diff --git a/CPV/src/fromscra.f90 b/CPV/src/fromscra.f90
index 01619fcab..683d352d0 100644
--- a/CPV/src/fromscra.f90
+++ b/CPV/src/fromscra.f90
@@ -144,6 +144,7 @@ SUBROUTINE from_scratch( )
     !
     if( iverbosity > 1 ) CALL dotcsc( eigr, cm_bgrp, ngw, nbsp )
     !
+    !
     ! ... initialize bands
     !
     CALL occn_info( f )
diff --git a/CPV/src/nl_base.f90 b/CPV/src/nl_base.f90
index d75639153..fb2fbf952 100644
--- a/CPV/src/nl_base.f90
+++ b/CPV/src/nl_base.f90
@@ -1238,22 +1238,3 @@ subroutine nlfq_bgrp_x( c_bgrp, eigr, bec_bgrp, becdr_bgrp, fion )
   !
   return
 end subroutine nlfq_bgrp_x
-
-! In principle this can go away .......
-SUBROUTINE MYDGEMM( TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC )
-#if defined(__CUDA)
-    use cudafor
-    use cublas
-#endif
-    CHARACTER*1, INTENT(IN) ::        TRANSA, TRANSB
-    INTEGER, INTENT(IN) ::            M, N, K, LDA, LDB, LDC
-    DOUBLE PRECISION, INTENT(IN) ::   ALPHA, BETA
-    DOUBLE PRECISION  :: A( LDA, * ), B( LDB, * ), C( LDC, * ) 
-#if defined(__CUDA)
-    attributes(device) :: A, B, C
-    CALL cublasdgemm(TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC)
-#else
-    CALL dgemm(TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC)
-#endif
-
-END SUBROUTINE MYDGEMM
diff --git a/CPV/src/ortho.f90 b/CPV/src/ortho.f90
index b41afae07..44695d146 100644
--- a/CPV/src/ortho.f90
+++ b/CPV/src/ortho.f90
@@ -632,18 +632,3 @@ END MODULE local_ortho_memory
 100   FORMAT(3X,'diff = ',D18.10,' iter = ', I5 )
       !
    END SUBROUTINE ortho_x
-
-
-
-
-SUBROUTINE qe_sync()
-#if defined(__CUDA)
-      USE cudafor
-#endif
-   INTEGER :: info
-#if defined (__CUDA)
-   info = cudaDeviceSynchronize()
-   IF( info /= 0 ) CALL errore('qe_sync',' error ',ABS(info))
-#endif
-   RETURN
-END SUBROUTINE
diff --git a/CPV/src/ortho_base.f90 b/CPV/src/ortho_base.f90
index 53db0b561..1194fe399 100644
--- a/CPV/src/ortho_base.f90
+++ b/CPV/src/ortho_base.f90
@@ -1312,22 +1312,3 @@ CONTAINS
    END SUBROUTINE bec_bgrp2ortho
 
 END MODULE orthogonalize_base
-
-
-! In principle this can go away .......
-SUBROUTINE MYDGER  ( M, N, ALPHA, X, INCX, Y, INCY, A, LDA )
-#if defined(__CUDA)
-    use cudafor
-    use cublas
-#endif
-!     .. Scalar Arguments ..
-    DOUBLE PRECISION ::  ALPHA
-    INTEGER          ::   INCX, INCY, LDA, M, N
-!     .. Array Arguments ..
-    DOUBLE PRECISION :: A( LDA, * ), X( * ), Y( * )
-#if defined(__CUDA)
-    attributes(device) :: A, X, Y
-#endif
-    CALL DGER  ( M, N, ALPHA, X, INCX, Y, INCY, A, LDA )
-
-END SUBROUTINE MYDGER
diff --git a/UtilXlib/Makefile b/UtilXlib/Makefile
index 7ca3159eb..9cc3c37ec 100644
--- a/UtilXlib/Makefile
+++ b/UtilXlib/Makefile
@@ -6,6 +6,7 @@ include ../make.inc
 MODFLAGS=$(MOD_FLAG).
 
 UTIL = clocks_handler.o \
+device_helper.o \
 device_util.o \
 divide.o \
 data_buffer.o \
diff --git a/UtilXlib/device_helper.f90 b/UtilXlib/device_helper.f90
new file mode 100644
index 000000000..4dab88356
--- /dev/null
+++ b/UtilXlib/device_helper.f90
@@ -0,0 +1,67 @@
+!
+! Copyright (C) 2003-2020 Quantum ESPRESSO group
+! This file is distributed under the terms of the
+! GNU General Public License. See the file `License'
+! in the root directory of the present distribution,
+! or http://www.gnu.org/copyleft/gpl.txt .
+!
+! This file initiated by Carlo Cavazzoni 2020
+!
+! Purpose: collect miscellaneus subroutines to help dealing with 
+!          accelerator devices
+
+
+!=----------------------------------------------------------------------------=!
+
+SUBROUTINE qe_device_sync()
+#if defined(__CUDA)
+      USE cudafor
+#endif
+   INTEGER :: info
+#if defined (__CUDA)
+   info = cudaDeviceSynchronize()
+   IF( info /= 0 ) CALL errore('qe_sync',' error ',ABS(info))
+#endif
+   RETURN
+END SUBROUTINE
+
+!=----------------------------------------------------------------------------=!
+
+! In principle this can go away .......
+SUBROUTINE MYDGER  ( M, N, ALPHA, X, INCX, Y, INCY, A, LDA )
+#if defined(__CUDA)
+    use cudafor
+    use cublas
+#endif
+!     .. Scalar Arguments ..
+    DOUBLE PRECISION ::  ALPHA
+    INTEGER          ::   INCX, INCY, LDA, M, N
+!     .. Array Arguments ..
+    DOUBLE PRECISION :: A( LDA, * ), X( * ), Y( * )
+#if defined(__CUDA)
+    attributes(device) :: A, X, Y
+#endif
+    CALL DGER  ( M, N, ALPHA, X, INCX, Y, INCY, A, LDA )
+
+END SUBROUTINE MYDGER
+
+!=----------------------------------------------------------------------------=!
+
+! In principle this can go away .......
+SUBROUTINE MYDGEMM( TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC)
+#if defined(__CUDA)
+    use cudafor
+    use cublas
+#endif
+    CHARACTER*1, INTENT(IN) ::        TRANSA, TRANSB
+    INTEGER, INTENT(IN) ::            M, N, K, LDA, LDB, LDC
+    DOUBLE PRECISION, INTENT(IN) ::   ALPHA, BETA
+    DOUBLE PRECISION  :: A( LDA, * ), B( LDB, * ), C( LDC, * )
+#if defined(__CUDA)
+    attributes(device) :: A, B, C
+    CALL cublasdgemm(TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC)
+#else
+    CALL dgemm(TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC)
+#endif
+
+END SUBROUTINE MYDGEMM