Improve cuda%zero()

cuda/psb_c_cuda_vect_mod.F90

@@ -668,7 +668,9 @@ contains
     use psi_serial_mod
     implicit none 
     class(psb_c_vect_cuda), intent(inout) :: x
-
+    ! Since we are overwriting, make sure to do it
+    ! on the GPU side
+    call x%set_dev()
     call x%set_scal(czero)
 
   end subroutine c_cuda_zero

cuda/psb_d_cuda_vect_mod.F90

@@ -668,7 +668,9 @@ contains
     use psi_serial_mod
     implicit none 
     class(psb_d_vect_cuda), intent(inout) :: x
-
+    ! Since we are overwriting, make sure to do it
+    ! on the GPU side
+    call x%set_dev()
     call x%set_scal(dzero)
 
   end subroutine d_cuda_zero

cuda/psb_i_cuda_vect_mod.F90

@@ -650,7 +650,9 @@ contains
     use psi_serial_mod
     implicit none 
     class(psb_i_vect_cuda), intent(inout) :: x
-
+    ! Since we are overwriting, make sure to do it
+    ! on the GPU side
+    call x%set_dev()
     call x%set_scal(izero)
 
   end subroutine i_cuda_zero

cuda/psb_s_cuda_vect_mod.F90

@@ -668,7 +668,9 @@ contains
     use psi_serial_mod
     implicit none 
     class(psb_s_vect_cuda), intent(inout) :: x
-
+    ! Since we are overwriting, make sure to do it
+    ! on the GPU side
+    call x%set_dev()
     call x%set_scal(szero)
 
   end subroutine s_cuda_zero

cuda/psb_z_cuda_vect_mod.F90

@@ -668,7 +668,9 @@ contains
     use psi_serial_mod
     implicit none 
     class(psb_z_vect_cuda), intent(inout) :: x
-
+    ! Since we are overwriting, make sure to do it
+    ! on the GPU side
+    call x%set_dev()
     call x%set_scal(zzero)
 
   end subroutine z_cuda_zero