Micro-optimizations: * MAKE-FIXNUM on PPC now takes one instruction rather than two; * signed-num INTEGER-LENGTH VOP on PPC does not require a temporary register; * Added unsigned-num INTEGER-LENGTH VOP on PPC; * Musings on processor-specific micro-optimizations added to OPTIMIZATIONS.
is no :key or :test-not argument and :test is one of the standardly acceptable arguments for make-hash-table. The previous algorithm ran in O(n^2) time; this algorithm runs in O(n) time, n the length of the list.
…n the situations in which you've unbroken it. ANSI test DECODE-UNIVERSAL-TIME.5 now passes.
…presentable range in DECODE- and ENCODE-UNIVERSAL-TIME (reported by Paul Dietz and his magical ANSI test suite)
…massive speed benefit. Over 3 times speedup on looped read-line.
should be the last sizable slice of whitespace canonicalization (but it yet remains at least to consider *.sh/*.S/*.foobar, to make canonicalness check run as part of tests/, and to remove annoying "echo" statements) (Anyone who ends up here with "cvs annotate" probably wants to look at the "tabby" tagged version.)
types that normalize to very large MEMBER types on characters.
added scripts to impose whitespace canonicalization added scripts to check that sources haven't drifted away from canonical whitespacing (as when someone doesn't use emacs' INDENT-TABS-MODE=NIL...) (Note that the scripts include an annoying level of "echo"-ish commands which should probably be removed once things seem to be working OK for everyone.) (Something about huge touch-every-file "cvs commit" of canonicalized-whitespace versions of all sources is clashing with my flaky internet connection, so there may be some further delay in committing the result of running the scripts.)
(This is a trivial placeholder which is just preparation for stripping tabs, and otherwise canonicalizing whitespace, in most of the sources.) Later today I plan to change the whitespace style in most of the system's source files. Thus, this version will be either the last or very very close to the last version where "cvs annotate" conveniently and cleanly runs back to the dawn of time. Thus, after this commit I will run "cvs tag tabby".
* fixed 32bitness in some places * untangled waitqueues, mutexes and spinlocks * updated the user manual * removed target-unithread.lisp and target-multithread.lisp and put everything into target-thread.lisp mainly for OOAO the docstrings, but also to provide as much as possible of the thread interface in unithread builds
Fix for memory fault error in foreign.test.sh ... we need to be able to have two different linkage-table entries for "address of data" and "instructions to jump to address" (DATAP=NIL/T) ... use a cons as the hash key rather than just the name ... (this area is not my speciality, so this fix may be suboptimal. Adjust as necessary)
...thanks to Thiemo Seufer.