1.0.16.27: function-ify ERROR-CALL and GENERATE-ERROR-CODE on x86-64
* Port of 1.0.16.10 to x86-64.
- [DBH] src/assembly/x86-64/alloc.lisp
- [DBH] src/assembly/x86-64/assem-rtns.lisp
- [DBH] src/compiler/generic/array.lisp
- [DBH] src/compiler/x86-64/arith.lisp
- [DBH] src/compiler/x86-64/array.lisp
- [DBH] src/compiler/x86-64/call.lisp
- [DBH] src/compiler/x86-64/cell.lisp
- [DBH] src/compiler/x86-64/macros.lisp
- [DBH] src/compiler/x86-64/move.lisp
- [DBH] src/compiler/x86-64/subprim.lisp
- [DBH] src/compiler/x86-64/type-vops.lisp
- [DBH] src/compiler/x86-64/values.lisp
- [DBH] src/compiler/x86/macros.lisp
- [DBH] version.lisp-expr
(May 8, 2008 03:05 PM)
1.0.16.26: dx allocation thru CAST nodes
* Allow DX allocation of LVARs thru cast nodes without type checks.
* Since it is not obvious to me that all uses of CAST-VALUE must be
in the same component as the cast itself, AVER that.
* Results of MAKE-ARRAY can once more be stack allocated. Regression
caused by different handling of TRULY-THE introducing cast nodes
where there previously were none.
* Tests.
- [DBH] NEWS
- [DBH] src/compiler/ir1util.lisp
- [DBH] src/compiler/locall.lisp
- [DBH] src/compiler/physenvanal.lisp
- [DBH] tests/dynamic-extent.impure.lisp
- [DBH] version.lisp-expr
(May 8, 2008 11:52 AM)
1.0.16.25: restore buildability on threaded x86-64
* GENERATE-ERROR-CODE is still a macro on x86-64, so no quoting
TLS-EXHAUSTED-ERROR.
- [DBH] src/assembly/x86-64/alloc.lisp
- [DBH] version.lisp-expr
(May 8, 2008 10:08 AM)
1.0.16.24: slightly more verbose errors for SB-FOO symbols in genesis
* ...faster to fix when the error tells you what the problematic
symbol was.
- [DBH] src/compiler/generic/genesis.lisp
- [DBH] version.lisp-expr
(May 8, 2008 08:53 AM)
1.0.16.23: type-inconsistencies in HASH-TABLE accessors
* DEFKNOWN types did not match the slot types. Change the defknowns
to match the DEFSTRUCT.
- [DBH] src/compiler/fndb.lisp
- [DBH] version.lisp-expr
(May 8, 2008 08:38 AM)
1.0.16.22: FIXED-ALLOC to use MAYBE-PSEUDO-ATOMIC on x86 and x86-64.
* PA is not needed with stack allocation.
* Also rename the first argument to MAYBE-PSEUDO-ATOMIC to
NOT-REALLY-P to avoid confusion.
- [DBH] src/compiler/x86-64/alloc.lisp
- [DBH] src/compiler/x86-64/macros.lisp
- [DBH] src/compiler/x86/alloc.lisp
- [DBH] src/compiler/x86/macros.lisp
- [DBH] version.lisp-expr
(May 8, 2008 08:21 AM)
1.0.16.21: lose informatively when the tls is full
Instead of eventually producing a segv or some random corruption:
(progv (loop for i below 5000
collect (make-symbol (format nil "xxx~D" i)))
(loop for i below 5000 collect i))
- [DBH] package-data-list.lisp-expr
- [DBH] src/assembly/x86-64/alloc.lisp
- [DBH] src/assembly/x86/alloc.lisp
- [DBH] src/code/interr.lisp
- [DBH] src/compiler/generic/interr.lisp
- [DBH] src/compiler/generic/parms.lisp
- [DBH] src/runtime/dynbind.c
- [DBH] src/runtime/thread.c
- [DBH] version.lisp-expr
(May 8, 2008 07:14 AM)
Last tested version: 1.0.16.27. Significant changes:
- from SBCL:x86_64 revision 1.0.16.26 to 1.0.16.27:
- from SBCL:x86_64 revision 1.0.16.25 to 1.0.16.26:
- from SBCL:x86_64 revision 1.0.16.23 to 1.0.16.24:
- 1D-ARRAYS increased from (.04)s to (.05)s (+9%)
- from SBCL:x86_64 revision 1.0.16.22 to 1.0.16.23:
- from SBCL:x86_64 revision 1.0.16.21 to 1.0.16.22:
- 1D-ARRAYS decreased from (.05 ± .002)s to (.04)s (-16%)
- BIGNUM/PARI-200-5 increased from (.04 ± .001)s to (.044000000)s (+9%)
- DESTRUCTIVE decreased from (.22 ± .003)s to (.209 ± .002)s (-5%)
- FPRINT/UGLY decreased from (1.164 ± .008)s to (1.105 ± .002)s (-5%)
- RICHARDS increased from (.46 ± .01)s to (.55 ± .03)s (+17%)
- from SBCL:x86_64 revision 1.0.16.20 to 1.0.16.21:
(May 8, 2008 12:00 AM)
Last tested version: 1.0.16.20. Significant changes:
- from SBCL:x86_64 revision 1.0.16.19 to 1.0.16.20:
- from SBCL:x86_64 revision 1.0.16.18 to 1.0.16.19:
- from SBCL:x86_64 revision 1.0.16.17 to 1.0.16.18:
(May 7, 2008 12:00 AM)
1.0.16.20: update GIT-FOR-SBCL-HACKERS
Make it a bit less misleading about the actual treatment of
version.lisp-expr.
- [DBH] doc/GIT-FOR-SBCL-HACKERS.txt
- [DBH] version.lisp-expr
(May 6, 2008 04:54 PM)
1.0.16.19: cleanups motivated by clisp host-2
- [DBH] src/code/cross-byte.lisp
- [DBH] src/code/gc.lisp
- [DBH] src/code/late-extensions.lisp
- [DBH] src/code/win32-pathname.lisp
- [DBH] src/compiler/array-tran.lisp
- [DBH] src/compiler/ir1report.lisp
- [DBH] version.lisp-expr
(May 6, 2008 04:43 PM)
1.0.16.18: Fixes to get clisp through host-1
Just a couple of ignores and code rearrangements.
- [DBH] src/code/early-source-location.lisp
- [DBH] src/code/source-location.lisp
- [DBH] src/compiler/ir1tran.lisp
- [DBH] src/compiler/policy.lisp
- [DBH] version.lisp-expr
(May 6, 2008 10:45 AM)
1.0.16.17: log bug #426: inlining failure involing multiple nested calls
* Not a regression, but apparently of CMUCL vintage.
- [DBH] BUGS
- [DBH] version.lisp-expr
(May 5, 2008 01:26 PM)
1.0.16.16: Use declared element type in AREF short-circuit transform
The short-circuit transformation introduced in 1.0.2.17
removed the system's understanding of the declared array element
type (as opposed to the upgraded array element type).
Reintroduce the cleverness, and hope that the use of type=
doesn't remove all the slowdown.
(Issue noted by vy on #lisp afternoon 2008-05-04 BST)
- [DBH] src/compiler/array-tran.lisp
- [DBH] tests/compiler.pure.lisp
- [DBH] version.lisp-expr
(May 4, 2008 08:14 PM)
1.0.16.15: fix TRANSFORM-LIST-ITEM-SEEK for ADJOIN with constant list arg
* Urk, missed a spot. There is probably little sense to open code ADJOIN
like ASSOC and MEMBER, so let's not.
* Tests.
- [DBH] src/compiler/seqtran.lisp
- [DBH] tests/list.pure.lisp
- [DBH] version.lisp-expr
(May 4, 2008 03:06 PM)
Last tested version: 1.0.16.17. Significant changes:
- from SBCL:x86_64 revision 1.0.16.16 to 1.0.16.17:
- FRPOLY/BIGNUM decreased from (.24 ± .003)s to (.22 ± .005)s (-8%)
(May 3, 2008 12:00 AM)
Last tested version: 1.0.16.16. Significant changes:
- from SBCL:x86_64 revision 1.0.16.15 to 1.0.16.16:
- from SBCL:x86_64 revision 1.0.16.14 to 1.0.16.15:
(May 2, 2008 12:00 AM)
Last tested version: 1.0.16.14. Significant changes:
- from SBCL:x86_64 revision 1.0.16.13 to 1.0.16.14:
(May 1, 2008 12:00 AM)
1.0.16.14:
Fix compiler warning.
- [DBH] src/runtime/cheneygc.c
- [DBH] version.lisp-expr
(April 29, 2008 11:43 PM)
1.0.16.13: use TRANSFORM-LIST-ITEM-SEEK for ADJOIN as well
* Now that the freeze was cancelled, do this properly...
- [DBH] NEWS
- [DBH] package-data-list.lisp-expr
- [DBH] src/code/list.lisp
- [DBH] src/compiler/seqtran.lisp
- [DBH] version.lisp-expr
(April 29, 2008 01:58 PM)
1.0.16.12: revert ADJOIN change from 1.0.16.5
* The correct expansion is a bit too hairy to implement nicely
with a compiler macro.
* Test-cases.
Second take using TRANSFORM-LIST-ITEM-SEEK coming after the freeze.
- [DBH] NEWS
- [DBH] src/code/list.lisp
- [DBH] tests/list.pure.lisp
- [DBH] version.lisp-expr
(April 28, 2008 09:48 PM)
Last tested version: 1.0.16.13. Significant changes:
- from SBCL:x86_64 revision 1.0.16.12 to 1.0.16.13:
(April 27, 2008 12:00 AM)
Last tested version: 1.0.16.12. Significant changes:
- from SBCL:x86_64 revision 1.0.16.11 to 1.0.16.12:
- BROWSE decreased from (.279 ± .002)s to (.25 ± .003)s (-12%)
- DESTRUCTIVE increased from (.2)s to (.22 ± .004)s (+6%)
- FIB-RATIO increased from (.04)s to (.04 ± .001)s (+13%)
- FRPOLY/FLOAT increased from (.399 ± .003)s to (.439)s (+9%)
- RICHARDS decreased from (.58 ± .03)s to (.495 ± .002)s (-15%)
- SLURP-LINES decreased from (.808 ± .005)s to (.697 ± .002)s (-14%)
- WALK-LIST/SEQ decreased from (.03)s to (.03)s (-13%)
(April 26, 2008 12:00 AM)
Last tested version: 1.0.16.11. Significant changes:
- from SBCL:x86_64 revision 1.0.16.10 to 1.0.16.11:
- 1D-ARRAYS increased from (.04)s to (.05)s (+9%)
- BROWSE increased from (.253 ± .002)s to (.279 ± .002)s (+9%)
- FIB-RATIO decreased from (.040000000)s to (.04)s (-7%)
- FRPOLY/BIGNUM decreased from (.257 ± .002)s to (.229 ± .002)s (-11%)
- MANDELBROT/DFLOAT decreased from (2.180 ± .008)s to (1.615 ± .003)s (-26%)
- PI-DECIMAL/BIG decreased from (.260 ± .001)s to (.244 ± .002)s (-6%)
- RICHARDS increased from (.466 ± .003)s to (.58 ± .03)s (+20%)
- SLURP-LINES increased from (.716 ± .006)s to (.808 ± .005)s (+11%)
- SUM-PERMUTATIONS increased from (1.592 ± .002)s to (1.687 ± .008)s (+6%)
- WALK-LIST/MESS increased from (.03 ± .001)s to (.03)s (+16%)
- from SBCL:x86_64 revision 1.0.16.9 to 1.0.16.10:
- FFT decreased from (.1)s to (.05 ± .001)s (-7%)
- RICHARDS decreased from (.49 ± .01)s to (.466 ± .003)s (-5%)
- WALK-LIST/MESS decreased from (.03)s to (.03 ± .001)s (-19%)
(April 25, 2008 12:00 AM)
1.0.16.11: ANY-REG is good for pritimitive-type T on x86/x86-64
* Report and test case by Stelian Ionescu.
- [DBH] NEWS
- [DBH] src/compiler/generic/primtype.lisp
- [DBH] tests/alien.impure.lisp
- [DBH] version.lisp-expr
(April 24, 2008 08:36 AM)
1.0.16.10: function-ify ERROR-CALL and GENERATE-ERROR-CODE on x86
* Saves ~120k in core size.
* Should be done for other backends and ERROR-CALL &co.
refactored appropriately.
- [DBH] src/assembly/x86/assem-rtns.lisp
- [DBH] src/compiler/generic/array.lisp
- [DBH] src/compiler/x86/arith.lisp
- [DBH] src/compiler/x86/array.lisp
- [DBH] src/compiler/x86/call.lisp
- [DBH] src/compiler/x86/cell.lisp
- [DBH] src/compiler/x86/macros.lisp
- [DBH] src/compiler/x86/move.lisp
- [DBH] src/compiler/x86/subprim.lisp
- [DBH] src/compiler/x86/type-vops.lisp
- [DBH] src/compiler/x86/values.lisp
- [DBH] version.lisp-expr
(April 24, 2008 04:08 AM)
1.0.16.9: restore buildability from CMUCL
CMUCL justifiably emits warning on (ecase ... (...) (t ...)).
Use ((t) ...) instead as the last clause. Reported by Andreas
Franke (sbcl-devel 2008-04-23).
- [DBH] NEWS
- [DBH] src/compiler/generic/vm-macs.lisp
- [DBH] src/compiler/generic/vm-tran.lisp
- [DBH] src/compiler/srctran.lisp
- [DBH] version.lisp-expr
(April 23, 2008 09:43 PM)
1.0.16.8: NCONC with dx &rest lists
* Remove MAYBE-INLINE declaration: with dx &rest list inlining NCONC doesn't
yield any real benefits.
* Also delete some dead code, and NCONC2 -- interestingly unlike with
APPEND, a compiler-macro to NCONC2 seems to hurt more then it helps
(not that it hurts in any real way).
- [DBH] NEWS
- [DBH] src/code/list.lisp
- [DBH] version.lisp-expr
(April 23, 2008 06:21 PM)
Last tested version: 1.0.16.9. Significant changes:
- from SBCL:x86_64 revision 1.0.16.8 to 1.0.16.9:
- 1D-ARRAYS increased from (.040000000)s to (.05)s (+12%)
- FFT increased from (.05)s to (.1)s (+10%)
- from SBCL:x86_64 revision 1.0.16.7 to 1.0.16.8:
- from SBCL:x86_64 revision 1.0.16.6 to 1.0.16.7:
- from SBCL:x86_64 revision 1.0.16.5 to 1.0.16.6:
- BIGNUM/ELEM-1000-100 decreased from (.119)s to (.11 ± .002)s (-6%)
- BROWSE decreased from (.276000000)s to (.256 ± .002)s (-7%)
- FIB-RATIO decreased from (.04)s to (.040000000)s (-6%)
- FPRINT/PRETTY increased from (2.248 ± .007)s to (2.403 ± .011)s (+6%)
- RICHARDS decreased from (.455 ± .003)s to (.43 ± .01)s (-5%)
- WALK-LIST/SEQ increased from (.03)s to (.03)s (+9%)
- from SBCL:x86_64 revision 1.0.16.4 to 1.0.16.5:
- FIB-RATIO increased from (.04)s to (.04)s (+9%)
- FRPOLY/BIGNUM decreased from (.240 ± .002)s to (.227 ± .002)s (-6%)
- LOAD-FASL increased from (.360 ± .002)s to (.379 ± .003)s (+5%)
- RICHARDS decreased from (.53 ± .01)s to (.455 ± .003)s (-14%)
- STRING-CONCAT increased from (34.139 ± .058)s to (37.210 ± .192)s (+8%)
- WALK-LIST/SEQ decreased from (.03)s to (.03)s (-9%)
(April 20, 2008 12:00 AM)