08:33:07 <bero> #startmeeting
08:33:07 <linarobot> Meeting started Tue Nov  3 08:33:07 2015 UTC.  The chair is bero. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:33:07 <linarobot> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:33:17 <bero> hi
08:33:53 <bero> so some good news first - looks like we'll get a few more people to help with memory optimizations really soon now - their accounts have already been created, so it's getting really close
08:33:55 <Qian_> hi all
08:34:18 <bero> no other big news -- let's do the status round
08:34:22 <yspan> cool
08:34:30 <bero> liuyq: do you want to start?
08:35:22 <liuyq> bero, sorry, could you ask somebody else for next, I am preparing the document
08:35:26 <bero> sure
08:35:33 <bero> yspan: do you want to start?
08:35:45 <yspan> yes
08:36:00 <yspan> What did you do?
08:36:00 <yspan> Rebase to LCR-15.10(L), and still hunting a segfault, fix one, get another one.. use logcat to output some debug messages
08:36:09 <yspan> What are you going to do?
08:36:09 <yspan> This should be the last one :p. After fixing bug, I will test it with pngtest and measure the memory usage of libpng and libpng-enc
08:36:15 <yspan> What problems do you have?
08:36:15 <yspan> One of a segfault: I thought __func__ is only function name, like pngwrite_info. But actually it contains the whole function signature: void pngwrite_info(a, b, c), like __PRETTY_FUNCTION__ does. So I switch to use __FUNCTION__ although it is old style.
08:36:56 <bero> Oh, one more thing for everyone: You all may have seen tasks for 15.11 showing up -- that's what Jakub and I were planning on Friday, let me know if there's anything you can't/don't want to do that has been assigned to you.
08:37:32 <yspan> that's all
08:37:40 <luther> bero,
08:37:48 <bero> yspan: good... are you using __FUNCTION__ to implement the dlopen/dlsym/dlclose cycle in a generic way?
08:37:58 <yspan> bero, yes
08:38:13 <bero> yspan: nice
08:38:20 <bero> luther: what's up? do you want to go next?
08:39:09 <luther> ok
08:39:17 * luther What did you do?
08:39:17 <luther> 1. Looked C++ allocator template
08:39:18 <luther> 2. Talked with my colleague who joined lead project yesterday,
08:39:18 <luther> he is expert on zram swapfs, he would like the investigation on zramfs
08:39:18 * luther What are you going to do?
08:39:18 <luther> 1. Go on look at C++ allocator template
08:39:20 * luther What problems do you have?
08:39:22 <luther> 1. None till now
08:39:24 <Qian_> i have another talk now, please call me whem my tuen
08:39:26 <luther> i just want to talk about zram fs
08:40:22 <bero> yes, that would be good...
08:40:29 * liuyq is ready to report
08:40:40 <luther> bero, thanks:-)
08:40:52 <bero> Qian_: ok, you'll go last so you have a chance to finish the other meeting
08:40:55 <luther> bero, his name is xing.yu or yu.xing
08:42:47 <bero> luther: thanks, I've assigned the task to him
08:44:04 <luther> bero, thanks :-)
08:45:54 <bero> luther: do you want to pick another task instead, or do you have enough work lined up already?
08:46:30 <Qian_> What did you do?
08:46:30 <Qian_> Investigate about SPEC CPU2K6
08:46:30 <Qian_> Compile the src on Ubuntu-14.04, meet some errors and fix(or by workaround)
08:46:32 <Qian_> Run the Samples for having impression.
08:46:33 <Qian_> What are you going to do?
08:46:36 <Qian_> the following works must be done when we transplant it to arm.
08:46:37 <Qian_> Cross-compile Perl and other open source libs on ARM,  especially Perl, since the core of CPU2k6 is written by Perl.
08:46:40 <Qian_> Change the framework of CPU2K6 because it will compile binaries when execute runspec on x86
08:46:40 <luther> bero, i want to study c++ and java code for a while, is that ok?
08:46:41 <Qian_> I will spend much time to investigate how SPEC uses to test cases and how SPEC collects the datas
08:46:44 <Qian_> What problems do you have?
08:46:45 <Qian_> How the SPEC uses the test cases and how SPEC collects the output datas.
08:46:47 <Qian_> There is a good chance we can’t succeed even spend much time since SPEC needs native compile and runs binaries generated during the build
08:47:12 <bero> luther: sure
08:47:22 <luther> bero, thanks:-)
08:48:16 <bero> Qian_: we had gcc built for AOSP a couple of years back, maybe we need to dig that up again... Any idea what else will be needed?
08:50:58 <bero> probably Qian_ is busy right now, so let's continue with liuyq
08:51:18 <liuyq> YongQin Liu <yongqin.liu@linaro.org>
08:51:19 <liuyq> What did you do?
08:51:19 <liuyq> Investigated on Bug 795 - Failing multimedia codecs on differences between ARMv7 and ARMv8
08:51:19 <liuyq> Found the md5sum difference is caused by the the c-only implementation and asm implementation
08:51:19 <liuyq> Found some ARM optimizations are not used for the 32bit binary on 64bit platforms.
08:51:20 <liuyq> Found the size of 32bit libc.a is different between flo and flounder platforms
08:51:22 <liuyq> 
08:51:24 <liuyq> What are you going to do?
08:51:26 <liuyq> update the test implementation for linaro-android-userspace-test, will use the 32bit asm implementations
08:51:29 <liuyq> find out the places that ASM optimizations are not applied for 32bit binary on 64bit platforms, and fix them, will open a ticket for this.
08:51:32 <liuyq> What problems do you have?
08:51:36 <liuyq> none
08:52:14 <liuyq> bero, now I am update the for framework/av repository and linaro-android-userspace-test  repository for bug 795 only
08:53:05 <bero> good... sunao, do you want to go next?
08:53:11 <sunao> What did you do?
08:53:13 <sunao> 1. Investigate the jemalloc memory allocator, change the parameters and test on emulator.
08:53:14 <sunao> what are you going to do?
08:53:15 <sunao> 1. Investigate the jemalloc memory allocator.
08:53:17 <sunao> 2. Investigate how android cache fonts and rendered graphics.
08:53:19 <sunao> what problems do you have?
08:53:20 <liuyq> bero, will check the ASM implementation in other places after I finished the update for bug 795
08:53:20 <sunao> 1.None till now.
08:54:20 <bero> good... zhizhoutian_, do you want to go next?
08:54:31 <zhizhoutian_> ok
08:54:39 <Qian__> bero: I have asked bogden for help, and he gives me some help
08:54:56 <Qian__> bero:i will try as he tells next
08:55:31 <bero> Qian__: I've googled it and found a posting from Karim Yaghmour about it, so chances are he has built it for AOSP before. I'll ask him what he did, maybe he can help
08:57:19 <zhizhoutian_> What did you do?
08:57:19 <zhizhoutian_> check why VFAT kernel tests all fail on Nexus9
08:57:19 <zhizhoutian_> https://projects.linaro.org/browse/LMG-896
08:57:19 <zhizhoutian_> this command caused this issue - mkfs.vfat -F 12 -i 0x12347900 /dev/block/loop7
08:57:19 <zhizhoutian_> mkfs.vfat: failed whilst writing reserved sector
08:57:19 <zhizhoutian_> Investigate Juno toolchain issues with C-Reduce tools
08:57:19 <zhizhoutian_> C-Reduce can not be installed with low llvm version.
08:57:20 <zhizhoutian_> I have tried all c-reduce branches(version), the lowest c-reduce version need above llvm-3.0
08:57:20 <zhizhoutian_> But llvm-2.8 is latest in Ubuntu.
08:57:21 <zhizhoutian_> What are you going to do?
08:57:21 <zhizhoutian_> What problems do you have?
08:57:33 <Qian__> bero: yes, he said he has built before without documents, i will ask him when meet problems
08:58:06 <zhizhoutian_> What did you do?
08:58:06 <zhizhoutian_> 1. check why VFAT kernel tests all fail on Nexus9
08:58:06 <zhizhoutian_> https://projects.linaro.org/browse/LMG-896
08:58:06 <zhizhoutian_> this command caused this issue - mkfs.vfat -F 12 -i 0x12347900 /dev/block/loop7
08:58:06 <zhizhoutian_> mkfs.vfat: failed whilst writing reserved sector
08:58:07 <zhizhoutian_> 2. Investigate Juno toolchain issues with C-Reduce tools
08:58:07 <zhizhoutian_> C-Reduce can not be installed with low llvm version.
08:58:08 <zhizhoutian_> I have tried all c-reduce branches(version), the lowest c-reduce version need above llvm-3.0
08:58:08 <zhizhoutian_> But llvm-2.8 is latest in Ubuntu.
08:58:09 <zhizhoutian_> What are you going to do?
08:58:09 <zhizhoutian_> What problems do you have?
08:58:17 <zhizhoutian_> this should be better.
09:00:37 <bero> they should really update their llvm... using anything before 3.7 is a pretty bad idea these days given how far it has come lately
09:00:48 <bero> but I guess you can always build 3.7 from source
09:01:06 <bero> did I forget anyone?
09:02:13 <zhizhoutian_> i see, you mean i can install llvm with source.
09:04:05 <bero> yes
09:04:59 <zhizhoutian_> i see, i will try it. thansk
09:07:02 <bero> ok, I think we have everyone...
09:07:05 <bero> did I forget anyone?
09:07:14 <bero> or does anyone want to bring up anything else?
09:08:17 <bero> apparently not -- so let's get back to the code... See you all around!
09:08:19 <bero> #endmeeting