appropriate instances of CONS-POINTER to use the new type.
* Replace RTL expression type @ADDRESS->FLOAT with new type
OBJECT->FLOAT.
* Introduce new internal switch USE-PRE/POST-INCREMENT?. Change code
generation of in-line consing to pay attention to this switch.
* Merge common parts of "machine/make" into new file "base/make".
On MIPS:
* Change code sequence that assigns type codes to assume that the type
field has a known value. This eliminates one instruction in every
type-code assignment. It assumes that the data segment bits have a
certain value, but the microcode already does this.
* Cache immediate constants in registers, and remember which registers
contain which constants. (This should be improved by noticing when
arithmetic operations are done on known constants and cacheing the
results.)
* Set USE-PRE/POST-INCREMENT? to false, saving one instruction in
every CONS, and multiple instructions in each call to VECTOR.
#| -*-Scheme-*-
-$Header: /Users/cph/tmp/foo/mit-scheme/mit-scheme/v7/src/compiler/machines/mips/machin.scm,v 1.5 1991/10/25 00:13:12 cph Exp $
-$MC68020-Header: machin.scm,v 4.22 90/05/03 15:17:20 GMT jinx Exp $
+$Header: /Users/cph/tmp/foo/mit-scheme/mit-scheme/v7/src/compiler/machines/mips/machin.scm,v 1.6 1991/10/25 12:24:32 cph Exp $
Copyright (c) 1988-91 Massachusetts Institute of Technology
\f
;;;; Architecture Parameters
-(define use-pre/post-increment? true)
+(define use-pre/post-increment? false)
(define endianness 'LITTLE)
(define-integrable addressing-granularity 8)
(define-integrable scheme-object-width 32)