diff --git a/.gitignore b/.gitignore
index a1aa80caa4..de6942f861 100644
--- a/.gitignore
+++ b/.gitignore
@@ -22,9 +22,7 @@ build/
 
 bin/*
 FoX
-LIBBEEF
 MBD
-include/fft_defs.h
 include/configure.h
 include/qe_cdefs.h
 install/config.log
@@ -33,6 +31,8 @@ install/configure.msg
 install/make-*
 install/configure-*
 install/uncompress-*
+*/make.depend
+*/*/make.depend
 test-suite/*/test.err.*
 test-suite/*/test.out.*
 test-suite/*/pwscf*
diff --git a/configure b/configure
index 936a957ba2..f3272a776e 100755
--- a/configure
+++ b/configure
@@ -58,8 +58,9 @@ elif [[ (-e make.inc) && (-e Makefile) ]] ; then
 fi
 test -e ./install/make.inc       && rm ./install/make.inc
 
-# compute dependencies (next line prevents bogus errors)
+# compute dependencies (next two lines avoid bogus errors)
 touch ./include/qe_cdefs.h
+touch ./include/configure.h
 ./install/makedeps.sh
 
 # run the autoconf configure with the given conf_flags 
diff --git a/include/defs.h.README b/include/defs.h.README
index cab9479b82..e4540c147b 100644
--- a/include/defs.h.README
+++ b/include/defs.h.README
@@ -27,11 +27,12 @@ You should not need to edit the following files:
          PREPROCESSING OPTIONS USED IN *.h FILES AND IN THE SOURCES
 
 Predefined by specific compilers:
-      _OPENMP            OpenMP parallelization (all compilers)
+*     _OPENMP            OpenMP parallelization (all compilers)
       __GFORTRAN__       GNU gfortran
       __INTEL_COMPILER   Intel compiler
 
 Hardware / Compiler:
+*     __CUDA      Compilation for NVidia GPUs
       __AIX       Ibm rs/6000 machines (obsolescent)
       __PGI       Portland Group compiler workarounds for compiler bugs
        _WIN32     Windows 32-bits
@@ -41,21 +42,21 @@ New algorithms:
       __DOUBLE_BUFFER          Use double buffer algorithm for parallel FFTs
 
 Parallel execution:
-      __MPI       	MPI parallelization
+*     __MPI       	MPI parallelization
       __MPI_MODULE      Compile with "USE mpi" instead of "include 'mpif.h'"
       __MPI_NONBLOCKING Use nonblocking routines (e.g. mpi_ibcast, mpi_igatherv)
 
 Libraries:
-      __FFTW        FFT routines from internal FFTW library (default)
-      __FFTW3       FFT routines from external FFTW v.3 library
-      __DFTI        FFT routines from Intel MKL library
+*     __FFTW        FFT routines from internal FFTW library (default)
+*     __FFTW3       FFT routines from external FFTW v.3 library
+*     __DFTI        FFT routines from Intel MKL library
       __LINUX_ESSL  Use blas/lapack/fft routines from IBM ESSL library
                    (linux version; aix version no longer supported)
-      __SCALAPACK   Use Scalapack routines instead of internal ones
+*     __SCALAPACK   Use Scalapack routines instead of internal ones
                     for parallel subspace diagonalization
       __ELPA_2015   Use ELPA library (compatible with ELPA 2015.x, 2016.05)
       __ELPA_2016   Use ELPA library (compatible with ELPA 2016.11, 2017.x, 2018.05)
-      __ELPA        Use ELPA library (compatible with ELPA 2018.11, 2019.x, 2020.x)
+*     __ELPA        Use ELPA library (compatible with ELPA 2018.11, 2019.x, 2020.x)
       __MASS        Use mathematical routines from IBM MASS library
                     (obsolete?)
       __SX6         Use MathKeisan libraries for NEC SX machines
@@ -66,9 +67,10 @@ Signals:
       __TRAP_SIGUSR1  Enable signal trapping (experimental): code will stop
                       and save data if executable is signaled with USR1
 IO / Reporting:
-      __HDF5           Enable I/O using parallel HDF5 (Fortran 2003
-                       features needed to compile)
+*     __HDF5           Enable I/O using parallel HDF5
       __CLOCK_SECONDS  Translate clock from HH:mm:ss to only seconds
 
-All other preprocessing flags are for debugging purposes and should not be
-used unless you know what you are doing
+The '*' in the first column marks the most important preprocessing flags.
+The other preprocessing flags, and those not included in this list, are
+either less likely to be useful, or are for debugging purposes and be used
+only by knowledgeable people.