-------- Original Message -------- Subject: Re: Bug affecting mpq_class::get_num() and mpq_class::get_den()? Date: Thu, 13 Jun 2002 11:29:25 +0200 From: gerardo.ballabio@unimib.it (Ballabio Gerardo - Dip. di Scienza dei Materiali) To: Kevin Ryde <user42@zip.com.au> CC: Roberto Bagnara <bagnara@cs.unipr.it>, Torbjorn Granlund <tege@swox.se> References: <87k7p9bdqi.fsf@zip.com.au> <20020610122024.C4790@nettunio.mater.unimib.it> <87adq1cpi8.fsf@zip.com.au> Hi, I've done the "symlinking" and had mpq_class::get_num/get_den return true mpz_class references (see attached files) -- but don't like it very much. The problem is, I had to change the implementation of mpz_class and mpq_class: originally they just contained an mpz_t and an mpq_t, respectively. Now mpz_class contains an mpz_t AND an mpz_ptr: the pointer "normally" points to the mpz_t, but can also be "redirected" to another mpz_t, in particular, to the numerator or denominator of an mpq_t. mpq_class contains an mpq_t AND two mpz_classes: this amounts to an mpq_t, two pointers, and two mpz_t. That is, sizeof(mpq_class) is more than twice that of mpq_t; furthermore, the two mpz_t are always unused. I'd call such an implementation quite bloated. Another possibility is to throw away the mpz_t, keeping just the pointer, and to allocate it dynamically at the time of construction (deallocation would be done by the destructor). This is certainly better on technical grounds, but yields some (small) overhead. However this overhead occurs just when constructing and destroying, which should always happen outside of time-critical portions of code. Of course there is a third choice, that is, leave everything as is. My "vote" goes to the second (dynamical allocation), but I would like to hear your advice. Gerardo -- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it