- Mar 30, 1998
- Mar 28, 1998
-
-
pw authored
error processing (should have noticed earlier, dang!). This change puts back the missing parts.
-
- Mar 27, 1998
- Mar 26, 1998
-
-
pw authored
-
pw authored
copy-symbol with non-nil optional arg now copies function and value slots along with a copy of property list. gensym now uses current value of *gensym-counter* by default. *gensym-counter* is only incremented if its value is used for the new symbol name.
-
pw authored
with alien package. Will need a re-think.
-
pw authored
as required by ANSI spec.
-
pw authored
Also export BOOLEAN from :common-lisp as required by ANSI spec.
-
pw authored
now signals that condition as required by ANSI spec.
-
pw authored
-
pw authored
-
- Mar 21, 1998
-
-
dtc authored
-
dtc authored
Add long-float support to sub-set-debug-var-slot for the x86 and sparc, and fix some complex-float bugs.
-
dtc authored
-
dtc authored
-
dtc authored
second return value indicating the number of locations associated with the TN.
-
dtc authored
-
dtc authored
-
dtc authored
-
dtc authored
-
dtc authored
print_generation_stats. This is necessary when this function is called directly from the allocation routines on a failure and the FP state is setup for Lisp.
-
- Mar 20, 1998
-
-
pw authored
text-get-last-insertion-position, text-get-baseline and compound-string-baseline now work. Add create-text-field. Note that (some) text-* functions appear to work on text-field widgets, but I need to add the similarly named text-field-* functions before this is fully supported. At least one Motif book suggests text-field is buggy in version 1.1 of Motif and suggests using only text widgets.
-
- Mar 11, 1998
-
-
dtc authored
-
dtc authored
Enhance the make-complex-*-float VOPs to better handle a destination on the stack.
-
dtc authored
single registers as the instruction scheduler in not informed about the usage of the odd FP registers. Enhancements for the complex-float VOPs: 1. Better coding of the creation of complex floats with a destination on the stack. 2. Realpart and imagpart part can now load directly from the stack saving an unnecessary load.
-
dtc authored
1. Better coding of the creation of complex floats with a destination on the stack. 2. Realpart and imagpart part can now load directly from the stack saving an unnecessary load.
-
dtc authored
single registers as the instruction scheduler in not informed about the usage of the odd FP registers. Enhancements for the complex-float VOPs: 1. Better coding of the creation of complex floats with a destination on the stack. 2. Realpart and imagpart part can now load directly from the stack saving an unnecessary load.
-
dtc authored
Make room for the extra FP registers of the sparc V9, and add support for these to many of the instructions.
-
- Mar 10, 1998
-
-
dtc authored
-
dtc authored
move-double-reg which uses a single V9 fmovd instruction rather than two single-reg moves when the backend feature :sparc-v9 is defined. The double register encoding may still need some work.
-
dtc authored
-
dtc authored
sc_DescriptorReg define; the error code had been using a fixed SC offset which becomes invalid when the SCs are modified, for example when adding the complex-float SCs.
-
dtc authored
destination on the stack.
-
dtc authored
byte-order. The raw data vector element swapping is now handled by dump-data-maybe-byte-swapping and these conditionals would break a bootstrap from a big-endian machine - perhaps these were once used to help such a bootstrap.
-
dtc authored
required. The byte swapping of signed-byte 16 vectors will also need some work.
-
dtc authored
elements. The number of elements was being over-calculated, causing garbage to be written beyond the result vector which caused sporadic GC failure when cross compiling.
-
dtc authored
-