Added XGER routine, kernel, and tuner

3 files changed, 147 insertions, 0 deletions

diff --git a/include/internal/database.h b/include/internal/database.h
index 08e449fa..f26e354c 100644
--- a/include/internal/database.h
+++ b/include/internal/database.h
@@ -73,6 +73,7 @@ class Database {
   static const DatabaseEntry XaxpySingle, XaxpyDouble, XaxpyComplexSingle, XaxpyComplexDouble;
   static const DatabaseEntry XdotSingle, XdotDouble, XdotComplexSingle, XdotComplexDouble;
   static const DatabaseEntry XgemvSingle, XgemvDouble, XgemvComplexSingle, XgemvComplexDouble;
+  static const DatabaseEntry XgerSingle, XgerDouble, XgerComplexSingle, XgerComplexDouble;
   static const DatabaseEntry XgemmSingle, XgemmDouble, XgemmComplexSingle, XgemmComplexDouble;
   static const DatabaseEntry CopySingle, CopyDouble, CopyComplexSingle, CopyComplexDouble;
   static const DatabaseEntry PadSingle, PadDouble, PadComplexSingle, PadComplexDouble;
diff --git a/include/internal/database/xger.h b/include/internal/database/xger.h
new file mode 100644
index 00000000..c9cfb6cd
--- /dev/null
+++ b/include/internal/database/xger.h
@@ -0,0 +1,88 @@
+
+// =================================================================================================
+// 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):
+//   Database generator <database.py>
+//
+// This file populates the database with best-found tuning parameters for the 'Xger' kernels.
+//
+// =================================================================================================
+
+namespace clblast {
+// =================================================================================================
+
+const Database::DatabaseEntry Database::XgerSingle = {
+  "Xger", Precision::kSingle, {
+    { // Intel CPUs
+      kDeviceTypeCPU, "Intel", {
+        { "Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz",        { {"WGS1",128}, {"WGS2",2}, {"WPT",4} } },
+        { "default",                                         { {"WGS1",128}, {"WGS2",2}, {"WPT",4} } },
+      }
+    },
+    { // Default
+      kDeviceTypeAll, "default", {
+        { "default",                                         { {"WGS1",128}, {"WGS2",2}, {"WPT",4} } },
+      }
+    },
+  }
+};
+
+// =================================================================================================
+
+const Database::DatabaseEntry Database::XgerComplexSingle = {
+  "Xger", Precision::kComplexSingle, {
+    { // Intel CPUs
+      kDeviceTypeCPU, "Intel", {
+        { "Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz",        { {"WGS1",512}, {"WGS2",8}, {"WPT",2} } },
+        { "default",                                         { {"WGS1",512}, {"WGS2",8}, {"WPT",2} } },
+      }
+    },
+    { // Default
+      kDeviceTypeAll, "default", {
+        { "default",                                         { {"WGS1",512}, {"WGS2",8}, {"WPT",2} } },
+      }
+    },
+  }
+};
+
+// =================================================================================================
+
+const Database::DatabaseEntry Database::XgerDouble = {
+  "Xger", Precision::kDouble, {
+    { // Intel CPUs
+      kDeviceTypeCPU, "Intel", {
+        { "Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz",        { {"WGS1",512}, {"WGS2",16}, {"WPT",1} } },
+        { "default",                                         { {"WGS1",512}, {"WGS2",16}, {"WPT",1} } },
+      }
+    },
+    { // Default
+      kDeviceTypeAll, "default", {
+        { "default",                                         { {"WGS1",512}, {"WGS2",16}, {"WPT",1} } },
+      }
+    },
+  }
+};
+
+// =================================================================================================
+
+const Database::DatabaseEntry Database::XgerComplexDouble = {
+  "Xger", Precision::kComplexDouble, {
+    { // Intel CPUs
+      kDeviceTypeCPU, "Intel", {
+        { "Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz",        { {"WGS1",512}, {"WGS2",1}, {"WPT",1} } },
+        { "default",                                         { {"WGS1",512}, {"WGS2",1}, {"WPT",1} } },
+      }
+    },
+    { // Default
+      kDeviceTypeAll, "default", {
+        { "default",                                         { {"WGS1",512}, {"WGS2",1}, {"WPT",1} } },
+      }
+    },
+  }
+};
+
+// =================================================================================================
+} // namespace clblast
diff --git a/include/internal/routines/level2/xger.h b/include/internal/routines/level2/xger.h
new file mode 100644
index 00000000..45ecea10
--- /dev/null
+++ b/include/internal/routines/level2/xger.h
@@ -0,0 +1,58 @@
+
+// =================================================================================================
+// 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 Xger routine. The precision is implemented using a template argument.
+//
+// =================================================================================================
+
+#ifndef CLBLAST_ROUTINES_XGER_H_
+#define CLBLAST_ROUTINES_XGER_H_
+
+#include "internal/routine.h"
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class Xger: public Routine<T> {
+ public:
+
+  // Members and methods from the base class
+  using Routine<T>::db_;
+  using Routine<T>::source_string_;
+  using Routine<T>::queue_;
+  using Routine<T>::GetProgramFromCache;
+  using Routine<T>::TestVectorX;
+  using Routine<T>::TestVectorY;
+  using Routine<T>::TestMatrixA;
+  using Routine<T>::RunKernel;
+  using Routine<T>::ErrorIn;
+
+  // Constructor
+  Xger(Queue &queue, Event &event, const std::string &name = "GER");
+
+  // Templated-precision implementation of the routine
+  StatusCode DoGer(const Layout layout,
+                   const size_t m, 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);
+
+ private:
+  // Static variable to get the precision
+  const static Precision precision_;
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_ROUTINES_XGER_H_
+#endif