Re: cool generators insight

From: David Abrahams (gclbb-jamboost_at_[hidden])
Date: 2003-06-05 10:18:50

David Abrahams <gclbb-jamboost_at_[hidden]> writes:

> and solutions are possible... but they *are* ambiguous, as shown by
> the output of my algorithm, which shows three solutions with cost 9:

I take it back: my algorithm is failing to eliminate some identical
states, and all those solutions are equivalent; let me fix that and it
should run faster.

However, let me also say that the optimal path which produces EST_EXE
from those sources in the fewest transformations looks nothing like
the one you wanted anyway, which is another argument for specifying
intermediate main targets. The optimal path is:

-->cpp------->obj--
/ \
ecpp --< >--
\ / \
-->cpp------->obj-- \
>-->est_exe
cpp------------------->obj-- /
\ /
>--
/
--> cpp---->obj--
/
nm_asm --<
\
--> static_data

which is described by all three results I posted earlier.
-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com