00:00:58 -!- drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has quit [Remote host closed the connection] 00:31:37 -!- fisxoj [~fisxoj@c-24-12-190-29.hsd1.il.comcast.net] has quit [Ping timeout: 246 seconds] 00:38:16 foreignFunction [~niksaak@ip-4761.sunline.net.ua] has joined #sbcl 00:40:23 -!- Bike [~Glossina@67-5-229-86.ptld.qwest.net] has quit [Ping timeout: 252 seconds] 00:42:11 Bike [~Glossina@67-5-229-86.ptld.qwest.net] has joined #sbcl 00:54:30 -!- davazp [~user@79.Red-79-153-96.dynamicIP.rima-tde.net] has quit [Remote host closed the connection] 01:04:31 -!- wbooze [~wbooze@xdsl-78-35-133-142.netcologne.de] has quit [Read error: Connection reset by peer] 01:05:48 -!- foreignFunction [~niksaak@ip-4761.sunline.net.ua] has quit [Quit: Leaving.] 01:08:00 drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has joined #sbcl 01:26:45 fisxoj [~fisxoj@c-24-12-190-29.hsd1.il.comcast.net] has joined #sbcl 01:30:04 -!- slyrus [~chatzilla@173-228-44-92.dsl.static.sonic.net] has quit [Ping timeout: 246 seconds] 01:44:47 Krystof: more on modular type derivation: when the derived (non-modular) type is a subtype of the result type (i.e. the implicit mask-signed-field is a noop), do you see a reason not to return that? 01:45:18 AFAICT, we *always* give up for signed modarith type derivation. 02:07:31 abarch` [~user@ip-5-147-122-209.unitymediagroup.de] has joined #sbcl 02:11:43 -!- abarch [~user@ip-5-147-122-209.unitymediagroup.de] has quit [Ping timeout: 246 seconds] 02:15:13 -!- fisxoj [~fisxoj@c-24-12-190-29.hsd1.il.comcast.net] has quit [Ping timeout: 246 seconds] 02:38:48 -!- christoph_debian [~christoph@ppp-188-174-156-77.dynamic.mnet-online.de] has quit [Ping timeout: 256 seconds] 02:52:04 christoph_debian [~christoph@ppp-188-174-149-247.dynamic.mnet-online.de] has joined #sbcl 04:05:14 -!- rpg [~rpg@216.243.156.16.real-time.com] has quit [Quit: rpg] 04:17:19 -!- milanj [~milanj@cable-94-189-145-213.dynamic.sbb.rs] has quit [Ping timeout: 264 seconds] 04:24:21 attila_lendvai [~attila_le@95.56.77.101] has joined #sbcl 04:24:21 -!- attila_lendvai [~attila_le@95.56.77.101] has quit [Changing host] 04:24:21 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 04:25:49 ehaliewicz [~user@50-0-51-11.dsl.static.sonic.net] has joined #sbcl 04:49:00 -!- Quadrescence [~quad@unaffiliated/quadrescence] has quit [Quit: This computer has gone to sleep] 05:46:05 yacks [~py@180.151.36.168] has joined #sbcl 05:56:57 angavrilov [~angavrilo@217.71.227.190] has joined #sbcl 06:08:50 prxq [~mommer@mnhm-4d01299b.pool.mediaWays.net] has joined #sbcl 06:13:40 milanj [~milanj@cable-94-189-145-213.dynamic.sbb.rs] has joined #sbcl 06:19:24 slyrus [~chatzilla@107.200.11.156] has joined #sbcl 06:21:47 kanru [~kanru@118-163-10-190.HINET-IP.hinet.net] has joined #sbcl 06:55:55 loke [~loke@203.127.16.194] has joined #sbcl 06:56:46 -!- milanj [~milanj@cable-94-189-145-213.dynamic.sbb.rs] has quit [Ping timeout: 276 seconds] 07:18:36 -!- ehaliewicz [~user@50-0-51-11.dsl.static.sonic.net] has left #sbcl 07:19:12 joshe [~joshe@216.151.3.96] has joined #sbcl 07:23:43 -!- joshe [~joshe@216.151.3.96] has quit [Remote host closed the connection] 07:24:07 joshe [~joshe@216.151.3.96] has joined #sbcl 07:25:22 -!- Bike [~Glossina@67-5-229-86.ptld.qwest.net] has quit [Ping timeout: 276 seconds] 07:32:40 -!- slyrus [~chatzilla@107.200.11.156] has quit [Ping timeout: 246 seconds] 07:33:10 slyrus [~chatzilla@107.200.11.156] has joined #sbcl 07:39:50 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Quit: Leaving.] 07:55:51 -!- fe[nl]ix [~quassel@pdpc/supporter/professional/fenlix] has quit [Ping timeout: 256 seconds] 08:03:16 fe[nl]ix [~quassel@pdpc/supporter/professional/fenlix] has joined #sbcl 08:18:27 -!- drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has quit [Remote host closed the connection] 08:29:03 -!- abarch` [~user@ip-5-147-122-209.unitymediagroup.de] has quit [Ping timeout: 260 seconds] 08:34:58 -!- slyrus [~chatzilla@107.200.11.156] has quit [Ping timeout: 246 seconds] 08:42:30 Odyessus [~odyessus@089144192070.atnat0001.highway.a1.net] has joined #sbcl 08:53:25 teggi [~teggi@123.20.107.130] has joined #sbcl 09:19:49 -!- Odyessus [~odyessus@089144192070.atnat0001.highway.a1.net] has quit [Quit: Colloquy for iPad - http://colloquy.mobi] 09:23:55 abarch [~user@2001:638:504:2093:21d:9ff:fe30:1f87] has joined #sbcl 09:51:39 nicdev [user@2600:3c03::f03c:91ff:fedf:4986] has joined #sbcl 10:46:32 -!- easye` [~user@213.33.70.157] has quit [Ping timeout: 240 seconds] 10:49:06 easye` [~user@213.33.70.157] has joined #sbcl 10:50:12 Sagane [~Sagane@177.100-226-89.dsl.completel.net] has joined #sbcl 10:54:23 foreignFunction [~niksaak@ip-4761.sunline.net.ua] has joined #sbcl 10:55:22 milanj [~milanj@cable-94-189-145-213.dynamic.sbb.rs] has joined #sbcl 11:05:21 -!- foreignFunction [~niksaak@ip-4761.sunline.net.ua] has quit [Read error: Connection reset by peer] 11:13:55 foreignFunction [~niksaak@ip-4761.sunline.net.ua] has joined #sbcl 11:37:46 -!- easye` [~user@213.33.70.157] has quit [Quit: ERC Version 5.3 (IRC client for Emacs)] 11:38:34 easye [~user@213.33.70.157] has joined #sbcl 11:51:28 -!- Blkt` is now known as Blkt 12:14:02 -!- milanj [~milanj@cable-94-189-145-213.dynamic.sbb.rs] has quit [Read error: Operation timed out] 12:18:36 Quadrescence [~quad@unaffiliated/quadrescence] has joined #sbcl 12:37:11 LiamH [~none@pdp8.nrl.navy.mil] has joined #sbcl 12:44:27 sdemarre [~serge@91.180.80.126] has joined #sbcl 12:55:36 -!- sdemarre [~serge@91.180.80.126] has quit [Ping timeout: 240 seconds] 12:59:06 -!- Sagane [~Sagane@177.100-226-89.dsl.completel.net] has quit [Read error: Connection reset by peer] 13:06:28 -!- foreignFunction [~niksaak@ip-4761.sunline.net.ua] has quit [Quit: Leaving.] 13:10:41 slyrus [~chatzilla@107.200.11.156] has joined #sbcl 13:30:16 drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has joined #sbcl 13:30:33 milanj [~milanj@82.117.199.26] has joined #sbcl 13:44:38 -!- milanj [~milanj@82.117.199.26] has quit [Quit: Leaving] 13:45:07 milanj [~milanj@82.117.199.26] has joined #sbcl 14:10:25 wbooze [~wbooze@xdsl-78-35-152-231.netcologne.de] has joined #sbcl 14:31:28 -!- Quadrescence [~quad@unaffiliated/quadrescence] has quit [Quit: This computer has gone to sleep] 14:42:31 nyef [~nyef@c-50-157-244-41.hsd1.ma.comcast.net] has joined #sbcl 14:42:44 G'morning all. 14:44:17 morning SBCL! 14:44:21 I found a fun little issue in gencgc last night, while I was working on moving the page table entries from bitfields to a flags word. 14:44:22 pkhuong: I do not see a reason not to tighten 14:44:37 (I think the expectation was that that case would be rare) 14:45:17 (For the curious, it makes conservatism a lot more conservative than it really "should" be.) 14:45:38 now I am curious 14:46:10 Have a look at code_page_p(), and see if you can figure out what's wrong with it, and what the effects are. 14:46:38 I have a fix in my local tree already. 14:46:50 And I also tracked down when it was introduced. 14:49:09 And, while we're on the topic of bug fixes, are we in code freeze yet? 14:50:25 oh lord I forgot :-) 14:50:40 So... is that a "no"? 14:51:17 I would guess that CODE_PAGE_FLAG is not defined how one might imagine it would be defined 14:51:41 is there a distinction between "may this page contain code" and "does this page contain code"? 14:51:41 It's defined as (BOXED_PAGE_FLAG | UNBOXED_PAGE_FLAG). 14:52:28 Code pages only contain code, other pages MAY contain code, as genesis doesn't break things up by page type. 14:52:52 ah, so & is not the right test, == is? 14:53:04 or in other words CODE_PAGE_FLAG is not a flag 14:53:57 Right, the "allocation" field is three bits wide, there's an "open region flag" which is a flag, and the "boxed page flag" really is a flag, but it's also part of the two-bit "allocation type" field which indicates if the page is a free page, a boxed page, an unboxed page, or a code page. 14:54:07 (Code counts as boxed.) 14:54:29 lovely 14:55:11 so does the fix for this make some of the gc failure cases go away on x86, do you know? 14:55:44 Indeed. And now that I'm thinking about it again, we can't "just" fix it, because I'm not sure that it will pin live functions that aren't in code pages. 14:56:22 nyef: don't we have code to scan for code objects enclosing some address? 14:56:47 I don't know that it gets applied here. 14:58:01 Sagane [~Sagane@177.100-226-89.dsl.completel.net] has joined #sbcl 15:02:33 hi everybody, i have read a recent paper about regalloc for my google project . in this paper there is a lot about inter-class (SC) aliasing 15:02:58 is such information already available in SBCL? 15:04:26 conflicts are tracked at the storage-base level. 15:04:41 multiple storage classes map to the same storage base, and not all storage classes are the same size. 15:05:45 the latter freedom is only used for stack slots nowadays, although we still pretend that all > byte GPRs actually take up two GPR slots. 15:07:45 That latter bit about GPRs is x86oid-specific, none of the other backends do so, although some of them might use a similar trick with double-float storage? 15:08:05 nyef: oh right, SPARC does that for doubles, iirc. 15:08:12 Haven't played with other platforms. 15:09:18 I'd expect to see similar games with storage for doubles vs. singles on the float stacks across the board. 15:12:23 Sigh. Did not want to dive into Groetschell Lovasz and Schrijver this morning, especially not for SBCL. 15:14:16 Trying to remember the trick to colour nested interval graphs: the sort of intersection graph we get for TNs that are live over their lexical scope. 15:14:59 -!- drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has quit [Remote host closed the connection] 15:16:25 -!- cmm [~cmm@bzq-79-177-123-220.red.bezeqint.net] has quit [Ping timeout: 246 seconds] 15:17:27 cmm [~cmm@bzq-79-177-123-220.red.bezeqint.net] has joined #sbcl 15:17:28 drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has joined #sbcl 15:20:11 pkhuong: what do you mean by Lovasz and schrijver? 15:24:34 This tome of mostly completely theoretical results, but that sometimes has useful stuff. If there's any concerned observer: I don't ever intend to commit an ellipsoid method in SBCL ;) 15:30:45 sdemarre [~serge@91.180.80.126] has joined #sbcl 15:35:19 -!- sdemarre [~serge@91.180.80.126] has quit [Ping timeout: 264 seconds] 15:38:15 pkhuong: so there is no aliasing between SC of different size? 15:38:27 SCs 15:39:56 pkhuong: thanx for the reference - a very cute book! 15:41:07 attila_lendvai [~attila_le@92.47.191.246] has joined #sbcl 15:41:07 -!- attila_lendvai [~attila_le@92.47.191.246] has quit [Changing host] 15:41:07 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 15:45:37 abarch: what do you mean? Conflicts between TNs are mapped at the level of (contiguous and potentially aligned) slot numbers in a storage base. 15:47:02 So if A and B are simultaneously live and are in storage classes that map to the same storage base, they won't be assigned conflicting slots in the SB. 15:47:27 If they aren't however, a given SB slot could be used for both A and B, even though they're in different SCs. 15:52:01 pkhuong: sure 15:52:58 the size and alignment of the SC only matters when determining which slots to look at for conflicts. 15:55:59 Bike [~Glossina@67-5-229-86.ptld.qwest.net] has joined #sbcl 16:01:20 bah. reverse LexBFS is reasonable to get a perfect elimination ordering, but it's in the wrong order: we'll tend to colour stack slots from the topmost lexical scope last. 16:01:36 I suggest double-reverse LexBFS 16:02:47 unfortunately the result won't be perfect ;) 16:03:36 pkhuong: i mean that for example one WORD-REG is in conflict with two BYTE-REGs 16:04:53 abarch: yes. TNs are conceptuallt coloured with multiple (storage-base, slot) locations. 16:21:03 awesome... The intuitive ordering for live ranges that correspond to lexical scopes (pack the parent before any children) is optimal for that class of intersection graphs (comparability graphs). 16:37:01 (http://www.dcs.gla.ac.uk/~pat/jchoco/clique/papersColour/Reed%20B.A.%20-%20Recent%20Advances%20in%20Algorithms%20and%20Combinatorics(2002)(368).pdf#page=79, p. 71) 17:09:02 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Quit: Leaving.] 17:31:31 -!- yacks [~py@180.151.36.168] has quit [Read error: Operation timed out] 17:34:33 fisxoj [~fisxoj@c-24-12-190-29.hsd1.il.comcast.net] has joined #sbcl 17:41:27 foreignFunction [~niksaak@ip-4761.sunline.net.ua] has joined #sbcl 17:48:00 yacks [~py@180.151.36.168] has joined #sbcl 18:07:27 -!- drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has quit [Remote host closed the connection] 18:09:46 drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has joined #sbcl 18:18:02 -!- ASau [~user@p5797FF73.dip0.t-ipconnect.de] has quit [Ping timeout: 252 seconds] 18:19:15 -!- teggi [~teggi@123.20.107.130] has quit [Remote host closed the connection] 18:21:21 ASau [~user@p5797FF73.dip0.t-ipconnect.de] has joined #sbcl 18:25:25 -!- fisxoj [~fisxoj@c-24-12-190-29.hsd1.il.comcast.net] has quit [Ping timeout: 256 seconds] 18:31:45 -!- drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has quit [Remote host closed the connection] 18:33:55 ASau` [~user@p4FF96B51.dip0.t-ipconnect.de] has joined #sbcl 18:37:43 -!- ASau [~user@p5797FF73.dip0.t-ipconnect.de] has quit [Ping timeout: 264 seconds] 18:38:08 -!- ASau` is now known as ASau 18:41:26 drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has joined #sbcl 19:16:52 -!- slyrus [~chatzilla@107.200.11.156] has quit [Ping timeout: 246 seconds] 20:03:18 -!- flip214 [~marek@unaffiliated/flip214] has quit [Ping timeout: 245 seconds] 20:06:07 Quadrescence [~quad@unaffiliated/quadrescence] has joined #sbcl 20:09:08 flip214 [~marek@86.59.100.100] has joined #sbcl 20:09:08 -!- flip214 [~marek@86.59.100.100] has quit [Changing host] 20:09:08 flip214 [~marek@unaffiliated/flip214] has joined #sbcl 20:11:53 -!- Quadrescence [~quad@unaffiliated/quadrescence] has quit [Quit: This computer has gone to sleep] 20:19:19 -!- flip214 [~marek@unaffiliated/flip214] has quit [Ping timeout: 256 seconds] 20:20:14 flip214 [~marek@86.59.100.100] has joined #sbcl 20:20:14 -!- flip214 [~marek@86.59.100.100] has quit [Changing host] 20:20:14 flip214 [~marek@unaffiliated/flip214] has joined #sbcl 20:57:12 -!- prxq [~mommer@mnhm-4d01299b.pool.mediaWays.net] has quit [Remote host closed the connection] 21:09:40 -!- milanj [~milanj@82.117.199.26] has quit [Quit: Leaving] 21:23:11 -!- ASau [~user@p4FF96B51.dip0.t-ipconnect.de] has quit [Read error: Connection reset by peer] 21:23:35 ASau` [~user@p4FF96B51.dip0.t-ipconnect.de] has joined #sbcl 21:27:14 -!- ASau` is now known as ASau 21:33:40 -!- angavrilov [~angavrilo@217.71.227.190] has quit [Remote host closed the connection] 21:37:05 oh. I have to test this: I think sb-bignum croaks on bignums with >= 2^32 limbs. 21:37:39 MOST-POSITIVE-BIGNUM strikes again? 21:38:21 no, just a bare 32 instead of n-word-bits somewhere in a type declaration. 21:39:49 scymtym_ [~user@ip-5-147-122-209.unitymediagroup.de] has joined #sbcl 21:40:27 on 32 bit platforms, though, m-p-b is relatively tiny. Only 16 M limbs. 21:41:28 At some point you run out of heap space to hold the bignum, anyway, surely? 21:42:10 I have 128 GB here 21:48:28 ... I remember when I couldn't even fit 128 KB onto a disk. 21:48:58 (Okay, that's a lie. But the trick to doing so involved a hole punch.) 21:49:30 so do I. I'm still amazed that this workstation has more L3 capacity than my first real hard drive. 21:52:18 -!- drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has quit [Remote host closed the connection] 21:53:27 I wouldn't be entirely surprised if any of my machines has more L2 or even L1 than my first hard drive. 22:11:45 Thinking about the code_page_p() thing, once we've dumped a warm core, all code objects should be in code pages, simply because of the full GC with no conservative roots as part of core-save. 22:12:18 nyef: i.e. whenever we do a partial pickup? 22:12:39 Ah, right. Partial pickup, exactly. 22:12:52 Partial pickup preserves page properties. 22:13:33 Even if it IS a stupid name for what it is and does. 22:15:55 So the trick would be to have genesis emit a page-table, to have a special mode that short-circuits the reference validation for pinning the way that code_page_p() does now for use on a non-partial-pickup core, or to make sure that there's always a boxed reference to any active functions or code objects in the conservative roots. 22:17:08 Oh, or to do a full GC of the cold-core when first loading it, thus invalidating the map file, but making sure that all of the code objects are on code pages. 22:18:00 1) sounds ideal, but I expect 2) will be more realistic 22:18:15 drmeister [~drmeister@d205-250-223-135.bchsia.telus.net] has joined #sbcl 22:19:46 Mmm. Reworking genesis is a project in and of itself, while adding gencgc_partial_pickup to the test terms in preserve_pointer() isn't very difficult. 22:20:25 -!- Sagane [~Sagane@177.100-226-89.dsl.completel.net] has quit [Read error: Connection reset by peer] 22:22:23 can you think of a malloc that'd work well for a mark and sweep-ish last generation? dlmalloc is small, but has huge overhead for conses. 22:24:01 ehaliewicz [~user@50-0-51-11.dsl.static.sonic.net] has joined #sbcl 22:24:58 I was thinking of having a freespace-widetag for mark-and-sweep, TBH. 22:25:42 It's still heap space, after all. 22:25:51 I'm almost considering a stupid bitmap + first address fit. 22:28:29 -!- LiamH [~none@pdp8.nrl.navy.mil] has quit [Quit: Leaving.] 22:33:25 Okay, I need to get going if I'm to have dinner at a reasonable time this evening. 22:33:57 I'll go with option two on the code_page_p() thing for now, and should have a patch ready for review tomorrow. 22:35:06 -!- nyef [~nyef@c-50-157-244-41.hsd1.ma.comcast.net] has quit [Quit: G'night all.] 22:41:25 minion: memo for nyef: do we also use code_page_p when opening allocation regions? 22:41:25 Remembered. I'll tell nyef when he/she/it next speaks. 22:46:02 sdemarre [~serge@91.180.80.126] has joined #sbcl 23:11:59 -!- sdemarre [~serge@91.180.80.126] has quit [Ping timeout: 260 seconds] 23:46:22 davazp [~user@79.Red-79-153-96.dynamicIP.rima-tde.net] has joined #sbcl