Merge pull request #99 from CNugteren/development

Update to version 0.9.0

1 files changed, 6 insertions, 6 deletions

diff --git a/src/kernels/level3/transpose_fast.opencl b/src/kernels/level3/transpose_fast.opencl
index d5c46a30..70156d3a 100644
--- a/src/kernels/level3/transpose_fast.opencl
+++ b/src/kernels/level3/transpose_fast.opencl
@@ -36,12 +36,12 @@ R"(
 
 // Transposes and copies a matrix. Requires both matrices to be of the same dimensions and without
 // offset. A more general version is available in 'padtranspose.opencl'.
-__attribute__((reqd_work_group_size(TRA_DIM, TRA_DIM, 1)))
-__kernel void TransposeMatrixFast(const int ld,
-                                  __global const realT* restrict src,
-                                  __global realT* dest,
-                                  const __constant real* restrict arg_alpha) {
-  const real alpha = arg_alpha[0];
+__kernel __attribute__((reqd_work_group_size(TRA_DIM, TRA_DIM, 1)))
+void TransposeMatrixFast(const int ld,
+                         __global const realT* restrict src,
+                         __global realT* dest,
+                         const real_arg arg_alpha) {
+  const real alpha = GetRealArg(arg_alpha);
 
   // Sets the group identifiers. They might be 'shuffled' around to distribute work in a different
   // way over workgroups, breaking memory-bank dependencies.