gmane.lisp.steel-bank.devel

Re: Running tests in win32 for 1.0.23

Nikodemus Siivola — 2008-12-01T22:08:26

Yes (sorry for the slow answer to your other questions.) Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: Running tests in win32 for 1.0.23

Matthew D. Swank — 2008-12-01T21:25:15

On Mon, 1 Dec 2008 17:12:16 +0000 (UTC) Matthew D. Swank gmail.com> wrote: This is a reply to a message I submitted using gmane. I noticed that it isn't showing up on gmane, but I did get a copy here at my regular, subscribed address. I also submitted a post to sbcl-general via gmane a few weeks ago (it was about specifying line-endings for streams read by read-line under win32); it also never showed up on gmane, but I was able to find a copy in the archives unanswered (I don't subscribe to general, help, or whatever it is). Are people seeing posts I make via gmane? Thanks, Matt

Re: [patch] hppa-hpux

Christophe Rhodes — 2008-12-01T20:33:34

Thiemo's points are good; for this one, though, there's my preferred option of using LISP_FEATURE_HPUX, which doesn't depend on the compiler at all (but on our own autogenerated header files). Best, Christophe ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: code objects on separate pages

Nikodemus Siivola — 2008-12-01T20:10:25

On Wed, Oct 29, 2008 at 9:53 AM, Nikodemus Siivola random-state.net> wrote: Essentially this, split into more digestable hunks, has been merged to CVS. Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: RFC: --script commandline argument

Nikodemus Siivola — 2008-12-01T19:49:44

On Sat, Oct 11, 2008 at 3:38 PM, Nikodemus Siivola random-state.net> wrote: I just realized something. SB-EXECUTABLE (the contrib, not the SAVE-LISP-AND-DIE extension of course) already has this problem. So maybe adding shebang lines to fasls would not be such a horrible step backwards after all -- just don't make them executable by default, but document the situation. Maybe. ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: [WISHPOINT] mark code portions belonging toinlined functions in DISASSEMBLE's output

Tobias C. Rittweiler — 2008-12-01T18:07:27

Make the annotation of inlined functions be optional, e.g. via adding an additional &key parameter. So general readability is not of concern anymore. It's ok not to cover every case, and still make it practically very useful. -T. ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: [patch] hppa-hpux

Thiemo Seufer — 2008-12-01T17:22:43

That's impressive work! Some random observation from a quick glance at the patch: - The patch is HUGE. It mixes generic bug fixes, spelling fixes, hppa fixes/changes and support for HP/UX. Splitting it in topical parts would make a review a lot easier. - The hppa/hpux conditionals need to be sorted out. The are apparently mostly used as synonyms. I saw places where hppa was used for OS-specifics, and I also saw places where hpux was used to make architecture-specific distinctions. - In the C runtime, both #ifdef hpux and #ifdef HPUX are bad choices. The standard reserves this namespace for applications, I presume the compiler also provides nicer (standard conformant) predefines like __hpux__ or _HPUX. Thiemo ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event a

Running tests in win32 for 1.0.23

Matthew D. Swank — 2008-12-01T17:12:16

Lots of messages then... /done with DEFSTRUCT /checking constructed structure /checking no read-only writers /doing INLINEP=T /doing INLINEP=NIL /done with INLINEP loop /done with PROGN for COLONTYPE=LIST Dynamic space usage is: 38,334,752 bytes. Read-only space usage is: 2,968 bytes. Static space usage is: 1,640 bytes. Control stack usage is: 2,008 bytes. Binding stack usage is: 200 bytes. Garbage collection is currently enabled. Breakdown for dynamic space: 9,175,968 bytes for 11,620 code objects. 8,736,784 bytes for 202,997 instance objects. 7,266,008 bytes for 908,251 cons objects. 5,039,760 bytes for 109,377 simple-vector objects. 2,787,336 bytes for 313,491 bignum objects. 7,700,032 bytes for 256,261 other objects. 40,705,888 bytes for 1,801,997 dynamic objects (space total.) Dynamic space usage is: 28,873,448 bytes. Read-only space usage is: 2,968 bytes. Static space usage is: 1,640 bytes. Control stack usage is: 2

Re: [patch] hppa-hpux

Brian Mastenbrook — 2008-12-01T16:49:41

Wow! And here I thought HPPA support was dead for sure. I assume it compiles itself? This was a persistent problem under Linux/ HPPA, which Christophe was eager to blame on Linux kernel bugs. Can you provide binaries for this on a regular basis (ideally monthly)? If not, can you provide access to a machine that can be used to build binaries? Have you tried this under HP/UX on Itanium? -- Brian Mastenbrook brian< at >mastenbrook.net http://brian.mastenbrook.net/ ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: [patch] hppa-hpux

V. Glenn Tarcea — 2008-12-01T16:39:06

Ughh.... My apologies I replied to the wrong email. I seem to be having a bad email day... :-) Glenn V. Glenn Tarcea gtarcea< at >umich.edu Hey buddy, can you paradigm? -----Original Message----- From: V. Glenn Tarcea [mailto:gtarcea< at >umich.edu] Sent: Monday, December 01, 2008 11:32 AM To: 'Larry Valkama'; 'sbcl-devel< at >lists.sourceforge.net' Subject: RE: [Sbcl-devel] [patch] hppa-hpux Thursday works for me. Glenn V. Glenn Tarcea gtarcea< at >umich.edu Hey buddy, can you paradigm? -----Original Message----- From: Larry Valkama [mailto:remlali< at >users.sourceforge.net] Sent: Monday, December 01, 2008 12:28 PM To: sbcl-devel< at >lists.sourceforge.net Subject: [Sbcl-devel] [patch] hppa-hpux Listserver ate my previous announcement ( too large attachment, waiting for approval ). take two: * decreased initial heap usage, works better with default kernels. * tested with sbcl 1.0.22.21 attaching the patch. have fun! /larry ------------------------------------------------------------------------- This SF.Net email is

Re: [patch] hppa-hpux

V. Glenn Tarcea — 2008-12-01T16:32:25

Thursday works for me. Glenn V. Glenn Tarcea gtarcea< at >umich.edu Hey buddy, can you paradigm? -----Original Message----- From: Larry Valkama [mailto:remlali< at >users.sourceforge.net] Sent: Monday, December 01, 2008 12:28 PM To: sbcl-devel< at >lists.sourceforge.net Subject: [Sbcl-devel] [patch] hppa-hpux Listserver ate my previous announcement ( too large attachment, waiting for approval ). take two: * decreased initial heap usage, works better with default kernels. * tested with sbcl 1.0.22.21 attaching the patch. have fun! /larry ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

[patch] hppa-hpux

Larry Valkama — 2008-12-01T17:28:23

------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Sbcl-devel mailing list Sbcl-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sbcl-devel

Re: faster generic arithmetic dispatch on x86 andx86-64

Nikodemus Siivola — 2008-12-01T15:48:49

On Mon, Dec 1, 2008 at 4:38 PM, Nikodemus Siivola random-state.net> wrote: Merged as 1.0.23.3. Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: FUNCTION-INFORMATION,using git to contribute to SBCL

Nikodemus Siivola — 2008-12-01T15:46:56

Fixed in 1.0.23.1, finally. (Leaf-where-from should have been :declared.) Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: [patch] fix STRING->UTF8's handling of the STARTargument

Nikodemus Siivola — 2008-12-01T15:46:16

Merged as 1.0.23.2, thank you! Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: Wrong :format-arguments in fill-pointer-error

Nikodemus Siivola — 2008-12-01T15:44:57

Merged as 1.0.23.4, thank you! Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: [Sbcl-help] sb-sys:vector-sap on multidimensionalarrays

Nikodemus Siivola — 2008-12-01T15:35:46

... I would like to stick this: ;;; User visible extension (declaim (ftype (function (simple-array) (values (simple-array * (*)) &optional)) simple-array-vector)) (defun simple-array-vector (array) "ARRAY must be a SIMPLE-ARRAY. If ARRAY is multidimensional, returns the underlying one-dimensional SIMPLE-ARRAY, which shares storage with ARRAY. Otherwise returns ARRAY." ;; KLUDGE: Without TRULY-THE the system is not smart enough to figure out that ;; (1) SIMPLE-ARRAY without ARRAY-HEADER-P is a vector (2) the data vector of ;; a SIMPLE-ARRAY is a vector. (truly-the (simple-array * (*)) (if (array-header-p array) (%array-data-vector array) array))) in SB-EXT. Any objections? Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize

Re: faster generic arithmetic dispatch on x86 andx86-64

Nikodemus Siivola — 2008-12-01T14:38:37

On Mon, Dec 1, 2008 at 3:32 PM, Nikodemus Siivola random-state.net> wrote: It's CLC, RET ... but it's quite a win! On this test: (defun test (n x y) (declare (fixnum n) (integer x y)) (dotimes (i n) (setf x (+ y x)) (setf x (+ y x)) (setf x (+ y x))) x) (time (test 100000000 42 1)) The previous version was 20% faster than HEAD on x86, yours is 30% faster. On x86-64 the improvement is a pfiffling 20%, biti winnage anyways! Updated patch attached. Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Sbcl-devel mailing list Sbcl-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/s

Re: faster generic arithmetic dispatch on x86 andx86-64

Nikodemus Siivola — 2008-12-01T13:32:36

The reason for the two JMPs was static branch prediction. ...which this would do fine with. I'll build and measure, and let you know. Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: faster generic arithmetic dispatch on x86 andx86-64

Nathan Froyd — 2008-12-01T13:25:18

On Mon, Dec 1, 2008 at 8:09 AM, Nikodemus Siivola random-state.net> wrote: Surely you can get away with a single (inst jmp :z DO-BODY) can't you? I also wonder if there'd be icache gains to be made by moving the body prior to the static-fun case, so you'd get routines that looked like: test rcx, 7 jnz DO-STATIC-FUN BODY ret DO-STATIC-FUN: ... this wins on icache and branch prediction (maybe?) for the common case of fixnum addition. You'd waste an extra 'ret' at the end due to DEFINE-ASSEMBLY-ROUTINE wonkiness, but that's not a big deal in the grand scheme of things. -Nathan ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/

faster generic arithmetic dispatch on x86 and x86-64

Nikodemus Siivola — 2008-12-01T13:09:40

Inspired by Slava's recent factor work on the same area: instead of two branches OR the values so that a single test is enough to determine if both are fixnums. Merging after freeze. Cheers, -- Nikodemus ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/_______________________________________________ Sbcl-devel mailing list Sbcl-devel< at >lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sbcl-devel