> > On Fri, Mar 9, 2018 at 8:38 PM, Waldek Hebisch <hebi...@math.uni.wroc.pl> > wrote: > >> > >> Waldek, your changes to fix this problem (SVN r2373 or git 6404dd01) > >> has made SBCL-1.4.5 to use 'make-array' instead of > >> 'sbcl_make_sized_vector', is that your intention? > > > > Yes. We use 'sbcl_make_sized_vector' because it was much faster. > > But with new sbcl the claim is that 'make-array' is much > > faster than it was. > > > > For small vector, 'sbcl_make_sized_vector' is faster than 'make-array' > in sbcl 1.4.5, it's 20% for vector with 10 elements.
We can tolerate 20%. > It is a bug to have "`(ceiling (* ,n ,n-bits) sb-vm:n-word-bits)" > in 'sbcl_make_sized_vector', we should fix it to > "`(ceiling (* ,n ,nb) sb-vm:n-word-bits)". > > Also I think it is better to use ALLOCATE-VECTOR-WITH-WIDETAG > than ALLOCATE-VECTOR. We try to do the same thing 'make-array' would do, but skipping time-consuming irrelevant stuff. 'sbcl' used to to do 'ALLOCATE-VECTOR' so we do so. -- Waldek Hebisch -- You received this message because you are subscribed to the Google Groups "FriCAS - computer algebra system" group. To unsubscribe from this group and stop receiving emails from it, send an email to fricas-devel+unsubscr...@googlegroups.com. To post to this group, send email to fricas-devel@googlegroups.com. Visit this group at https://groups.google.com/group/fricas-devel. For more options, visit https://groups.google.com/d/optout.