update README

README.md

@@ -1,4 +1,55 @@
 # parallelproj
 code for parallel TOF and NONTOF projections
 
-## foo bar
+## dependencies
+- cmake>=3.9 (3.9 version needed to detect CUDA correctly)
+- recent c compiler
+- CUDA (tested with 10.1.105)
+
+for the examples:
+- python (tested with 3.7.6)
+- numpy  (tested with 1.18.1)
+
+## build the project
+```
+cd my_project_dir
+mkdir build
+cmake ..
+make 
+make install
+```
+
+The CMakeLists.txt is configured to search for CUDA.
+If CUDA is not present, compilation of the CUDA lib is skipped.
+
+After a successfull build and install, the compiled libs
+should appear in the ./lib directory.
+
+If doxygen is present, the documentation is rendered in 
+html and latex in the ./doc directory.
+
+## run examples
+All python examples should have the "-h" option to show the meaning of the command line options.
+
+OpenMP CPU examples:
+```
+cd my_project_dir
+cd examples
+
+python openmp_nontof_sino_scaling.py --nv 1
+python openmp_tof_sino_scaling.py --nv 1 --nrep 1
+
+python openmp_nontof_lm_scaling.py --ne 1e6
+python openmp_tof_lm_scaling.py --ne 1e6
+```
+CUDA examples:
+```
+cd my_project_dir
+cd examples
+
+python cuda_nontof_sino_scaling.py --nv 1
+python cuda_tof_sino_scaling.py --nv 1 --nrep 1
+
+python cuda_nontof_lm_scaling.py --ne 1e6
+python cuda_tof_lm_scaling.py --ne 1e6
+```