2006-04-28  Gavin Salam  <salam@lpthe.jussieu.fr>

	* switched sort.f90 to be my own version.
  	* removed unused interpolation routines from interpolation.f90
	
2006-04-07  Gavin Salam  <salam@lpthe.jussieu.fr>

	* Added the example_f90/small_fast_tab.f90 program to examine
	timings and accuracies for a (by default) small tabulation.

	Default parameters give 10^{-4} accuracy at 3 ms/evolution over
	the small region (Rojo's request: 9-e3<x<0.75, Q<50). Timings
	increase to about 8.5ms/evolution for a larger range (1e-6<x,
	Q<14000) and I haven't checked accuracies (NB 6ms with better
	ifort optimization). Note also that eps dependence in integrations
	still needs to be checked.

        Timings below are for small table, with default parameters (see
	svn of this date!), on tycho and gaia (3.4GHz P4).
	
	                          init (s)   evln (ms)     x,Q point (mu s)
	ifort -O                     2.2       3.0               2.8
	ifort -O3 -axP               2.1       2.1               2.6
        lf95 -O                      2.7       2.2               2.2 
        gfortran -O3 -march=i686     5.9     130               180  
	g95 -O3 -march=i686          7.8      21                22            
	
	* increased default interpolation order in the pdf_tabulation to 3
	(2 was definitely too low and 4 might be better). Found that a
	slightly small spacing than used in the f77 routine could be
	beneficial (e.g. 0.07). NB: haven't checked correctness of results
	with -axP on intel (there are reports [banfi] of sporadic
	non-understood problems).
	

2006-04-06  Gavin Salam  <salam@lpthe.jussieu.fr>

	Version 1.0
	
	* imported sources from the PDFevln package, with the addition of
	some f77 hooks and the removal of all the main programs; also a
	directory structure has been introduced so that the library source
	in is src/ while the example and testing programs are in separate
	directories.