Now passing alpha/beta to the kernel as arguments as before fp16 support; in case of fp16 arguments are cast on host and in kernel

1 files changed, 4 insertions, 4 deletions

diff --git a/src/kernels/level3/transpose_pad.opencl b/src/kernels/level3/transpose_pad.opencl
index 2de0c7bd..2e20d667 100644
--- a/src/kernels/level3/transpose_pad.opencl
+++ b/src/kernels/level3/transpose_pad.opencl
@@ -31,9 +31,9 @@ __kernel void TransposePadMatrix(const int src_one, const int src_two,
                                  const int dest_one, const int dest_two,
                                  const int dest_ld, const int dest_offset,
                                  __global real* dest,
-                                 const __constant real* restrict arg_alpha,
+                                 const real_arg arg_alpha,
                                  const int do_conjugate) {
-  const real alpha = arg_alpha[0];
+  const real alpha = GetRealArg(arg_alpha);
 
   // Local memory to store a tile of the matrix (for coalescing)
   __local real tile[PADTRA_WPT*PADTRA_TILE][PADTRA_WPT*PADTRA_TILE + PADTRA_PAD];
@@ -95,10 +95,10 @@ __kernel void TransposeMatrix(const int src_one, const int src_two,
                               const int dest_one, const int dest_two,
                               const int dest_ld, const int dest_offset,
                               __global real* dest,
-                              const __constant real* restrict arg_alpha,
+                              const real_arg arg_alpha,
                               const int upper, const int lower,
                               const int diagonal_imag_zero) {
-  const real alpha = arg_alpha[0];
+  const real alpha = GetRealArg(arg_alpha);
 
   // Local memory to store a tile of the matrix (for coalescing)
   __local real tile[PADTRA_WPT*PADTRA_TILE][PADTRA_WPT*PADTRA_TILE + PADTRA_PAD];