Merge branch 'level1_routines' into development

4 files changed, 84 insertions, 2 deletions

diff --git a/include/clblast.h b/include/clblast.h
index ac16188f..5e5c5a46 100644
--- a/include/clblast.h
+++ b/include/clblast.h
@@ -174,6 +174,13 @@ StatusCode Dotc(const size_t n,
                 const cl_mem y_buffer, const size_t y_offset, const size_t y_inc,
                 cl_command_queue* queue, cl_event* event = nullptr);
 
+// Euclidian norm of a vector: SNRM2/DNRM2/ScNRM2/DzNRM2
+template <typename T>
+StatusCode Nrm2(const size_t n,
+                cl_mem nrm2_buffer, const size_t nrm2_offset,
+                const cl_mem x_buffer, const size_t x_offset, const size_t x_inc,
+                cl_command_queue* queue, cl_event* event = nullptr);
+
 // =================================================================================================
 // BLAS level-2 (matrix-vector) routines
 // =================================================================================================
diff --git a/include/clblast_c.h b/include/clblast_c.h
index a5563951..dcb3ae3a 100644
--- a/include/clblast_c.h
+++ b/include/clblast_c.h
@@ -260,6 +260,24 @@ StatusCode PUBLIC_API CLBlastZdotc(const size_t n,
                                    const cl_mem y_buffer, const size_t y_offset, const size_t y_inc,
                                    cl_command_queue* queue, cl_event* event);
 
+// Euclidian norm of a vector: SNRM2/DNRM2/ScNRM2/DzNRM2
+StatusCode PUBLIC_API CLBlastSnrm2(const size_t n,
+                                   cl_mem nrm2_buffer, const size_t nrm2_offset,
+                                   const cl_mem x_buffer, const size_t x_offset, const size_t x_inc,
+                                   cl_command_queue* queue, cl_event* event);
+StatusCode PUBLIC_API CLBlastDnrm2(const size_t n,
+                                   cl_mem nrm2_buffer, const size_t nrm2_offset,
+                                   const cl_mem x_buffer, const size_t x_offset, const size_t x_inc,
+                                   cl_command_queue* queue, cl_event* event);
+StatusCode PUBLIC_API CLBlastScnrm2(const size_t n,
+                                   cl_mem nrm2_buffer, const size_t nrm2_offset,
+                                   const cl_mem x_buffer, const size_t x_offset, const size_t x_inc,
+                                   cl_command_queue* queue, cl_event* event);
+StatusCode PUBLIC_API CLBlastDznrm2(const size_t n,
+                                   cl_mem nrm2_buffer, const size_t nrm2_offset,
+                                   const cl_mem x_buffer, const size_t x_offset, const size_t x_inc,
+                                   cl_command_queue* queue, cl_event* event);
+
 // =================================================================================================
 // BLAS level-2 (matrix-vector) routines
 // =================================================================================================
diff --git a/include/internal/routines/level1/xnrm2.h b/include/internal/routines/level1/xnrm2.h
new file mode 100644
index 00000000..b3fffef6
--- /dev/null
+++ b/include/internal/routines/level1/xnrm2.h
@@ -0,0 +1,55 @@
+
+// =================================================================================================
+// 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 Xnrm2 routine. The precision is implemented using a template argument.
+//
+// =================================================================================================
+
+#ifndef CLBLAST_ROUTINES_XNRM2_H_
+#define CLBLAST_ROUTINES_XNRM2_H_
+
+#include "internal/routine.h"
+
+namespace clblast {
+// =================================================================================================
+
+// See comment at top of file for a description of the class
+template <typename T>
+class Xnrm2: 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>::context_;
+  using Routine<T>::GetProgramFromCache;
+  using Routine<T>::TestVectorX;
+  using Routine<T>::TestVectorDot;
+  using Routine<T>::RunKernel;
+  using Routine<T>::ErrorIn;
+
+  // Constructor
+  Xnrm2(Queue &queue, Event &event, const std::string &name = "NRM2");
+
+  // Templated-precision implementation of the routine
+  StatusCode DoNrm2(const size_t n,
+                    const Buffer<T> &nrm2_buffer, const size_t nrm2_offset,
+                    const Buffer<T> &x_buffer, const size_t x_offset, const size_t x_inc);
+
+ private:
+  // Static variable to get the precision
+  const static Precision precision_;
+};
+
+// =================================================================================================
+} // namespace clblast
+
+// CLBLAST_ROUTINES_XNRM2_H_
+#endif
diff --git a/include/internal/utilities.h b/include/internal/utilities.h
index b6307a85..35f76722 100644
--- a/include/internal/utilities.h
+++ b/include/internal/utilities.h
@@ -61,6 +61,7 @@ constexpr auto kArgBOffset = "offb";
 constexpr auto kArgCOffset = "offc";
 constexpr auto kArgAPOffset = "offap";
 constexpr auto kArgDotOffset = "offdot";
+constexpr auto kArgNrm2Offset = "offnrm2";
 constexpr auto kArgAlpha = "alpha";
 constexpr auto kArgBeta = "beta";
 
@@ -113,6 +114,7 @@ struct Arguments {
   size_t c_offset = 0;
   size_t ap_offset = 0;
   size_t dot_offset = 0;
+  size_t nrm2_offset = 0;
   T alpha = T{1.0};
   T beta = T{1.0};
   size_t x_size = 1;
@@ -121,7 +123,7 @@ struct Arguments {
   size_t b_size = 1;
   size_t c_size = 1;
   size_t ap_size = 1;
-  size_t dot_size = 1;
+  size_t scalar_size = 1;
   // Tuner-specific arguments
   double fraction = 1.0;
   // Client-specific arguments
@@ -149,7 +151,7 @@ struct Buffers {
   Buffer<T> b_mat;
   Buffer<T> c_mat;
   Buffer<T> ap_mat;
-  Buffer<T> dot;
+  Buffer<T> scalar;
 };
 
 // =================================================================================================