01:06.10 | *** join/#maemo-ssu joshgillies (~josh@124-171-115-220.dyn.iinet.net.au) |
01:16.06 | *** join/#maemo-ssu andre__ (~andre@wikimedia/aklapper) |
02:34.59 | *** join/#maemo-ssu AndrewX192 (~andrew@20aa4b03b846.click-network.com) |
02:34.59 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
02:41.46 | *** join/#maemo-ssu AndrewX192 (~andrew@131.191.23.122) |
02:41.46 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
02:46.47 | *** join/#maemo-ssu AndrewX192_ (~andrew@131.191.69.108) |
02:46.47 | *** join/#maemo-ssu AndrewX192_ (~andrew@unaffiliated/andrewx192) |
02:51.47 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
02:57.25 | *** join/#maemo-ssu AndrewX192 (~andrew@131.191.69.108) |
02:57.26 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
03:02.32 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
03:07.33 | *** join/#maemo-ssu AndrewX192 (~andrew@131.191.23.122) |
03:07.33 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
03:30.51 | *** join/#maemo-ssu AndrewX192_ (~andrew@131.191.69.108) |
03:30.51 | *** join/#maemo-ssu AndrewX192_ (~andrew@unaffiliated/andrewx192) |
03:35.51 | *** join/#maemo-ssu AndrewX192 (~andrew@unaffiliated/andrewx192) |
03:53.40 | *** join/#maemo-ssu amiconn_ (amiconn@rockbox/developer/amiconn) |
03:56.20 | jon_y | merlin1991: ping |
04:03.35 | *** join/#maemo-ssu DocScrutinizer05 (~HaleBopp@openmoko/engineers/joerg) |
04:03.59 | jon_y | merlin1991: did you ever encounter something about a missing libfakeroot-sysv.so when doing http://wiki.maemo.org/CSSU-thumb_toolchain_setup_%28gcc4.7.2-linaro%29 ? |
04:14.29 | merlin1991 | nope, but I never set one up |
04:14.47 | merlin1991 | freemangordon: ^^ |
04:36.40 | *** join/#maemo-ssu _xnt14 (~xnt14@xceleo.us) |
04:44.45 | jon_y | merlin1991: I copied the so from FREMANTLE_ARMEL |
04:44.49 | jon_y | no idea if it is OK |
04:47.49 | jon_y | merlin1991: also, is there a way for scratchbox to access outside the chroot? |
05:18.13 | merlin1991 | nope |
06:56.20 | *** join/#maemo-ssu dhbiker (~dhbiker@APN-123-117-7-gprs.simobil.net) |
07:21.04 | jon_y | oh haha wow, I've been building the kernel wrong all these time |
07:21.26 | jon_y | need to extract kernel sources first into the git checkout |
07:21.34 | jon_y | nothing about that in the README |
08:29.55 | *** join/#maemo-ssu jonwil (~jonwil@27-33-137-199.static.tpgi.com.au) |
09:12.54 | *** join/#maemo-ssu Estel_ (~Estel@Maemo/community/contributor/Estel-) |
10:29.18 | *** join/#maemo-ssu M4rtinK (~M4rtinK@mail.melf.eu) |
11:03.18 | *** join/#maemo-ssu NIN101 (~NIN@p5DD28B5D.dip0.t-ipconnect.de) |
12:05.19 | *** join/#maemo-ssu BCMM (~BCMM@unaffiliated/bcmm) |
12:21.08 | *** join/#maemo-ssu arcean (~arcean@aacs56.neoplus.adsl.tpnet.pl) |
12:28.36 | *** join/#maemo-ssu freemangordon (~freemango@130-204-50-168.2074221835.ddns.cablebg.net) |
12:30.10 | freemangordon | merlin1991: x86 |
12:41.27 | *** join/#maemo-ssu kolp (~quassel@212.255.23.54) |
12:46.08 | *** join/#maemo-ssu luf (~luf@ip-89-102-208-114.net.upcbroadband.cz) |
12:59.23 | luf | freemangordon, merlin1991: what about thle libxml2 merge request? I think we agreed it's bug and should be fixed. |
12:59.39 | luf | https://gitorious.org/community-ssu/libxml2/merge_requests/1 |
13:03.18 | DocScrutinizer05 | ((is there a way for scratchbox to access outside the chroot?)) depends on definition of "access". You can do all sorts of wget apt-get etc. You also can bindmount dirs form host into chroot, that's how a chroot works |
13:04.06 | DocScrutinizer05 | actually SB can do apt-get only because host's /dev is bindmounted into the chroot |
13:05.09 | DocScrutinizer05 | those bindmounts are also the reason why it's a VERY BAD idea to rm -rf /scratchbox |
13:05.28 | DocScrutinizer05 | ask mohammad if you wanna know how bad an idea that is |
13:20.33 | jon_y | DocScrutinizer05: I forgot about bind mount |
13:30.23 | luf | gregoa: ping |
13:30.23 | gregoa | luf: You sent me a contentless ping. This is a contentless pong. Please provide a bit of information about what you want and I will respond when I am around. |
13:32.26 | luf | I built libxml2 deb package and it put into depends (using shlibs) zlib1g (>= 1:1.2.1). But I patch it and the required version of zlib1g should be 1.2.3. I don't see where the shlibs takes the 1.2.1 version. |
13:32.38 | luf | gregoa: ^^^ |
14:17.04 | *** join/#maemo-ssu arcean_ (~arcean@aacs56.neoplus.adsl.tpnet.pl) |
14:18.41 | *** join/#maemo-ssu arcean_ (~arcean@aacs56.neoplus.adsl.tpnet.pl) |
14:19.38 | *** join/#maemo-ssu zogg_ (~zoggrules@bzq-109-64-160-7.red.bezeqint.net) |
14:23.35 | *** join/#maemo-ssu arcean (~arcean@aaeo165.neoplus.adsl.tpnet.pl) |
14:53.33 | *** join/#maemo-ssu Martix (~martix@4.177.broadband3.iol.cz) |
15:13.56 | gregoa | luf: zlib1g has a shlibs file that says: "libz 1 zlib1g (>= 1:1.2.1)" and that gets inserted in your libxml2 builg via the "${shlib:Depends}" in debian/control. if you need 1.2.3, then just adding "zlib1g (>= 1:1.2.3)" to the Depends line in debian/control should be enough. (or bumping the shlibs in zlib source package and rebuild and build libxml2 against that one) |
15:30.45 | *** join/#maemo-ssu arcean_ (~arcean@aaeo165.neoplus.adsl.tpnet.pl) |
15:38.05 | freemangordon | luf: https://gitorious.org/community-ssu/libpng |
15:53.08 | luf | freemangordon: coool |
15:54.08 | luf | freemangordon: We need to revert commits with the NEON opts I added. |
15:54.26 | freemangordon | luf: I am testing now with test suite on png website, so far so good |
15:54.37 | freemangordon | luf: no, my changes are on top of those |
15:55.23 | freemangordon | luf: BTW I don;t guarantee there won;t be some buffer overflow, please check aligment part :D |
15:55.43 | luf | Ahhh I see. |
15:56.47 | freemangordon | luf: well, it should be ok, but i'll prefer another one to review it |
15:56.55 | freemangordon | *i'd |
15:59.43 | luf | freemangordon, merlin1991: what about thle libxml2 merge request? I think we agreed it's bug and should be fixed. |
16:00.10 | freemangordon | I am ok |
16:00.24 | freemangordon | but only merlin1991 can merge |
16:00.47 | luf | ok. |
16:02.28 | freemangordon | luf: BTW libpng neon patch can be used without alignment part, but performance is worse |
16:03.20 | freemangordon | now it scores 2.220, unaligned ~2.400 |
16:03.35 | luf | Nice. |
16:14.02 | luf | freemangordon: png_read_filter_row: unsigned int bpp = (png_ptr->pixel_depth + 7) >> 3; - what does it do? |
16:14.21 | freemangordon | calculates bytes per pixel |
16:14.32 | freemangordon | but htis is original code |
16:14.41 | luf | I see. |
16:16.15 | luf | freemangordon: did you add prev_row to the struct? |
16:16.25 | freemangordon | no |
16:16.34 | freemangordon | I didn't touch ABI |
16:16.35 | luf | I see stupid question ... |
16:21.05 | luf | freemangordon: just one thing. 15 upstream has png_ptr->old_prev_row_size = row_bytes + 64; below row_buf setting. I see you keep it above ... |
16:22.38 | luf | s/old_prev_row_size/old_big_row_buf_size/ |
16:28.55 | freemangordon | luf: lemme check |
16:29.22 | luf | Also I don't think it's good to set old_prev_row_size prior to prev_row allocation. |
16:29.41 | luf | I think also prev_row should be 16-bit aligned. |
16:30.07 | luf | Based on libpng15 code: png_ptr->row_buf = temp - extra - 1/*filter byte*/; png_ptr->prev_row = temp - extra - 1/*filter byte*/; |
16:30.11 | freemangordon | luf: your first wuestio: it does not really matter which one will be set first |
16:30.20 | freemangordon | *question |
16:31.11 | freemangordon | luf: prev_row alignment is done in real-time, that was the hack I was talking aboot |
16:31.34 | freemangordon | upstream 15 has an additional struct member to keep unaligned pointer |
16:31.41 | freemangordon | we don't :) |
16:32.20 | freemangordon | so I keep unaligned pointer and calculate/pass aligned when needed |
16:32.33 | freemangordon | a bit of overhead, but no option there aiui |
16:33.04 | freemangordon | not that I like that much :D |
16:33.18 | luf | So row_bytes is always 16-bit aligned? |
16:33.27 | freemangordon | no |
16:33.51 | freemangordon | aiui we alloc row_bytes+64 |
16:34.29 | freemangordon | that's called big_row_buf or smtg |
16:34.54 | freemangordon | row pointer is aligned within this buffer |
16:35.09 | freemangordon | so we have padding before and after the real row |
16:35.51 | freemangordon | I am not sure i fully understand why this padding is needed, that's why i wanted a code review |
16:38.08 | freemangordon | valgrind reports no bad memory access, no memleaks :) |
16:38.54 | *** join/#maemo-ssu dhbiker (~dhbiker@APN-123-51-72-gprs.simobil.net) |
16:41.47 | luf | what does png_memset_check do? |
16:43.18 | freemangordon | NFC, but I gues it does memset :D |
16:45.05 | luf | So what's the reason to align16 in png_mem_check? |
16:45.27 | luf | I think it's a bug. |
16:45.29 | freemangordon | luf: which line? |
16:46.29 | luf | 3179 |
16:46.42 | freemangordon | luf: png_memset_check(png_ptr, align16(png_ptr->prev_row), 0, |
16:46.42 | freemangordon | png_ptr->rowbytes + 1); |
16:46.46 | freemangordon | luf: that code? |
16:47.08 | luf | yes. |
16:47.22 | freemangordon | no, it is ok, we don't need the whole buffer zeroed |
16:47.30 | freemangordon | only the part that is used |
16:48.07 | luf | Yeah allocated is rowbytes+64 so it doesn't overflow. |
16:48.26 | freemangordon | png_ptr->rowbytes is the "real" row bytesize, not the "big buffer" size |
16:50.19 | luf | I don't see why the buffer around is needed. But I see no problem in you code. |
16:50.56 | luf | Just one more question. Why you changed if ((png_uint_32)row_bytes + 1 > (png_uint_32)65536L) ? |
16:51.58 | luf | I see no reason for such change. |
16:57.49 | freemangordon | luf: aiui I think the padding is just in case :D |
16:58.28 | freemangordon | luf: which line is that change? |
16:59.44 | freemangordon | luf: BTW 64 bytes is a bit of overkill, upstream 1.5 use 48 |
17:01.35 | freemangordon | luf: re why i changed ^^^ - aiui this code becomes redundant, TBH I don;t see a reason for it to be there even in not-patched version |
17:01.46 | freemangordon | *non-patched |
17:09.04 | freemangordon | luf: results for http://en.wikipedia.org/wiki/File:PNG_transparency_demonstration_2.png - http://pastebin.com/UYvdgcsb |
17:09.14 | freemangordon | impressive |
17:11.16 | kerio | heh |
17:11.28 | kerio | do the test on the true transparent one |
17:16.30 | freemangordon | kerio: wanna do it yourself? i'll upload the binaries |
17:16.59 | kerio | is the thumby deb for libpng somewhere? |
17:17.08 | freemangordon | just a minute |
17:18.52 | freemangordon | kerio: http://merlin1991.at/~freemangordon/libpng/ |
17:19.03 | freemangordon | .deb is thumb-compiled, despite the name |
17:22.01 | freemangordon | kerio: transparency shhould not affect decoding time |
17:23.25 | luf | freemangordon: You're right. That code has to write some drunken dev. As "Row has too many bytes to allocate in memory." is behind some memory (re)allocation :D |
17:24.37 | luf | freemangordon: Frommy POV the align part is ok. |
17:26.13 | luf | merlin1991: ping |
17:28.17 | luf | freemangordon: more interesting should be ARM againist NEON with same thumb/non-thumb and gcc. |
17:28.49 | luf | However 30% speed up is nice ;) |
17:29.13 | kerio | freemangordon: it's ok if pngtest.png and pngout.png are different, right? |
17:29.39 | luf | kerio: it should be ... due to zlib etc. |
17:31.19 | luf | kerio: let's try rename the pngout.png and test it on that new ;) |
17:32.44 | freemangordon | luf: what do you mean ARM against NEON? |
17:32.54 | luf | http://pastebin.com/UYvdgcsb |
17:33.16 | luf | I see ARM, gcc 4.2.1 vs NEON+THUMB, gcc 4.7.2: |
17:33.32 | kerio | CPU time used = 31.420 seconds (decoding 2.370, encoding 29.040 , other 0.010 seconds) |
17:34.06 | freemangordon | luf: THUMB makes almost no difference |
17:34.15 | kerio | hold on, i already had a thumby one |
17:34.16 | freemangordon | see encoding times |
17:34.27 | freemangordon | kerio: yes |
17:34.40 | luf | freemangordon: and different gcc also? |
17:34.48 | freemangordon | looks like |
17:34.55 | luf | ok. |
17:35.08 | kerio | CPU time used = 30.800 seconds (decoding 1.580, encoding 29.170 , other 0.050 seconds) |
17:35.14 | kerio | holy balls |
17:35.23 | kerio | freemangordon: what's this new libpng? |
17:35.26 | kerio | NEONized one? |
17:35.43 | freemangordon | luf: why don't you test it, pngtest binary is here http://merlin1991.at/~freemangordon/libpng/ |
17:35.46 | freemangordon | kerio: yes |
17:35.52 | luf | kerio: yes NEONized + memory alignment |
17:35.54 | kerio | thumbify it, too! |
17:36.01 | freemangordon | it is thumbified |
17:36.21 | luf | freemangordon: I'm still running non-thumb version. |
17:36.46 | freemangordon | luf just build the .deb dammit :P |
17:36.56 | freemangordon | luf: ok, i'll build it for you |
17:37.03 | freemangordon | hold on a minute |
17:37.07 | luf | BTW I see several reboots on my dev N900 yesterday. I installed tklocks and others from cssu-devel. |
17:37.18 | luf | freemangordon: don't do it. |
17:37.23 | luf | I'm leaving now ... |
17:37.31 | freemangordon | luf: reboots? |
17:37.46 | freemangordon | do you have sp-rich-core installed? |
17:38.10 | freemangordon | luf: before you leave |
17:38.20 | luf | freemangordon: I have to leave now. I'll continue later or tomorrow. |
17:38.23 | freemangordon | where is neon-patched zlib? |
17:38.38 | freemangordon | ok |
17:39.01 | luf | No I don't have. |
17:39.09 | freemangordon | ok |
17:39.10 | luf | You know where is the source. |
17:39.30 | luf | http://merlin1991.at/~luf |
17:39.40 | luf | http://merlin1991.at/~luf/zlib/source/ |
17:40.42 | luf | BTW today's benchmarks showed me less improvement zlib ... |
17:56.16 | *** join/#maemo-ssu dafox (~dafox@ip51cc571d.speed.planet.nl) |
17:57.02 | kerio | freemangordon: btw, /dev/shm is neat to do tests |
18:11.37 | *** join/#maemo-ssu MrPingu (~MrPingute@86.92.226.97) |
18:52.34 | *** join/#maemo-ssu Martix (~martix@4.177.broadband3.iol.cz) |
18:58.50 | *** join/#maemo-ssu Martix (~martix@4.177.broadband3.iol.cz) |
19:28.25 | *** join/#maemo-ssu arcean (~arcean@aaeo165.neoplus.adsl.tpnet.pl) |
20:08.30 | *** join/#maemo-ssu arcean (~arcean@aaeo165.neoplus.adsl.tpnet.pl) |
20:19.37 | *** join/#maemo-ssu MrPingu (~MrPingute@86.92.226.97) |
20:19.38 | *** join/#maemo-ssu andre__ (~andre@wikimedia/aklapper) |
20:25.14 | *** join/#maemo-ssu Skry (~skry@77.109.215.222) |
20:38.15 | *** join/#maemo-ssu Skry (~skry@77.109.215.222) |
21:37.07 | *** join/#maemo-ssu joshgillies (~josh@58-6-102-109.dyn.iinet.net.au) |
21:59.41 | *** join/#maemo-ssu Oposum (~ruin@ruin.mcn.hs-mittweida.de) |
22:37.17 | *** join/#maemo-ssu nox- (noident@freebsd/developer/nox) |
22:58.00 | *** join/#maemo-ssu dafox (~dafox@ip51cc571d.speed.planet.nl) |
23:30.27 | *** join/#maemo-ssu jonwil (~jonwil@27-33-137-199.static.tpgi.com.au) |