x1. Improve the seeding for routines that use the LCG seed generator.
It can only generate 130,000 different initial states. We should
change it to provide 2^31 different initial states (this will also
prevent the high bits being zero). DONE - we now use a 32-bit
generator.

x2. Make things work on 64 bit platforms. Sometimes we just need to
use & 0xffffffff on 32 bit quantities that are returned to the user
where we rely on 32 bit arithmetic, sometimes it is more complicated
(e.g. the Tausworthe generator relies on 32 bit arithmetic at all
stages). DONE, needs to be tested on an actual 64 bit machine though
DONE (I tested it on the Alpha (ancho).

x3. Add reimplementations of the Numerical Recipes generators, since
lots of people use those. DONE, for ran0,1,2,3

x4. Add functions for getting an integer in the range [0,N] or [1,N],
like rolling dice. Be aware that division might not make a smooth
distribution, and a small modulus might be correlated for some generators.
DONE, for [0,N-1]

x5. provide a RAND_SCALE precomputed scale factor for getting [0,1] In
numerical recipes they divide by (MAX+1) which we can't handle in our
current scheme. They also replace 1 by 1-eps where eps=1.2e-7. Yuk.
DONE

6. run the DIEHARD tests on all the generators, especially the ones we
are listing as "Simulation Quality" -- some of those are a bit old and
might fail one or two diehard tests.

7. Add drand48,   DONE
       NAG,       missing, gave up! 
       CRAY,      DONE 
       CDC 48-bit missing 

x8. Get the macros from the faster MT19937 generator and use them. We
need to make MT be the fastest of the simulation quality generators if
it is the default. DONE. It didn't improve the speed on other
platforms, so I just used the tricks which also worked on the pentium
(e.g. changing mag[x&1] to x&1 ? mag[1] : mag[0])

x9. Send to M.Makumoto for review. DONE, no reply as of 4 Nov 1998 though.

x10. Add random/srandom DONE.

11. Check out the bug fix to mrand48 that was made in glibc2, pr757

12. Check out  the following paper,

On the anomaly of ran1() in Monte Carlo pricing of financial
derivatives; Akira Tajima , Syoiti Ninomiya , and Shu Tezuka ; Winter
simulation , 1996, Page 360, from ACM

13. The following papers have been published, I think we refer to them
(or could do),

    Pierre L'Ecuyer, "Tables of Linear Congruential Generators of different
    size and good lattice structure", Mathematics of Computation, Vol 68,
    No 225, Jan 1999, p249-260

    Pierre L'Ecuyer, "Tables of Maximally equidistributed combined LSFR
    generators", ibid, p261-270

14. Look at this paper: I. Vattulainen, "Framework for testing random numbers
    in parallel calculations", Phys. Rev. E, 59, 6, June 1999, p7200

15. Check that RANLXS will work on machines with non-standard with of
float/dbl (original has checks for DBL_MANT_DIG ..)

16. Rewrite test.c to use a loop over a list of all the generators.
Too much cut and paste! DONE