Greetings! Robert Boyer <[EMAIL PROTECTED]> writes:
> I've run the newest 2.7.0 ANSI on ACL2 2.9.3 on "certify books" and got: > > real 61m33.062s > user 58m10.680s > sys 1m26.220s > > Thanks so much. > > With readtable-case and equalp hash tables knocked out so quickly by you, it > almost feels like it might be downhill to pure ANSI from here (for you). The Well, we're now under 5% failures on the tester, and very many of these stem from a single cause. Don't know about downhill, but there is progress for which I am grateful. > biggest quagmires I see ahead are FORMAT (and backquote's finest details). I > wonder if it could be best simply to steal those from Clisp, CMU Lisp, ABCL, > or SBCL if they are done there in Lisp. > Always looking for the right balance here. Fixsing the GCL code has the following benefits 1) At least one developer becomes familiar with it andcan fix it in the future 2) the older stuff usually has all the hard stuff needed for performance already coded, 3) having a diversity of code is beneficial in comparing the lisps for their mutual improvement, 4) don't have to isolate unfamiliar code and parse the hooks surrounding it vs. 1) likely already bug free and widely tested. We've done a mixture thus far, pcl, conditions, destructuring bind, ansi loop are all inhaled, but for smaller components I've tried to salvage the existing code. I'd hate to lose format.c given the obvious amout of work that has gone into it. BTW, the author of femlisp, a finite element package in lisp, is trying to port to GCL, and needs define-symbol-macro. I cannot see how this can be done without going into the guts of eval, which I have managed to avoid thus far. Am I overlooking something hoefully much more simple? Take care, > Bob > > > -- Camm Maguire [EMAIL PROTECTED] ========================================================================== "The earth is but one country, and mankind its citizens." -- Baha'u'llah _______________________________________________ Gcl-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gcl-devel
