- Aug 05, 1992
-
-
ram authored
LOOP is supposed to have setting semantics.
-
- Jun 05, 1992
-
-
ram authored
is specified (need to get the right type of zero.)
-
- Apr 08, 1992
-
-
wlott authored
try to run when the 1 is real. Otherwise it does fun things like call minusp on #C(0 1).
-
- Feb 03, 1992
-
-
ram authored
rational-specific. Include CONSTANT-ARGUMENT in various arg type restrictions so that we don't get silly efficiency notes. Changed multiply recoding to left-associate the sum so that we are less likely to run out of non-descriptor registers.
-
- Feb 02, 1992
-
-
ram authored
intermediate of (- 0 x), since this is not a correct transformation. (- 0 0.0) is 0.0, not -0.0. Fixed the (- 0 x) transform to be restricted to rational args.
-
ram authored
variable substitution can't swap it back again. Fixed multiple recoding to include many TRULY-THE's in the expansion so that the resulting shift-and-add code would actually open coded. Added comprehensive handling of arithmetic and logical identities when an arg is -1, 0 or +1. Fixed the (- 0 x) ==> (%negate x) transform to be inhibited when the 0 is more contagious than X, or is -0.0.
-
- Nov 14, 1991
-
-
wlott authored
-
- Nov 12, 1991
-
-
ram authored
-
- Nov 09, 1991
-
-
wlott authored
-
- Oct 03, 1991
-
-
ram authored
-
- Sep 03, 1991
-
-
ram authored
-
- Feb 20, 1991
-
-
ram authored
-
- Jan 30, 1991
-
-
ram authored
-
- Jan 22, 1991
-
-
ram authored
unnecessary shifting.
-
- Jan 12, 1991
-
-
ram authored
-
- Dec 12, 1990
-
-
ram authored
explicit supplied-p so that anyone who does (truncate x nil) gets what they deserve.
-
- Dec 04, 1990
-
-
wlott authored
'(unsigned-byte 0). Use '(integer 0 0) instead.
-
- Dec 01, 1990
-
-
wlott authored
more bits that the shorter argument, even if it was negative. So I totally rewrote the derive-type optimizers for LOGAND, LOGIOR, and LOGXOR in case there were other problems.
-
- Nov 10, 1990
-
-
wlott authored
-
- Oct 10, 1990
-
-
ram authored
change so that we don't "optimize" (< *foo* *foo*) to NIL.
-
- Oct 06, 1990
-
-
ram authored
-
- Oct 05, 1990
-
-
ram authored
any necessary efficency note
-
- Oct 03, 1990
-
-
ram authored
in the efficency note.
-
- Sep 27, 1990
-
-
ram authored
an arg is a FIXNUM, since there is a good implementation in that case.
-
- Sep 25, 1990
-
-
wlott authored
in the final expression.
-
- Sep 24, 1990
-
-
wlott authored
-
- Aug 24, 1990
-
-
wlott authored
-
- Jun 01, 1990
- May 30, 1990
-
-
cvs2git authored
-
- May 23, 1990
-
-
ram authored
types do intersect. The old code was correct, but missed some opportunities.
-
- May 09, 1990
-
-
ram authored
-
- Mar 28, 1990
-
-
ram authored
-
- Mar 19, 1990
-
-
wlott authored
-
wlott authored
-
wlott authored
small, (e.g. (integer 0 15)) it will shift the range of the shiftee to obtain the domain. Changed the %ldb transform to map to (logand (ash num (- pos)) (mask size)) if the result is going to be a fixnum. Added transforms for floor, mod, truncate, and rem that check to see if the argument is a constant power of 2, and if so, change the divide into a shift.
-
- Mar 10, 1990
-
-
ram authored
INTEGER (REAL) arguments.
-
- Mar 05, 1990
-
-
cvs2git authored
-
- Feb 08, 1990
-
-
ram authored
call the appropriate subprimitive only if the number is a COMPLEX/RATIO. This simplifies the primitives and allows the type check to be squeezed out in cases where we know the type at compile time.
-
- Feb 06, 1990
-
-
wlott authored
-