- Feb 21, 1990
-
-
ram authored
-
ram authored
the structure before initializing the slots.
-
ram authored
these are always standard for now, they are now always required.
-
wlott authored
-
ram authored
SAVE-LISP init function.
-
ram authored
must be at the end of worldload.
-
ram authored
DEBUG-FUNCTION-START-PC until we actually compute this correctly.
-
ram authored
-
ch authored
-
- Feb 20, 1990
-
-
ch authored
-
wlott authored
-
wlott authored
Fixed reference to value-cell-type to be value-cell-header-type.
-
wlott authored
-
wlott authored
-
wlott authored
Added automatic generation of slot-ref and slot-set for :rest t slots.
-
wlott authored
Added closure-function-header and value-cell-header.
-
wlott authored
compile-time.
-
wlott authored
-
wlott authored
Changed call to load-foreign to emit load-foreign instruction. (oops)
-
wlott authored
Added header word stuff.
-
wlott authored
-
ram authored
initially making it.)
-
ram authored
-
ram authored
-
ram authored
compiled-debug-blocks. Changed compiled-location to compiled-debug-variable. Added DEBUG-FUNCTION-TLF-NUMBER and DEBUG-FUNCTION-START-PC slots. Allow the DEBUG-FUNCTION-ARGUMENTS to be any kind of vector so that it can be an i-vector in fixed-arg lambdas with no deleted arguments.
-
ram authored
localify so that GGS doesn't get confused by objects being trashed with GC forwards. In TRANSPORT-FUNCTION, we must be prepared to find closure-entry objects (and treat them the same as normal entries.) Also, we must check if the entry for a closure is pure before transporting the entry. In various places, I had to flush uses of INFO, since this accesses global variables (*info-environment*).
-
ram authored
wrote COERCE-TO-SMALLEST-ELTYPE and used it in various places to dump i-vectors when possible.
-
ram authored
vector in the VOP-SAVE-SET, rather than doing the old info-arg hack.
-
ram authored
VOP-GENERATOR-NUMBER.
-
ram authored
DO-LIVE-TNS macros.
-
ram authored
order, and starts with the bind block. This required disabling the code that attempts to drop through on tail calls. When I redo call to get tail calls right, this should start working again, at least in the simple cases we really care about.
-
ram authored
-
- Feb 19, 1990
-
-
ram authored
compiler.
-
ram authored
compiler. Installing an interpreted definition was breaking the compiler/eval, presumably because some calls aren't inline expanded yet.
-
ram authored
-
ram authored
IR1-TOP-LEVEL. Changed these interfaces to take TLF numbers instead of source paths, since that is what they really operate on.
-
ram authored
longer does this. Also, changed the PATH arg to IR1-TOP-LEVEL to be a TLF number (i.e. 0 instead of '(0)).
-
ram authored
FIND-SOURCE-PATHS. Now we call FIND-SOURCE-PATHS before doing our recursive PROCESS-FORM walk, and we just pass around the TLF number, instead of constructing a source path as we go. This allowed PROCESS-FORM to be changed to expand macros.
-
ram authored
instead of always defining a transform for string/=*.
-
ram authored
that it gets the correct cleanup computed by NODE-ENDS-BLOCK.
-