1 2012-12-16 00:44:11 <BlueMatt> sipa: hmm...I thought I had added that previously, Ill go do that. re: different hash init, meh...if Im gonna bother changing it, might as well just be nHashNum + nTweak
2 2012-12-16 00:44:12 <BlueMatt> sipa: hmm...I thought I had added that previously, Ill go do that. re: different hash init, meh...if Im gonna bother changing it, might as well just be nHashNum + nTweak
3 2012-12-16 00:44:18 <BlueMatt> (its not like murmur3 is that bad)
4 2012-12-16 00:48:05 <sipa> but but... you've got no idea how much cpu time it cost me to find the optimal multiplier!
5 2012-12-16 00:48:34 <sipa> more seriously, it doesn't matter indeed
6 2012-12-16 00:50:07 <sipa> or at least it shouldn't
7 2012-12-16 00:50:08 <sipa> or at least it shouldn't
8 2012-12-16 00:50:35 <sipa> BlueMatt: the partial hash tree probably needs some examples to illustrate, but meh...
9 2012-12-16 01:07:04 <swaits> what's the best practice for determining a final balance for an arbitrary address? apparently I can't determine this just from bitcoind's RPC?
10 2012-12-16 01:07:05 <BlueMatt> sipa: heh, so you did actually run something for optimal multiplier (I was thinking of coding that up...)
11 2012-12-16 01:07:29 <sipa> BlueMatt: yes, i was intrigued by the question :)
12 2012-12-16 01:07:33 <BlueMatt> sipa: and...yea, I was lazy with the pmt docs, I just figured your comment was good enough and anyone else can dig into the code
13 2012-12-16 01:07:51 <BlueMatt> sipa: its certainly a fun problem
14 2012-12-16 01:07:51 <gmaxwell> I assumed sipa searched for a good multiplier.
15 2012-12-16 01:07:52 <BlueMatt> sipa: its certainly a fun problem
16 2012-12-16 01:08:01 <gmaxwell> (thats what I would have done! :P)
17 2012-12-16 01:08:36 <BlueMatt> gmaxwell: I assumed he did something, not run simulations, though thats also what I would have done (not that it counts for as much...)
18 2012-12-16 01:10:29 <sipa> BlueMatt: the number i came up with guarantees a 10-bit difference between any two inits for any tweak, and on average close to 100 bits
19 2012-12-16 01:10:33 <sipa> eh, 11 bits
20 2012-12-16 01:12:37 <BlueMatt> nice
21 2012-12-16 01:13:10 <sipa> it's not guaranteed to be the best possible, but likely very clos
22 2012-12-16 01:13:11 <sipa> it's not guaranteed to be the best possible, but likely very clos
23 2012-12-16 01:13:44 <BlueMatt> well now I cant but change it...
24 2012-12-16 01:13:45 <BlueMatt> well now I cant but change it...
25 2012-12-16 01:13:57 <sipa> haha :D
26 2012-12-16 01:22:04 <BlueMatt> might as well state this here for anyone else (dont see any evidence it effects x86): probably a bad idea to run bitcoind on arm, it seems to like to crash with bad_allocs after a while
27 2012-12-16 01:25:16 <sipa> what are you trying to run it on?
28 2012-12-16 01:26:22 <BlueMatt> http://www.hardkernel.com/renewal_2011/products/prdt_info.php?g_code=g133999328931 (Samsung Exynos 4412)
29 2012-12-16 01:40:34 <jgarzik> BlueMatt: BE or LE?
30 2012-12-16 01:40:35 <jgarzik> BlueMatt: BE or LE?
31 2012-12-16 01:40:57 <jgarzik> if LE, it is probably unaligned accesses somewhere
32 2012-12-16 01:41:13 <BlueMatt> LE, yea, thats why gmaxwell suggested, it sounds like a probably source
33 2012-12-16 01:41:31 <BlueMatt> itl run for a few days before falling over, but its probably not safe...
34 2012-12-16 01:44:55 <BlueMatt> https://github.com/bitcoin/bitcoin/issues/2111
35 2012-12-16 01:45:44 <sipa> bad_alloc only gets thrown by new afaik
36 2012-12-16 01:46:03 <BlueMatt> I believe so
37 2012-12-16 01:51:54 <phantomcircuit> hmm
38 2012-12-16 01:52:33 <phantomcircuit> im starting to think that adding a default nice(1) to bitcoin{d,-qt} would be a good idea...
39 2012-12-16 01:52:34 <phantomcircuit> im starting to think that adding a default nice(1) to bitcoin{d,-qt} would be a good idea...
40 2012-12-16 01:53:01 <Luke-Jr> phantomcircuit: ionice moreso
41 2012-12-16 01:53:30 <phantomcircuit> yeah
42 2012-12-16 01:54:16 <gmaxwell> phantomcircuit: ioload s much much less of an issue on git.
43 2012-12-16 01:55:23 <phantomcircuit> still there's very little reason that bitcoin should be running with the same priority as the rest of the system
44 2012-12-16 01:55:41 <phantomcircuit> actually just certain threads but the same basic point still holds
45 2012-12-16 01:58:00 <gmaxwell> is there something like ionice in windows?
46 2012-12-16 01:58:27 <gmaxwell> also, would nicing it out just result in @#$#@$#@ windows killing it? :(
47 2012-12-16 01:59:17 <gmaxwell> "this application hasn't responded in 3ms. Because we're confused and think we're a cooperative multitasking operating system we're going to put this big idiot button in front of you right now to kill your program."
48 2012-12-16 02:01:48 <sipa> Vista and above: http://msdn.microsoft.com/en-us/library/windows/hardware/ff549710(v=vs.85).aspx
49 2012-12-16 02:02:22 <sipa> OSX doesn't have anything similar, it seems
50 2012-12-16 02:02:23 <sipa> OSX doesn't have anything similar, it seems
51 2012-12-16 02:02:47 <Luke-Jr> "(Windows Drivers)"
52 2012-12-16 02:02:57 <Luke-Jr> don't think we want to write a driver :|
53 2012-12-16 02:03:58 <phantomcircuit> sipa, yeah you're looking at kernel driver api
54 2012-12-16 02:04:36 <sipa> ha!
55 2012-12-16 02:05:00 <sipa> that's probably a good way to get better windows performance: have bitcoin run in kernel space :p
56 2012-12-16 02:20:47 <phantomcircuit> sipa, io priority on windows is linked to cpu priority
57 2012-12-16 02:20:48 <phantomcircuit> sipa, io priority on windows is linked to cpu priority
58 2012-12-16 02:21:03 <phantomcircuit> so the only thing that needs to happen is SetPriorityClass
59 2012-12-16 02:21:04 <phantomcircuit> so the only thing that needs to happen is SetPriorityClass
60 2012-12-16 02:21:23 <phantomcircuit> possibly SetThreadPriority so the gpu doesn't get stuck
61 2012-12-16 02:21:24 <phantomcircuit> possibly SetThreadPriority so the gpu doesn't get stuck
62 2012-12-16 02:21:28 <phantomcircuit> *gui
63 2012-12-16 02:39:17 <nus> http://www.bitsum.com/pl_io_priority.php
64 2012-12-16 02:46:07 <pingdrive> does anyone know how to splice a two dimension java array?
65 2012-12-16 02:48:18 <pingdrive> grr slice
66 2012-12-16 02:52:22 <gmaxwell> phantomcircuit: I'd worry about the gui thread spinning on locks held by background stuff.
67 2012-12-16 02:52:23 <gmaxwell> phantomcircuit: I'd worry about the gui thread spinning on locks held by background stuff.
68 2012-12-16 02:54:30 <phantomcircuit> gmaxwell, yeah there is that risk
69 2012-12-16 02:54:35 <phantomcircuit> but done properly it could work
70 2012-12-16 02:54:36 <phantomcircuit> but done properly it could work
71 2012-12-16 02:54:48 <phantomcircuit> might not be worth the effort though
72 2012-12-16 03:00:23 <gmaxwell> phantomcircuit: might as well just have a bitocoin nice level that inserts yields in between expensive operations
73 2012-12-16 03:00:45 <phantomcircuit> hmm
74 2012-12-16 03:00:47 <phantomcircuit> yeah i guess
75 2012-12-16 04:20:52 <gmaxwell> sipa: seems we don't handle the blktree not being openable gracefully:
76 2012-12-16 04:20:53 <gmaxwell> CLevelDB(): error opening database environment Invalid argument: not an sstable (bad magic number)
77 2012-12-16 04:20:56 <gmaxwell> bitcoin in AppInit()
78 2012-12-16 04:34:12 <gmaxwell> The leveldb code is _much_ more robust against corrupt database input than bdb, even with the checksums disabled.
79 2012-12-16 04:34:18 <gmaxwell> BDB was trivial to make crash with corruption.
80 2012-12-16 04:39:56 <gmaxwell> oh neat. I got a testnet database into a state that has no blocks and can't be recovered from with reindex or loadblock.
81 2012-12-16 05:05:55 <sta> screw you gmaxwell
82 2012-12-16 05:06:23 <sta> ^ that was a reflex to seeing too many gmaxwell lines on my screen
83 2012-12-16 05:07:18 <sta> and it's ugly and green
84 2012-12-16 05:07:46 <sta> should have kept the ignore
85 2012-12-16 11:14:12 <st1> i won't say it. i won't argue. with some people, you don't want to argue
86 2012-12-16 11:14:13 <st1> i won't say it. i won't argue. with some people, you don't want to argue
87 2012-12-16 11:31:48 <st1> because he added some easy ones, just after the "event"
88 2012-12-16 11:31:49 <st1> because he added some easy ones, just after the "event"
89 2012-12-16 11:31:55 <st1> but he doesn't get me
90 2012-12-16 11:31:56 <st1> but he doesn't get me
91 2012-12-16 12:42:21 <MC-Eeepc> anyone tried running ecdsa checking on GPU?
92 2012-12-16 12:50:34 <sipa> MC-Eeepc: won't be too easy to implement i think
93 2012-12-16 12:50:35 <sipa> MC-Eeepc: won't be too easy to implement i think
94 2012-12-16 12:51:22 <sipa> MC-Eeepc: oclvanitygen has a EC addition on GPU
95 2012-12-16 12:51:23 <sipa> MC-Eeepc: oclvanitygen has a EC addition on GPU
96 2012-12-16 12:52:01 <MC-Eeepc> cool
97 2012-12-16 12:52:40 <sipa> but to do ECDSA you need EC multiplications as well
98 2012-12-16 12:54:46 <MC-Eeepc> if the parallelisation can be done properly wouldnt it be rather fast
99 2012-12-16 12:57:29 <sipa> very rough estimate: if a GPU can do 350 MH/s, it'd do 100k signature verifications per second
100 2012-12-16 12:57:52 <MC-Eeepc> that sounds fast
101 2012-12-16 12:59:30 <sipa> i may be a factor 2-3 off, but yes
102 2012-12-16 16:20:12 <Scrat> how well can bitcoind handle thousands of accounts?
103 2012-12-16 16:42:22 <rdymac> how is the 0.8 going? Can it be downloaded to test?
104 2012-12-16 16:42:30 <rdymac> is there binary for mac already?
105 2012-12-16 16:42:49 <sipa> Scrat: not very well
106 2012-12-16 16:43:16 <sipa> rdymac: gavin has certainly built mac binaries already for testing, but none are publically available afaik
107 2012-12-16 16:43:48 <rdymac> Anyway I could get one for testing?
108 2012-12-16 16:43:49 <rdymac> Anyway I could get one for testing?
109 2012-12-16 16:44:07 <Scrat> sipa: so is it using a linear lookup?
110 2012-12-16 16:44:08 <Scrat> sipa: so is it using a linear lookup?
111 2012-12-16 16:44:19 <Scrat> or serial rather
112 2012-12-16 16:44:21 <sipa> Scrat: yup
113 2012-12-16 16:44:22 <sipa> Scrat: yup
114 2012-12-16 16:44:42 <sipa> Scrat: well, no, but some operations do require a linear scan of all wallet transactions or wallet keys
115 2012-12-16 16:44:43 <sipa> Scrat: well, no, but some operations do require a linear scan of all wallet transactions or wallet keys
116 2012-12-16 16:49:39 <rdymac> I saw a link here a few days ago, but I missed it
117 2012-12-16 16:50:59 <sipa> https://bitcointalk.org/index.php?topic=129861.0 ?
118 2012-12-16 17:29:34 <sipa> gmaxwell: btw... a blktree database corrupted such that -reindex doesn't repair it? wth?
119 2012-12-16 19:01:44 <winemaker> hello
120 2012-12-16 19:02:14 <winemaker> what software / lib's etc do i need to compile bitcoin-qt - i'm using win8
121 2012-12-16 19:03:11 <Luke-Jr> winemaker: I think there's a total of one person who compiles Bitcoin-Qt on Windows anymore (Diapolo), and he's not here right now.
122 2012-12-16 19:03:21 <Luke-Jr> winemaker: it *should* be documented in the doc/ directory though
123 2012-12-16 19:03:39 <Luke-Jr> winemaker: note that official Windows releases are compiled on Ubuntu
124 2012-12-16 19:04:06 <winemaker> i'm so sorry but win8 fucked up my ubuntu and i was to lazy to reinstall grub
125 2012-12-16 19:04:35 <sipa> nothing to be sorry about, but it just means that it's hard to help you
126 2012-12-16 19:04:49 <Luke-Jr> ^
127 2012-12-16 19:05:04 <sipa> as most developers are linux users, and windows releases are built on linux too
128 2012-12-16 19:33:00 <etotheipi_> is it true the genesis transaction can't be spent?
129 2012-12-16 19:33:05 <etotheipi_> What is preventing it?
130 2012-12-16 19:33:40 <sipa> the fact that it is initialized without being added to the transaction database by every version of the reference client
131 2012-12-16 19:33:58 <sipa> maybe a bug, maybe intentionally
132 2012-12-16 19:35:04 <etotheipi_> oh, it would be seen as invalid if a spend showed up in the blockchain (and thus wouldn't really be in the blockchain that most people are following)
133 2012-12-16 19:35:33 <sipa> yes
134 2012-12-16 19:35:57 <etotheipi_> then how does the client verify the genesis hash? or does it not?
135 2012-12-16 19:36:19 <sipa> it doesn't
136 2012-12-16 19:36:27 <sipa> it's hardcoded
137 2012-12-16 19:36:36 <gmaxwell> The genesis block is hardcoded, but thats orthorgonal to it being unspendable.
138 2012-12-16 19:36:48 <etotheipi_> got it
139 2012-12-16 19:36:52 <gmaxwell> E.g. it would be perfectly possible to validate it but not insert it into the index.
140 2012-12-16 19:37:06 <sipa> or the other way around
141 2012-12-16 21:53:21 <nus> ehm, the bitcoint.it ssl certificate has expired
142 2012-12-16 21:55:29 <sipa> ;;later tell MagicalTux bitcoin.it's SSL certificate has expired
143 2012-12-16 21:55:30 <gribble> The operation succeeded.
144 2012-12-16 23:19:26 <elux> "FYI, the SSL cert for the wiki site has expired: https://en.bitcoin.it/wiki/Main_Page There's no obvious owner on the page that I can contact."
145 2012-12-16 23:19:29 <elux> https://bitcointalk.org/index.php?topic=131136.0
146 2012-12-16 23:21:41 <Luke-Jr> elux: yep, MagicalTux has been notified
147 2012-12-16 23:22:02 <MagicalTux> it's pending issuance by CA