4 files changed, 7 insertions, 25 deletions
diff --git a/src/routines/level2/xgemv.cpp b/src/routines/level2/xgemv.cpp
index 21fb397c..4e32ba41 100644
--- a/src/routines/level2/xgemv.cpp
+++ b/src/routines/level2/xgemv.cpp
@@ -22,7 +22,7 @@ namespace clblast {
 // Constructor: forwards to base class constructor
 template <typename T>
 Xgemv<T>::Xgemv(Queue &queue, EventPointer event, const std::string &name):
-    Routine(queue, event, name, {"Pad", "Xgemv"}, PrecisionValue<T>()) {
+    Routine(queue, event, name, {"Pad", "Xgemv", "XgemvFast", "XgemvFastRot"}, PrecisionValue<T>()) {
   source_string_ =
     #include "../../kernels/level2/xgemv.opencl"
     #include "../../kernels/level2/xgemv_fast.opencl"
@@ -122,16 +122,10 @@ StatusCode Xgemv<T>::MatVec(const Layout layout, const Transpose a_transpose,
   }
   if (fast_kernel_rot) {
     kernel_name = "XgemvFastRot";
-    global_size = m_real / db_["WPT3"];
+    global_size = m_real;
     local_size = db_["WGS3"];
   }
 
-  // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
-  auto alpha_buffer = Buffer<T>(context_, 1);
-  auto beta_buffer = Buffer<T>(context_, 1);
-  alpha_buffer.Write(queue_, 1, &alpha);
-  beta_buffer.Write(queue_, 1, &beta);
-
   // Retrieves the Xgemv kernel from the compiled binary
   try {
     const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -140,8 +134,8 @@ StatusCode Xgemv<T>::MatVec(const Layout layout, const Transpose a_transpose,
     // Sets the kernel arguments
     kernel.SetArgument(0, static_cast<int>(m_real));
     kernel.SetArgument(1, static_cast<int>(n_real));
-    kernel.SetArgument(2, alpha_buffer());
-    kernel.SetArgument(3, beta_buffer());
+    kernel.SetArgument(2, GetRealArg(alpha));
+    kernel.SetArgument(3, GetRealArg(beta));
     kernel.SetArgument(4, static_cast<int>(a_rotated));
     kernel.SetArgument(5, a_buffer());
     kernel.SetArgument(6, static_cast<int>(a_offset));
diff --git a/src/routines/level2/xger.cpp b/src/routines/level2/xger.cpp
index 353047d2..29cffe0c 100644
--- a/src/routines/level2/xger.cpp
+++ b/src/routines/level2/xger.cpp
@@ -56,10 +56,6 @@ StatusCode Xger<T>::DoGer(const Layout layout,
   status = TestVectorY(n, y_buffer, y_offset, y_inc);
   if (ErrorIn(status)) { return status; }
 
-  // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
-  auto alpha_buffer = Buffer<T>(context_, 1);
-  alpha_buffer.Write(queue_, 1, &alpha);
-
   // Retrieves the kernel from the compiled binary
   try {
     const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -68,7 +64,7 @@ StatusCode Xger<T>::DoGer(const Layout layout,
     // Sets the kernel arguments
     kernel.SetArgument(0, static_cast<int>(a_one));
     kernel.SetArgument(1, static_cast<int>(a_two));
-    kernel.SetArgument(2, alpha_buffer());
+    kernel.SetArgument(2, GetRealArg(alpha));
     kernel.SetArgument(3, x_buffer());
     kernel.SetArgument(4, static_cast<int>(x_offset));
     kernel.SetArgument(5, static_cast<int>(x_inc));
diff --git a/src/routines/level2/xher.cpp b/src/routines/level2/xher.cpp
index ed8ba9e9..6dd95938 100644
--- a/src/routines/level2/xher.cpp
+++ b/src/routines/level2/xher.cpp
@@ -70,10 +70,6 @@ StatusCode Xher<T,U>::DoHer(const Layout layout, const Triangle triangle,
   // Creates a matching version of alpha
   const auto matching_alpha = GetAlpha(alpha);
 
-  // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
-  auto alpha_buffer = Buffer<T>(context_, 1);
-  alpha_buffer.Write(queue_, 1, &matching_alpha);
-
   // Retrieves the kernel from the compiled binary
   try {
     const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -81,7 +77,7 @@ StatusCode Xher<T,U>::DoHer(const Layout layout, const Triangle triangle,
 
     // Sets the kernel arguments
     kernel.SetArgument(0, static_cast<int>(n));
-    kernel.SetArgument(1, alpha_buffer());
+    kernel.SetArgument(1, GetRealArg(matching_alpha));
     kernel.SetArgument(2, x_buffer());
     kernel.SetArgument(3, static_cast<int>(x_offset));
     kernel.SetArgument(4, static_cast<int>(x_inc));
diff --git a/src/routines/level2/xher2.cpp b/src/routines/level2/xher2.cpp
index 50572cea..3d57a9b9 100644
--- a/src/routines/level2/xher2.cpp
+++ b/src/routines/level2/xher2.cpp
@@ -58,10 +58,6 @@ StatusCode Xher2<T>::DoHer2(const Layout layout, const Triangle triangle,
   status = TestVectorY(n, y_buffer, y_offset, y_inc);
   if (ErrorIn(status)) { return status; }
 
-  // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
-  auto alpha_buffer = Buffer<T>(context_, 1);
-  alpha_buffer.Write(queue_, 1, &alpha);
-
   // Retrieves the kernel from the compiled binary
   try {
     const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -69,7 +65,7 @@ StatusCode Xher2<T>::DoHer2(const Layout layout, const Triangle triangle,
 
     // Sets the kernel arguments
     kernel.SetArgument(0, static_cast<int>(n));
-    kernel.SetArgument(1, alpha_buffer());
+    kernel.SetArgument(1, GetRealArg(alpha));
     kernel.SetArgument(2, x_buffer());
     kernel.SetArgument(3, static_cast<int>(x_offset));
     kernel.SetArgument(4, static_cast<int>(x_inc));