Depend on parallel, doc improvements

By depending on parallel instead of importing it, functions to set up
and stop a cluster are always available when mkin is loaded.

The use of multicore processing in mmkin on Windows is now documented in
the help file, which brings mkin closer to a version 1.0 #9.

1 files changed, 11 insertions, 0 deletions

diff --git a/man/mmkin.Rd b/man/mmkin.Rd
index 80360cd7..9b836242 100644
--- a/man/mmkin.Rd
+++ b/man/mmkin.Rd
@@ -76,6 +76,17 @@ plot_sep(fits.0[[1, 1]])
 # Plotting with mmkin (single brackets, extracting an mmkin object) does not
 # allow to plot the observed variables separately
 plot(fits.0[1, 1])
+
+# On Windows, we can use multiple cores by making a cluster using the parallel
+# package, which gets loaded with mkin, and passing it to mmkin, e.g.
+cl <- makePSOCKcluster(12)
+f <- mmkin(c("SFO", "FOMC", "DFOP"),
+  list(A = FOCUS_2006_A, B = FOCUS_2006_B, C = FOCUS_2006_C, D = FOCUS_2006_D),
+  cluster = cl, quiet = TRUE)
+print(f)
+# We get false convergence for the FOMC fit to FOCUS_2006_A because this
+# dataset is really SFO, and the FOMC fit is overparameterised
+stopCluster(cl)
 }
 
 }