kernel from the compiled binary
+  try {
+    auto& program = GetProgramFromCache();
+    auto kernel = Kernel(program, "Xher2");
+
+    // Sets the kernel arguments
+    kernel.SetArgument(0, static_cast<int>(n));
+    kernel.SetArgument(1, alpha);
+    kernel.SetArgument(2, x_buffer());
+    kernel.SetArgument(3, static_cast<int>(x_offset));
+    kernel.SetArgument(4, static_cast<int>(x_inc));
+    kernel.SetArgument(5, y_buffer());
+    kernel.SetArgument(6, static_cast<int>(y_offset));
+    kernel.SetArgument(7, static_cast<int>(y_inc));
+    kernel.SetArgument(8, a_buffer());
+    kernel.SetArgument(9, static_cast<int>(a_offset));
+    kernel.SetArgument(10, static_cast<int>(a_ld));
+    kernel.SetArgument(11, static_cast<int>(is_upper));
+    kernel.SetArgument(12, static_cast<int>(is_rowmajor));
+
+    // Launches the kernel
+    auto global_one = CeilDiv(Ceil(n, db_["WGS1"]), db_["WPT"]);
+    auto global_two = CeilDiv(Ceil(n, db_["WGS2"]), db_["WPT"]);
+    auto global = std::vector<size_t>{global_one, global_two};
+    auto local = std::vector<size_t>{db_["WGS1"], db_["WGS2"]};
+    status = RunKernel(kernel, global, local);
+    if (ErrorIn(status)) { return status; }
+
+    // Waits for all kernels to finish
+    queue_.Finish();
+
+    // Succesfully finished the computation
+    return StatusCode::kSuccess;
+  } catch (...) { return StatusCode::kInvalidKernel; }
+}
+
+// =================================================================================================
+
+// Compiles the templated class
+template class Xher2<float>;
+template class Xher2<double>;
+template class Xher2<float2>;
+template class Xher2<double2>;
+
+// =================================================================================================
+} // namespace clblast
diff --git a/src/routines/level2/xhpr.cc b/src/routines/level2/xhpr.cc
index abe00669..24d7ae95 100644
--- a/src/routines/level2/xhpr.cc
+++ b/src/routines/level2/xhpr.cc
@@ -34,7 +34,7 @@ StatusCode Xhpr<T,U>::DoHpr(const Layout layout, const Triangle triangle,
                             const Buffer<T> &x_buffer, const size_t x_offset, const size_t x_inc,
                             const Buffer<T> &ap_buffer, const size_t ap_offset) {
 
-  // 
+  // Specific Xhpr functionality is implemented in the kernel using defines
   return DoHer(layout, triangle, n, alpha,
                x_buffer, x_offset, x_inc,
                ap_buffer, ap_offset, n, true);
diff --git a/src/routines/level2/xspr.cc b/src/routines/level2/xspr.cc
index 5159ad50..7ef41fba 100644
--- a/src/routines/level2/xspr.cc
+++ b/src/routines/level2/xspr.cc
@@ -34,7 +34,7 @@ StatusCode Xspr<T>::DoSpr(const Layout layout, const Triangle triangle,
                           const Buffer<T> &x_buffer, const size_t x_offset, const size_t x_inc,
                           const Buffer<T> &ap_buffer, const size_t ap_offset) {
 
-  // 
+  // Specific Xspr functionality is implemented in the kernel using defines
   return DoHer(layout, triangle, n, alpha,
                x_buffer, x_offset, x_inc,
                ap_buffer, ap_offset, n, true);
diff --git a/src/routines/level2/xsyr.cc b/src/routines/level2/xsyr.cc
index 755fde0d..c01fa2d3 100644
--- a/src/routines/level2/xsyr.cc
+++ b/src/routines/level2/xsyr.cc
@@ -34,7 +34,7 @@ StatusCode Xsyr<T>::DoSyr(const Layout layout, const Triangle triangle,
                           const Buffer<T> &x_buffer, const size_t x_offset, const size_t x_inc,
                           const Buffer<T> &a_buffer, const size_t a_offset, const size_t a_ld) {
 
-  // 
+  // Specific Xsyr functionality is implemented in the kernel using defines
   return DoHer(layout, triangle, n, alpha,
                x_buffer, x_offset, x_inc,
                a_buffer, a_offset, a_ld);
diff --git a/src/routines/level2/xsyr2.cc b/src/routines/level2/xsyr2.cc
new file mode 100644
index 00000000..6db55085
--- /dev/null
+++ b/src/routines/level2/xsyr2.cc
@@ -0,0 +1,52 @@
+
+// =================================================================================================
+// 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 Xsyr2 class (see the header for information about the class).
+//
+// =================================================================================================
+
+#include "internal/routines/level2/xsyr2.h"
+
+#include <string>
+
+namespace clblast {
+// =================================================================================================
+
+// Constructor: forwards to base class constructor
+template <typename T>
+Xsyr2<T>::Xsyr2(Queue &queue, Event &event, const std::string &name):
+    Xher2<T>(queue, event, name) {
+}
+
+// =================================================================================================
+
+// The main routine
+template <typename T>
+StatusCode Xsyr2<T>::DoSyr2(const Layout layout, const Triangle triangle,
+                            const size_t n,
+                            const T alpha,
+                            const Buffer<T> &x_buffer, const size_t x_offset, const size_t x_inc,
+                            const Buffer<T> &y_buffer, const size_t y_offset, const size_t y_inc,
+                            const Buffer<T> &a_buffer, const size_t a_offset, const size_t a_ld) {
+
+  // Specific Xsyr2 functionality is implemented in the kernel using defines
+  return DoHer2(layout, triangle, n, alpha,
+                x_buffer, x_offset, x_inc,
+                y_buffer, y_offset, y_inc,
+                a_buffer, a_offset, a_ld);
+}
+
+// =================================================================================================
+
+// Compiles the templated class
+template class Xsyr2<float>;
+template class Xsyr2<double>;
+
+// =================================================================================================
+} // namespace clblast