Version 1.2 - 08.06.13

• Sparse mode is now fully integrated into ADDA and significantly improved (#98). It is designed for very porous particles.
• Linked to new OpenCL FFT library from AMD (clAmdFft), it is now the default one in OpenCL mode (#144). Speed improvement is a few times depending on the GPU. But there are some limitations (#157).
• New iterative solver BCGS2 - an improved version of BICGSTAB (#24).
• New command line option argument -init_field read <filenameY> [<filenameX>] to initialize an iterative solver by the field read from file (#2).
• Behavior of WKB initial field was slightly improved. Before WKB result was assigned to E_exc, now it is assigned to E (similar to the behavior of -init_field read).
• Subversion revision number was added to the output of -V command line option (#125).
• Timing for generation of incident beam was added.
• Memory usage during prognosis was slightly reduced.
• Finally removed deprecated options -Cpr_mat and -sg_format ddscat.
• Fixed inaccuracy in input/scat_params.dat (r1159).
• A few minor bugs were fixed (#83, #159, #162).
• Several improvements of tests/2exec to accommodate OpenCL and sparse modes, as well as comparison of different FFT implementations (r1192, r1200, r1203).
• Added wiki pages: SparseMode, InstallingclAmdFft, CodeDesign (#75), and CodeStyleGuide (#74).
• Added new folder devtools. It contains a couple of text files, bash scripts for preparing files for release and other tasks.
• Added doxygen config file (doc/doxy/Doxyfile) for automatic generation of extensive docs and GraphViz files (doc/gv/) with schemes describing ADDA design (for svg files in CodeDesign).
• Several improvements to Makefile:
  • Improved the automatic dependency generation - now it is faster and more robust.
  • Full support for compilation in 32-bit mode on 64-bit OS, using EXTRA_FLAGS=-m32, also for misc/ packages.
  • New make targets: cleanfull, cleanruns, cleanrunsseq, cleanrunsmpi, cleanrunsocl. They enable cleaning output from ADDA runs.