Nikodemus Siivola [Tue, 6 Nov 2007 15:01:06 +0000 (15:01 +0000)]
1.0.11.14: thread-safe FORM-NUMBER-TRANSLATIONS
* Instead of keeping global tables, allocate new ones for each
call. Since common case seems to be a "decently small" form,
use a list instead of hash-table.
...but we could really do with a better internal set representation
for arbitrary objects -- both here and in MEMBER type machinery.
Nikodemus Siivola [Tue, 6 Nov 2007 14:25:37 +0000 (14:25 +0000)]
1.0.11.13: dead code deletion
* *IR1-LAMBDA-DEBUG-FUN* and *IR1-BLOCK-DEBUG-BLOCK* have been
unused for a looong time...
Nikodemus Siivola [Tue, 6 Nov 2007 14:21:50 +0000 (14:21 +0000)]
1.0.11.12: MAPHASH and WITH-HASH-TABLE documentation
* paraphrase the hash-table traversal/side-effect rule
Nikodemus Siivola [Tue, 6 Nov 2007 14:17:40 +0000 (14:17 +0000)]
1.0.11.11: expand zero-object WITH-PINNED-OBJECTS to a PROGN
* x86 and x86-64 only.
Nikodemus Siivola [Tue, 6 Nov 2007 13:43:25 +0000 (13:43 +0000)]
1.0.11.10: remove useless WITHOUT-GCING
* The surrounding WITH-RECURSIVE-SYSTEM-SPINLOCK
has :WITHOUT-GCING T.
Nikodemus Siivola [Tue, 6 Nov 2007 13:33:13 +0000 (13:33 +0000)]
1.0.11.9: whitespace
* Fix indentation of target-hash-table.lisp.
Rudi Schlatte [Mon, 5 Nov 2007 04:30:37 +0000 (04:30 +0000)]
1.0.11.8: also fix install-info on Debian
Also reported by Eugene Ossintsev
Rudi Schlatte [Mon, 5 Nov 2007 04:14:41 +0000 (04:14 +0000)]
1.0.11.7: Don't try to register info-* files using install-info
Reported by Eugene Ossintsev
NIIMI Satoshi [Sun, 4 Nov 2007 15:54:19 +0000 (15:54 +0000)]
1.0.11.6: canonicalize whitespaces only on updated files
Cyrus Harmon [Sat, 3 Nov 2007 08:54:31 +0000 (08:54 +0000)]
1.0.11.5: update tests for newer macos version
* use MAC_OS_X_VERSION_10_5 instead of __DARWIN_UNIX03
Rudi Schlatte [Sat, 3 Nov 2007 08:05:03 +0000 (08:05 +0000)]
1.0.11.4: Unset GREP_OPTIONS when filtering nm output in linux-nm
* Nuke it since there is no grep option (that I could see) to
reverse -i or ignore the GREP_OPTIONS variable
* Patch by Daniel Lowe
William Harold Newman [Fri, 26 Oct 2007 12:48:14 +0000 (12:48 +0000)]
1.0.11.3: belatedly recognizing STREAM-ERROR and PARSE-ERROR printing is OK
* (In my 1.0.9.60 patch, I fixed a READER-ERROR printing bug and
wrote that parallel PARSE-ERROR and STREAM-ERROR printing bugs
remained and that I would fix them. Now finally returning to
fix them, I find that they were fixed already, the tests work
with with no additional fixes needed.)
* so I just un-#+NILed-out the tests
* also made a few trivial rearrangements in
src/code/condition.lisp, moving things around without changing
meaning
Rudi Schlatte [Fri, 26 Oct 2007 06:23:07 +0000 (06:23 +0000)]
1.0.11.2: defer package creation of defpackage
Patch by Zach Beane to sbcl-devel 2007-10-09
Rudi Schlatte [Fri, 26 Oct 2007 04:33:34 +0000 (04:33 +0000)]
1.0.11.1: Handle set-but-empty environment variables
... SBCL_HOME= sbcl crashed on startup
... Patch by Michael Weber
William Harold Newman [Thu, 25 Oct 2007 21:28:03 +0000 (21:28 +0000)]
1.0.11: release, will be tagged as sbcl_1_0_11
Rudi Schlatte [Mon, 22 Oct 2007 04:16:12 +0000 (04:16 +0000)]
1.0.10.55: Remove non-portable arguments to find, xargs
NIIMI Satoshi [Sun, 21 Oct 2007 23:25:35 +0000 (23:25 +0000)]
1.0.10.54: Fix build with old gcc
Surround code fragment with braces for variable declarations.
Cyrus Harmon [Sun, 21 Oct 2007 23:19:12 +0000 (23:19 +0000)]
1.0.10.53: fix undefined-alien errors on darwin
* the new mach exception handling stuff was checking to see if there
were two faults on the page by checking the region's protection
status and bailing out assuming that there were two faults on the
same page. The problem was that it was following this code path for
undefined-alien errors (and presumably control-stack-exhausted)
thereby just resignalling an EXC_BAD_ACCESS, ad infinitum. Moved
the checks for the page address being undefined_alien_address, or
the control stack guard pages up before checking the page
protection.
Rudi Schlatte [Sun, 21 Oct 2007 05:15:04 +0000 (05:15 +0000)]
1.0.10.52: fix clean.sh for whitespace-containing source directories
NIIMI Satoshi [Sat, 20 Oct 2007 06:27:20 +0000 (06:27 +0000)]
1.0.10.51: New function: THREAD-YIELD
An interface to sched_yield(). And use it in busy loops in some tests.
Nikodemus Siivola [Fri, 19 Oct 2007 13:57:11 +0000 (13:57 +0000)]
1.0.10.50: short-circuit LVAR type derivation when type becomes T
* Even if there are uses we have not considered yet,
%LVAR-DERIVED-TYPE can stop if the type turns into the universal
type.
Nikodemus Siivola [Fri, 19 Oct 2007 13:25:03 +0000 (13:25 +0000)]
1.0.10.49: deadline refinements
* Deadlines are per-thread. (Children do no inherit their parents
deadlines.)
* SIGNAL-DEADLINE estabilishes a DEFER-DEADLINE restart.
* Handle SIGNAL-DEADLINE returning due to deferred deadlines where
necessary.
* Documentation.
Nikodemus Siivola [Thu, 18 Oct 2007 13:23:12 +0000 (13:23 +0000)]
1.0.10.48: new function TRY-RESTART
* Factor out all "find-restart and maybe invoke-restart" logic to one
place.
Nikodemus Siivola [Thu, 18 Oct 2007 12:56:51 +0000 (12:56 +0000)]
1.0.10.47: proper fix for "high-debug-known-function-inlining"
* Add ALLOW-INSTRUMENTING slot to CLAMBDA, taking it's value
from *ALLOW-INSTRUMENTING*.
* Require LAMBDA-ALLOW-INSTRUMENTING to be true for emission
of BIND/UNBIND-SENTINEL.
* Remove the earlier KLUDGE workaround.
Juho Snellman [Sun, 14 Oct 2007 23:40:55 +0000 (23:40 +0000)]
1.0.10.46: Inline the local functions in MAYBE-REHASH
* Fixes small hash-table speed regression introduced in
1.0.10.14 (and hidden by the other improvements in that
commit).
* 15-30% speedup over .14 on common hash operations.
Nikodemus Siivola [Thu, 11 Oct 2007 15:33:55 +0000 (15:33 +0000)]
1.0.10.45: save typecheck-function in slot-table even if location is not saved
* Otherwise non-standard metaclasses for which optimized instance
accessors can be used don't get the typecheck-fun from the wrapper
in the function returned from MAKE-OPTIMIZED-STD-WRITER-FUNCTION.
* Test-case.
* Missing NEWS entry for .44.
Nikodemus Siivola [Thu, 11 Oct 2007 13:13:07 +0000 (13:13 +0000)]
1.0.10.44: BUFFER-P, not BUFFERP
* Reported by Yuri Vishnevsky on sbcl-devel.
* Test-case coming Real Soon.
NIIMI Satoshi [Thu, 11 Oct 2007 09:20:36 +0000 (09:20 +0000)]
1.0.10.43: Fix NaN comparison on x86-64
Comisd and comiss sets all of ZF, PF, and CF of EFLAGS if its
operand contains NaN. Because PF is never set by normal
comparison, we use it to check NaN. This is same as "=" vops.
NIIMI Satoshi [Thu, 11 Oct 2007 03:19:27 +0000 (03:19 +0000)]
1.0.10.42: change dynamic space area on FreeBSD
If KVA_PAGES is extended from 256 to 384 (1GB to 1.5GB), stack starts
0x9FC00000. So change dynamic space from 0x60000000-0xA0000000 to
0x58000000-0x98000000.
Nowadays, 4GB of address space is too small, isn't it?
NIIMI Satoshi [Wed, 10 Oct 2007 21:36:49 +0000 (21:36 +0000)]
1.0.10.41: Remove unnecessary sched_yield()
Don't call sched_yield() on thread creation and retirement, since
the cost of sched_yield() is very high on FreeBSD 7.
Rudi Schlatte [Mon, 8 Oct 2007 11:02:52 +0000 (11:02 +0000)]
1.0.10.40: Fix build for windows when source directory contains spaces
Rudi Schlatte [Mon, 8 Oct 2007 08:34:44 +0000 (08:34 +0000)]
1.0.10.39: .gitignore additions for windows
Rudi Schlatte [Mon, 8 Oct 2007 06:19:26 +0000 (06:19 +0000)]
1.0.10.38: fix get-host-by-address on Darwin
sockaddr_in has a length field on Darwin but not on linux
Cyrus Harmon [Mon, 8 Oct 2007 04:30:21 +0000 (04:30 +0000)]
1.0.10.37: fix x86-64/darwin build
* use MAC_OS_X_VERSION_10_5 instead of __DARWIN_UNIX03 as the test
for new style headers as __DARWIN_UNIX03 is defined when building
x86-64 on 10.4
Cyrus Harmon [Mon, 8 Oct 2007 01:46:30 +0000 (01:46 +0000)]
1.0.10.36: support for Darwin versions that support __DARWIN_UNIX03
* on some versions of darwin symbols without a leading _ are dropped
by the linker. Use a leading _ on Darwin ldso_stub functions.
* Add a FIXME about 128-bit integeer passing being broken on x86-64
* _DARWIN_UNIX03 support
** os_context_t is now __darwin_ucontext
** eip -> __eip (and friends) - use macro hackery EIP etc...
** use darwin_ucontext/darwin_mcontext instead of struct
ucontext/mcontext where appropriate.
** struct ucontext -> os_context_t where appropriate
* allow socket-error for the aceptable errors in inet-socket error
test cases
* Add a fix me about a rumored OS bug fix that in name-service tests
Cyrus Harmon [Sun, 7 Oct 2007 20:21:58 +0000 (20:21 +0000)]
1.0.10.35: fix sb-posix test on linux
* look for sys instead of wheel when testing getgrnam
Andreas Fuchs [Sun, 7 Oct 2007 11:00:33 +0000 (11:00 +0000)]
1.0.10.34: Fix sb-posix:get{{gr,pw}nam},pwuid,grgid} for non-existing entries.
Used to signal memory fault, now returns NIL on queries for non-existing
users or groups.
Thiemo Seufer [Sat, 6 Oct 2007 21:53:45 +0000 (21:53 +0000)]
1.0.10.33: Lesson: Test before commit.
Cyrus Harmon [Sat, 6 Oct 2007 17:55:05 +0000 (17:55 +0000)]
1.0.10.32: better mach exception handling
* Patch from Ole Myren Rohne for better mach exception handling. This
deals with the case where the kernel started a signal handler
before our mach exception handler got a chance to run. This was
causing us to inappropriately SIGILL and die. This patch fixes
this, along with some general cleanup issues, including the
call_handler_on_thread wrapper function.
* TODO: this logic needs to be applied to the x86-64/darwin port as
well
Cyrus Harmon [Sat, 6 Oct 2007 16:53:53 +0000 (16:53 +0000)]
1.0.10.31: sb-posix support for getgrnam and getgrgid
* add alien-group and group protocol-class
* add define-gr-call macro a la define-pw-call
* alien definitions for getgrnam and getgrgid
* trivial test for above calls
NIIMI Satoshi [Sat, 6 Oct 2007 13:26:32 +0000 (13:26 +0000)]
1.0.10.30: Fix build without sb-thread
Nikodemus Siivola [Fri, 5 Oct 2007 14:48:40 +0000 (14:48 +0000)]
1.0.10.29: MUTEX refactoring & optimization
* (SETF MUTEX-VALUE) removed entirely. Using it was never sane,
and in the new setup entirely broken.
* On futex platforms, keep track of state of the mutex: free, taken,
or contested (taken and may have one ore more threads waiting).
* On release, if the mutex is not contested, there is no need to
perform the wakeup.
* Document assumptions made by GET-MUTEX & RELEASE-MUTEX better.
* Update FASL version.
Nikodemus Siivola [Fri, 5 Oct 2007 14:00:08 +0000 (14:00 +0000)]
1.0.10.28: export semaphore interface
* Semaphores are a fundamental threading construct -- export them.
Clean up the interface slightly: not (SETF SEMAPHORE-COUNT), note
that being a subclass of STRUCTURE-OBJECT is not guaranteed, etc.
NIIMI Satoshi [Fri, 5 Oct 2007 11:30:20 +0000 (11:30 +0000)]
1.0.10.27: Change unportable find option "-path"
The find option "-path" is not portable. Use "-prune" not to
recurse into CVS directory.
Rudi Schlatte [Fri, 5 Oct 2007 09:21:28 +0000 (09:21 +0000)]
1.0.10.26: More sb-posix documentation, thanks to Richard M Kreuter
Thiemo Seufer [Fri, 5 Oct 2007 00:55:27 +0000 (00:55 +0000)]
1.0.10.25: Build fix, add a missing bit of hashtable rewrite.
Thiemo Seufer [Thu, 4 Oct 2007 23:02:59 +0000 (23:02 +0000)]
1.0.10.24: Don't include CVS cruft in binary distributions.
Thiemo Seufer [Thu, 4 Oct 2007 23:01:57 +0000 (23:01 +0000)]
1.0.10.23: Shuffle def* foo around for less compiler warnings.
Thiemo Seufer [Thu, 4 Oct 2007 23:00:26 +0000 (23:00 +0000)]
1.0.10.22: Code formatting.
Thiemo Seufer [Thu, 4 Oct 2007 22:59:32 +0000 (22:59 +0000)]
1.0.10.21: MIPS will soon need an explicit -fPIC flag, too.
Thiemo Seufer [Thu, 4 Oct 2007 22:56:15 +0000 (22:56 +0000)]
1.0.10.20: Code provision for a future MIPS gencgc port.
Thiemo Seufer [Thu, 4 Oct 2007 22:54:48 +0000 (22:54 +0000)]
1.0.10.19: Check also for MIPS foreign_function_call lossage
... and fix a spelling typo.
Thiemo Seufer [Thu, 4 Oct 2007 22:52:51 +0000 (22:52 +0000)]
1.0.10.18: Define __USE_GNU only if it wasn't already defined.
Thiemo Seufer [Thu, 4 Oct 2007 22:48:45 +0000 (22:48 +0000)]
1.0.10.17: Reformat string, and fix compiler warning.
Rudi Schlatte [Thu, 4 Oct 2007 05:12:34 +0000 (05:12 +0000)]
1.0.10.16: Add some documentation for sb-posix
Mutilated contrib/sb-posix/README, removing
non-implementation-specific bits and referring to sb-alien instead.
... Some examples would be nice.
Rudi Schlatte [Tue, 2 Oct 2007 07:17:26 +0000 (07:17 +0000)]
1.0.10.15: Some sb-posix constants for stat().
Define S_IFMT and its friends S_IFIFO, S_IFDIR etc.
Juho Snellman [Sun, 30 Sep 2007 23:18:50 +0000 (23:18 +0000)]
1.0.10.14: remove locking and gc inhibition from hash-tables, power of 2 sizes
This commit removes a bunch of bottlenecks from the hash-table
implementation. It speeds up GETHASH, (SETF GETHASH) and
REMHASH by a factor of 2-4x (on platforms with a real
WITH-PINNED-OBJECTS) depending on the operation. On the flip
side, no automatic locking is done on tables any more, so
multi-threaded applications must do their own locking. (The
locking done by SBCL was always just an implementation detail,
not a part of the external interface). By popular demand it's
also still safe to have multiple readers on the same table
without locking.
Originally GCs were inhibited during most hash-table
operations for two reasons. To prevent the GC from rehashing a
table while a Lisp-side operation is going on, and to prevent
the GC from moving the key after the hash-value has been
calculated.
More recently, most hash-tables operations have acquired a
lock on the table in order to prevent two concurrent writers
from corrupting the chains. While it's never been the intent
for the standard data structures to be automatically
thread-safe in SBCL, this locking had to be done since corrupt
tables could lead to infinite GC loops.
Both the locking and the without-gcing are expensive
operations relative to the total cost of a hash-table lookup.
This commit removes both the gc inhibition and the locks.
Additionally we switch to power of two table size, which
allows calculating a cheaper hash -> bucket with cheaper
operations than MOD.
* The GC no longer does the rehashing itself, but just marks
the hash-table as needing a rehash, which will then be done
Lisp-side when the table is next accessed. While it's
possible to find cases where the former behaviour has better
performance, they're very contrived.
* The hash-table operations that work on the chains now check
for loops in the chains, and signal an error if one is found.
* The hash-table operations now pin the key before calculating
the hash value (needed for EQ-based hash functions).
* Add a GC epoch value that GETHASH can use to check whether
a GC happened during the lookup. This is needed since another
thread calling GETHASH on the same table might have caused it
to be rehashed.
* Kill the old MUST-REHASH vector header, and replace it with a
slot in the HASH-TABLE structure. The overloading of the header
caused missed rehashings when both the GC and %%PUTHASH modified
it at the same time.
* Switch to power of two table sizes, with a slightly more complex
hash value -> bucket calculation than just taking the low bits,
which in many cases have a very skewed distribution for the existing
SBCL hash functions. Still a lot faster than using MOD.
* Leave in locking and GC inhibition during rehashing (needed to
allow multiple readers to coexist) and for weak hash-tables
(they need some GC support, and the code is much simpler when
all of the logic is in the GC instead of interleaved in the GC and
Lisp-side). Neither of these cases is performance critical.
William Harold Newman [Sat, 29 Sep 2007 13:27:29 +0000 (13:27 +0000)]
1.0.10.13: trivial change: Print more explicit message upon entering LDB.
William Harold Newman [Fri, 28 Sep 2007 12:56:23 +0000 (12:56 +0000)]
1.0.10.12: logged bug 417
Andreas Fuchs [Thu, 27 Sep 2007 19:25:25 +0000 (19:25 +0000)]
1.0.10.11: Make FEATUREP accept only symbols or conses.
One more restriction on feature expressions: They can contain only conses
or symbols. The previous implementation allowed conses and atoms.
Nikodemus Siivola [Thu, 27 Sep 2007 15:56:06 +0000 (15:56 +0000)]
1.0.10.10: theoretical #!+long-float fix to MAKE-MEMBER-TYPE
* 0.0l0, not 0.0lo. Thanks to Peter Graves.
Nikodemus Siivola [Thu, 27 Sep 2007 15:50:43 +0000 (15:50 +0000)]
1.0.10.9: symbol-macro expansion uses *MACROEXPAND-HOOK*
Thanks to Tobias Rittweiler.
Nikodemus Siivola [Thu, 27 Sep 2007 15:43:25 +0000 (15:43 +0000)]
1.0.10.8: correct nested DX implementation
* Instead of just checking for BASIC-COMBINATION-P,
HANDLE-NESTED-DYNAMIC-EXTENT needs to do the same check for each USE
as RECHECK-DYNAMIC-EXTENT-LVARS does.
* Tests.
Nikodemus Siivola [Wed, 26 Sep 2007 16:00:44 +0000 (16:00 +0000)]
1.0.10.7: multiply-used DX LVARS
* HANDLE-NESTED-DYNAMIC-EXTENT maps over all the uses of the LVAR,
collecting their argument lvars as well.
* RECHECK-DYNAMIC-EXTENT-LVARS accepts multiply-used DX LVARs,
checking that all uses support stack allocation.
* UPDATE-UVL-LIVE-SETS accepts multiply-used DX LVARs, doing that
lifetime merging with all uses.
* ...and OOPS, move the NEWS entries of .5 and .6 to a new section for
1.0.11...
Nikodemus Siivola [Wed, 26 Sep 2007 15:54:58 +0000 (15:54 +0000)]
1.0.10.6: nested DX allocation
* RECOGNIZE-DYNAMIC-EXTENT-LVARS needs to propagate DX information to
combination arguments, so that
(LET ((X (LIST (LIST 1 2) (LIST 3 4))))
(DECLARE (DYNAMIC-EXTENT X))
(FOO X))
does the right thing.
Nikodemus Siivola [Wed, 26 Sep 2007 15:44:23 +0000 (15:44 +0000)]
1.0.10.5: dynamic-extent CONS
* Extend EMIT-FIXED-ALLOC to support stack-allocation.
* Appropriate DEFOPTIMIZER for CONS.
Note: it seems like it should be a simple matter to support stack
allocation of all primitive objects allocated using the :ALLOC-TRANS
framework.
NIIMI Satoshi [Wed, 26 Sep 2007 11:42:32 +0000 (11:42 +0000)]
1.0.10.4: Use variable for SIG_MEMORY_FAULT on FreeBSD
When SIGBUS is signalled on FreeBSD 7, or SIGSEGV is signalled on
FreeBSD 6 or earlier, it's very hard to debug if both of the signals
are hooked for GC.
Instead to hook both signals, define SIG_MEMORY_FAULT as variable
and set it at startup.
NIIMI Satoshi [Wed, 26 Sep 2007 11:37:34 +0000 (11:37 +0000)]
1.0.10.3: Grovel suseconds_t
Instead to detect if suseconds_t exists by OS and architecture,
check it in grovel-features.sh.
NIIMI Satoshi [Wed, 26 Sep 2007 11:30:50 +0000 (11:30 +0000)]
1.0.10.2: Don't link unnecessary undefines.c on FreeBSD/i386
undefines.c is not required anymore on FreeBSD/i386. Since I'm not
sure it's required or not on other *BSD, add it to OS_SRC in
Config.x86-*bsd.
NIIMI Satoshi [Wed, 26 Sep 2007 11:24:01 +0000 (11:24 +0000)]
1.0.10.1: Cleanup code
Shut GCC warnings up.
William Harold Newman [Tue, 25 Sep 2007 16:18:24 +0000 (16:18 +0000)]
1.0.10: release, will be tagged as sbcl_1_0_10
NIIMI Satoshi [Wed, 19 Sep 2007 04:11:44 +0000 (04:11 +0000)]
1.0.9.64: Use umtx system call on FreeBSD to emulate futex.
Andreas Fuchs [Tue, 18 Sep 2007 08:05:28 +0000 (08:05 +0000)]
1.0.9.63: Disallow (:not) in #+/#- expressions.
NOT in feature expressions is specified to accept exactly one argument.
The previous implementation of FEATUREP allowed zero or one argument,
so tighten that check.
Might need to fix CLRFI-1, as well (-:
NIIMI Satoshi [Tue, 18 Sep 2007 06:43:42 +0000 (06:43 +0000)]
1.0.9.62: Performance and stability improvement of threading on FreeBSD
* Use GCC's Thread-Local Storage to store current thread.
(Tested on Linux x86 and x86-64, too)
* Restore lisp level TLS segment register at interrupt handler.
Christophe Rhodes [Mon, 17 Sep 2007 18:38:16 +0000 (18:38 +0000)]
1.0.9.61: Fix %report-reader-error
It was missing an apply (or, equivalently, a ~?) for printing
the simple-condition part of the error.
William Harold Newman [Sun, 16 Sep 2007 12:05:16 +0000 (12:05 +0000)]
1.0.9.60: partial fix for bug in STREAM-ERROR :REPORT
* added a new (partially #+NILed out) test case for a bug in
STREAM-ERROR :REPORT, where it expects STREAM-ERROR to have
SIMPLE-CONDITION-like properties
* partial fix for bug in test case:
** Define SB-INT:SIMPLE-READER-ERROR which portably behaves
the way that pre-ANSI code expected READER-ERROR to behave.
** Redo most internal references to READER-ERROR as references
to SB-INT:SIMPLE-READER-ERROR.
* (This is only a partial fix because PARSE-ERROR and
STREAM-ERROR still have similar issues.)
Christophe Rhodes [Wed, 12 Sep 2007 16:14:16 +0000 (16:14 +0000)]
1.0.9.59: Fix ash/smod61 on x86-64 for constant large shifts.
Thanks to Paul Khuong for the translation.
Christophe Rhodes [Wed, 12 Sep 2007 15:37:05 +0000 (15:37 +0000)]
1.0.9.58: Fix x86 smod30 ash with a constant large argument.
The usual thing: the VOP was written in the belief that the
compiler had already proved that the shift couldn't be
too large.
(Include test cases which should catch the analogous problem on
x86-64 as well as the problem just fixed.)
Nikodemus Siivola [Mon, 10 Sep 2007 21:42:20 +0000 (21:42 +0000)]
1.0.9.57: allow pv-optimizations for typechecking (SETF SLOT-VALUE)
* For each optimized access, store both the slot location and the
typecheck function in the permutation vector. If the write is in
safe code, use the typecheck function to validate the new value
before using it.
* Approximately ~5x as fast as SAFE-SET-SLOT-VALUE -- which is still
needed in a few places as ACCESSOR-SET-SLOT-VALUE doesn't do type
checking.
Nikodemus Siivola [Mon, 10 Sep 2007 20:55:11 +0000 (20:55 +0000)]
1.0.9.56: faster typechecking/optimized std-writer-method-functions
* Fetch the type-check-fun from wrapper-slot-table, not
from the slot-definition.
Nikodemus Siivola [Mon, 10 Sep 2007 20:34:23 +0000 (20:34 +0000)]
1.0.9.55: trivial src/pcl/vector.lisp cleanup
* There hasn't been a :DEFAULT kind in permutation vector code
for ages -- delete the code that handled it.
Nikodemus Siivola [Mon, 10 Sep 2007 14:29:38 +0000 (14:29 +0000)]
1.0.9.54: clean up old pv updating code
* Since the vectors that are updated are referred to by stale
wrappers, this seems to be quite useless: if an instance has an
invalid wrapper it will still have the old layout as well.
William Harold Newman [Mon, 10 Sep 2007 13:31:45 +0000 (13:31 +0000)]
1.0.9.53: trivial typo fixes
(I had batched these up as a "git baby steps" commit, but now
that's been delayed...)
Nikodemus Siivola [Mon, 10 Sep 2007 12:36:59 +0000 (12:36 +0000)]
1.0.9.52: copy propagation interfering with local calls
* Don't allow copy propagation to a local call argument, thus
preserving parallel assignment semantics -- to judge by the comment
above OK-COPY-REF, this is what it was ment to do in the first
place. Reported by Paul Khuong on sbcl-devel.
Nikodemus Siivola [Mon, 10 Sep 2007 12:14:42 +0000 (12:14 +0000)]
1.0.9.51: SB-CLTL2: implement FUNCTION-INFORMATION, touch VARIABLE-INFORMATION
* Based on work done by Larry D'Anna.
* Rewire VARIABLE-INFORMATION in a similar manner. Add a FIXME
note about lexically apparent special bindings, and document
the current state of affairs. Improve the documentation string.
* Tests, and more tests for VARIABLE-INFORMATION as well.
Nikodemus Siivola [Sun, 9 Sep 2007 12:39:16 +0000 (12:39 +0000)]
1.0.9.50: O(1) weak pointer scavenging
* Use self-pointer instead of NULL to mark the end of scavenged weak
pointer list in GC, which allows identifying unscavenged pointers
by the NULL next pointers.
* Scavenging a single weak pointer in gencgc is now an O(1) operation
instead of O(#scanned pointers so far).
Thanks to Paul Khuong.
Christophe Rhodes [Sun, 9 Sep 2007 07:46:18 +0000 (07:46 +0000)]
1.0.9.49: NetBSD 2.0 build fix (Aymeric Vincent sbcl-devel 2007-09-05)
No ptsname() and friends on NetBSD.
Christophe Rhodes [Sun, 9 Sep 2007 07:32:28 +0000 (07:32 +0000)]
1.0.9.48: texi2pdf rework (Aymeric Vincent sbcl-devel 2007-09-05)
It looks likely that texi2dvi is more ubiquitous than texi2pdf,
so just apply the patch.
... (plus whitespace and .cvsignore collateral changes).
Nikodemus Siivola [Sat, 8 Sep 2007 17:35:47 +0000 (17:35 +0000)]
1.0.9.47: VALID-WRAPPER-OF
* Renamed CHECK-OBSOLETE-INSTANCE/WRAPPER-OF, and fixed the semantics
so that it always returns the fresh wrapper.
* Add FIXME re CHECK-WRAPPER-VALIDITY name.
Nikodemus Siivola [Sat, 8 Sep 2007 17:32:22 +0000 (17:32 +0000)]
1.0.9.46: take pv-slot value from wrapper-slot-table
* Use NIL instead of T in the CAR of the SLOT-TABLE cells to indicate
"slot location not saved for some reason".
* Better slot location computation for SLOT-TABLES: remaining wrapper
copies now also copy the slot-table, and bootstrapping computes the
slot locations for rest of the interesting cases.
* Use SLOT-TABLE to obtain the slot location for permutation vectors
instead of WRAPPER-CLASS-SLOTS and WRAPPER-INSTANCE-SLOTS-LAYOUT.
Faster, and slowly point the way to getting rid of WRAPPER/LAYOUT
distinction.
* We deal with slot names in permutation vectors: remove the unused
code that computed things for non-slot-name places in them.
Nikodemus Siivola [Sat, 8 Sep 2007 16:20:35 +0000 (16:20 +0000)]
1.0.9.45: cleaner & thread-safe pv-table interning
* Rename the global interning tables and comment on
their purpose for easier understanding.
* Simplify by-slot indexing: PV-OFFSETS are not needed in the global
table.
* Lock around hash-table accesses.
Nikodemus Siivola [Sat, 8 Sep 2007 15:54:16 +0000 (15:54 +0000)]
1.0.9.44: clean up PV-TABLE constructor
Nikodemus Siivola [Sat, 8 Sep 2007 15:15:45 +0000 (15:15 +0000)]
1.0.9.43: .PV-CELL., use .PV. directly
* Now that .CALLS. are gone we can get rid of the extra indirection.
(Maybe we have to add it back later, but worry about that then.)
* Since .PV. is magical, also localize its bindings to vector.lisp,
instead of exposing the variable in PV-BINDING1's interface (which
is used elsewhere as well.)
Nikodemus Siivola [Sat, 8 Sep 2007 14:38:02 +0000 (14:38 +0000)]
1.0.9.42: remove stale CALLS support
* Permutation vector code contained infrastructure for optimizing
GF calls inside method bodies in a similar manner as slot accesses
are optimized, but this support was never finished.
While it would be nice to have it, right now clarity of the code
seems like a higher priority -- it seems to be simple enough to
reinstate when we want, and the end result is liable to be easier to
understand.
* Add a big FIXME with a short explanation of the optimization and
a pointer to this commit.
NIIMI Satoshi [Sat, 8 Sep 2007 00:18:18 +0000 (00:18 +0000)]
1.0.9.41: Fix test on unthreaded config.
Nikodemus Siivola [Fri, 7 Sep 2007 09:51:37 +0000 (09:51 +0000)]
1.0.9.40: fix 64 bit build
* Need to cast (BACKEND_PAGE_SIZE-1) to an unsigned long before bit
flipping, otherwise we're potentially cutting the high bits off the
allocation address.
Nikodemus Siivola [Thu, 6 Sep 2007 13:36:23 +0000 (13:36 +0000)]
1.0.9.39: thread stack memory leaks
* Since 1.0.9.30 we pad & align the per-thread areas after
allocation -- but we need to still pass the original address
and size to os_invalidate(), or else we leak.
* Also refactor the freeable_thread_stack stuff slightly for
less OAOOM.
* Whitespace in tests.
Nikodemus Siivola [Thu, 6 Sep 2007 11:49:15 +0000 (11:49 +0000)]
1.0.9.38: fix COMPARE-AND-SWAP
* :COMPARE-AND-SWAP-VOPS, not :COMPARE-AND-SWAP-VOP in *FEATURES*.
* Threaded cas tests. (threads.impure.lisp currently broken for
unrelated reasons since .30 or so.)
Thiemo Seufer [Wed, 5 Sep 2007 20:40:33 +0000 (20:40 +0000)]
1.0.9.37: MIPS shared libraries will need an explicit -fPIC soon-ish.
Thiemo Seufer [Wed, 5 Sep 2007 20:36:10 +0000 (20:36 +0000)]
1.0.9.36: Match the correct atomicity bit.
The wonders of copy & paste ...