From: Juho S. <js...@us...> - 2005-11-03 12:41:20
|
Update of /cvsroot/sbcl/sbcl In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv6126 Modified Files: NEWS version.lisp-expr Log Message: 0.9.6.14: Faster allocation on x86-64 (25% speedup on memory-intensive CL-BENCH tests, 5% on more generic stuff like COMPILER): * Inline allocation was using a memory-to-register XCHG (latency 16 on Athlon 64) on the fast path. Use a temporary register instead. * Change the temp-tn from r13 to r11, which has a shorter encoding (results in smaller core and better icache behaviour) TODO: Check whether the XCHG issue also caused the bizarre P4 performance problems with the (disabled) x86 inline allocation support, and whether anything can be done to fix the problem. Using the same solution is probably impossible due to the lack of extra registers. Index: NEWS =================================================================== RCS file: /cvsroot/sbcl/sbcl/NEWS,v retrieving revision 1.906 retrieving revision 1.907 diff -u -d -r1.906 -r1.907 --- NEWS 2 Nov 2005 22:19:47 -0000 1.906 +++ NEWS 3 Nov 2005 12:41:07 -0000 1.907 @@ -16,6 +16,7 @@ merged with *DEFAULT-PATHNAME-DEFAULTS*. * optimization: performance improvements to IO on file streams of :ELEMENT-TYPE CHARACTER + * optimization: much faster memory allocation on x86-64 changes in sbcl-0.9.6 relative to sbcl-0.9.5: * bug fix: add a workaround to SBCL looping infinitely at startup on Index: version.lisp-expr =================================================================== RCS file: /cvsroot/sbcl/sbcl/version.lisp-expr,v retrieving revision 1.2529 retrieving revision 1.2530 diff -u -d -r1.2529 -r1.2530 --- version.lisp-expr 2 Nov 2005 22:19:47 -0000 1.2529 +++ version.lisp-expr 3 Nov 2005 12:41:07 -0000 1.2530 @@ -17,4 +17,4 @@ ;;; checkins which aren't released. (And occasionally for internal ;;; versions, especially for internal versions off the main CVS ;;; branch, it gets hairier, e.g. "0.pre7.14.flaky4.13".) -"0.9.6.13" +"0.9.6.14" |