- Feb 13, 1992
-
-
wlott authored
-
- Dec 22, 1991
-
-
ram authored
cost is missing.
-
- Nov 11, 1991
-
-
ram authored
-
- Nov 08, 1991
-
-
ram authored
by move VOPs. If the other operand has a known representation, then use the appropriate cost. This is particularly useful for moves into wired passing locations, etc.
-
- Jun 19, 1991
-
-
ram authored
coercion would be inserted after the ALLOCATE-FRAME VOP, which caused register save computation to get confused.
-
- May 03, 1991
- Apr 24, 1991
- Feb 20, 1991
-
-
ram authored
-
- Feb 11, 1991
-
-
ram authored
-
- Feb 04, 1991
-
-
ram authored
-
- Nov 13, 1990
-
-
wlott authored
from the correct backend for runtime.
-
- Nov 11, 1990
-
-
ram authored
done by the primitive type of the operands. Also, added stuff to delete moves to :NORMAL TNs that have no reads. Changed the existing special-casing of unread results to only punt on :NORMAL TNs. Also, fixed the move/coerce generation stuff to just leave alone the MOVEs that are o.k., instead of deleting the old MOVE and inserting a new one.
-
- Nov 03, 1990
-
-
wlott authored
information.
-
- Oct 24, 1990
-
-
ram authored
an expensive coercion (like number consing.)
-
- Aug 16, 1990
-
-
ram authored
-
- Jul 23, 1990
-
-
ram authored
-
- Jun 19, 1990
-
-
wlott authored
-
- Jun 15, 1990
-
-
wlott authored
where the NFP is allocated to be a call to make-number-stack-pointer-tn instead of a make-representation-tn.
-
- Jun 06, 1990
-
-
ram authored
disallowed by the primitive type.
-
- May 30, 1990
-
-
cvs2git authored
-
- May 15, 1990
-
-
wlott authored
-
- May 14, 1990
-
-
ram authored
-
- May 11, 1990
-
-
wlott authored
the NFP tn.
-
- Apr 30, 1990
-
-
ram authored
-
- Apr 27, 1990
-
-
wlott authored
being used.
-
- Apr 22, 1990
-
-
wlott authored
of just spinning forever marking the same tn as being on the number stack.
-
- Apr 17, 1990
-
-
ram authored
-
- Apr 13, 1990
-
-
ram authored
the value in the VOP-INFO to see if there is an inconsistency. Added CHECK-MOVE-FUNCTION-CONSISTENCY. When computing costs, ignore references due to the MOVE VOP. Changed stuff over to the new scheme of having the NFP be per-component rather than per-function. Added new argument moving stuff.
-
- Mar 27, 1990
-
-
ram authored
-
- Mar 08, 1990
-
-
ram authored
-