sbcl Log


Commit Date  
[a8b9d7] by Douglas Katzman Douglas Katzman

Don't cons in INITIAL-THREAD-FUNCTION-TRAMPOLINE before clearing *gc-inhibit*

Also in HANDLE-THREAD-EXIT don't cons

2014-06-29 03:47:10 Tree
[ad41f3] by Douglas Katzman Douglas Katzman

Remove unnecessary cases from PCL-INSTANCE-P

2014-06-29 03:30:54 Tree
[664fac] by Jan Moringen Jan Moringen

Section on writing tests in HACKING

Based on Stas' feedback

2014-06-28 22:20:41 Tree
[e568bb] (sbcl-1.2.1) by Christophe Rhodes Christophe Rhodes

1.2.1: will be tagged as "sbcl-1.2.1"

2014-06-28 20:46:17 Tree
[94d94d] by Stas Boukarev Stas Boukarev

Fix some immediate encoding issues on ARM.

Function with a lot of arguments easily overflow the immediate limits.

There are still problems with stack load and stores with large offsets.

2014-06-28 16:10:05 Tree
[8bed36] by Douglas Katzman Douglas Katzman

Revert "Derive that (INTERN x "KEYWORD") produces a KEYWORD"

This reverts commit 1d796c22c86e8be5e49857edb3637cc9b70d0d94.

2014-06-25 05:01:37 Tree
[0c54db] by Douglas Katzman Douglas Katzman

Enforce types returned from CLOS stream-read-char (etc) methods.

As per discussion on sbcl-devel, these are necessary as a consequence
of change 7cdaadca removing checks from callers of READ-CHAR et. al.

2014-06-25 04:56:30 Tree
[c06b8c] by Christophe Rhodes Christophe Rhodes

support for DragonFly BSD

x86(-64) only for now, and threads are currently non-functional.
Closes lp#1292845, thanks to Vasily Postnicov.

2014-06-20 10:48:35 Tree
[233cf1] by Douglas Katzman Douglas Katzman

Make -EMPTY-,-UNIVERSAL-,-WILD- types into defglobals

(Renaming with double-earmuffs would be gratuitous noise.)

2014-06-19 13:14:08 Tree
[9bbf8d] by Douglas Katzman Douglas Katzman

Add a remark about meta-compile.

2014-06-19 12:58:31 Tree
[1d796c] by Douglas Katzman Douglas Katzman

Derive that (INTERN x "KEYWORD") produces a KEYWORD

2014-06-19 12:52:41 Tree
[758126] by Douglas Katzman Douglas Katzman

Don't memoize the trivial case of a few 2-arg type functions.

TYPE=, TYPE-{INTERSECTION,UNION}2, and CSUBTYPEP
can test (EQ type1 type2) faster than computing the hash.

2014-06-18 22:28:38 Tree
[957829] by Stas Boukarev Stas Boukarev

Optimize assembly/x86-64/arith.lisp.

Mostly dealing with fixnum type checking.

2014-06-18 20:21:48 Tree
[7cdaad] by Douglas Katzman Douglas Katzman

Add some type-derivers

2014-06-18 16:44:59 Tree
[214b67] by Douglas Katzman Douglas Katzman

Advise compiler that *{HANDLER,RESTART}-CLUSTERS* are always-bound.

2014-06-18 16:44:59 Tree
[efc799] by Stas Boukarev Stas Boukarev

Fix a GC assert error message.

It was using %l for printing a number, change it to %ld.

2014-06-18 16:14:24 Tree
[478ca1] by Douglas Katzman Douglas Katzman

Reconcile error messages between {MAKE,ADJUST}-ARRAY.

Due to ADJUST-ARRAY's guts being extremely confusing, this code tries to
perform more checks up front and re-orders them to be more user-friendly.

There was no obvious reason why one place should say
"~S is not an array with a fill-pointer"
and another
"~S is not of type (SATISFIES ARRAY-HAS-FILL-POINTER-P)".
or
:INITIAL-CONTENTS may not be specified with the :INITIAL-ELEMENT or :DISPLACED-TO option.
vs
Neither :INITIAL-ELEMENT nor :INITIAL-CONTENTS can be specified along with :DISPLACED-TO

So those messages are more uniform now, and as a bonus,
DATA-VECTOR-FROM-INITS takes only 7 args intead of 9.

2014-06-18 04:38:13 Tree
[42a3ce] by Douglas Katzman Douglas Katzman

Improve cache hit rate for some memoized functions.

TYPE= and CSUBTYPEP forced eviction another entry at least
70% to 80% of the time. Enlarging them fixed that.
Also we now choose (arbitrarily) one of two possible lines
for any cache entry, giving another slight performance boost.

The :hash-function specified in DEFUN-CACHED must not
perform masking any more. It's done automatically.

2014-06-18 02:57:36 Tree
[4b7709] by Douglas Katzman Douglas Katzman

Further change to BINDING* to make it more really right.

2014-06-18 00:21:53 Tree
[48793e] by Stas Boukarev Stas Boukarev

Remove TRACE-TABLE-OFFSET slot from CODE.

It stored the byte offset to the location where the code ended and
where the trace table started, but since there's no trace tables, it
was only used for getting the byte size of the code.
The code size is already stored in the CODE-SIZE slot, but in words
and rounded up to two words.
Store verbatim bytes instead and do the conversions when necessary.

2014-06-17 21:41:10 Tree
[d31e78] by Stas Boukarev Stas Boukarev

Remove trace-table.

It hasn't been ever used in SBCL, and it was supposed to be used for
making precise backtraces for the precise GC.
There are probably better ways to achieve precises backtraces.

2014-06-17 21:41:10 Tree
[114719] by Douglas Katzman Douglas Katzman

Really make BINDING* accept declarations

2014-06-17 16:06:41 Tree
[4c0c12] by Douglas Katzman Douglas Katzman

Make BINDING* accept declarations, kinda.

It's not perfect, but this works now:
(BINDING* (((Q R) (FLOOR X Y))
(S (+ R 3)))
(DECLARE (TYPE MUMBLE S)) ...)

2014-06-16 22:50:20 Tree
[588eb5] by Paul Khuong Paul Khuong

Fix a caching bug in life.lisp

We'd sometimes cache too aggressively and mistakenly conclude that a
TN had no global conflict in an IR2 block. This would then lead to
having multiple conflict structs for the same block/TN pair, and,
finally, AVERs in graph colouring regalloc.

Reported by Douglas Katzman. Fixes lp#1327008.

2014-06-15 21:26:10 Tree
[398361] by Stas Boukarev Stas Boukarev

Restore the docstring of sb-bsd-sockets:get-protocol-by-name.

2014-06-14 17:39:41 Tree
Older >