08:33:40 <bero> #startmeeting
08:33:40 <linarobot> Meeting started Wed Nov  4 08:33:40 2015 UTC.  The chair is bero. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:33:40 <linarobot> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:33:43 <bero> who is here?
08:33:59 <zhizhoutian_> me is here
08:34:03 * zhizhoutian_ is here
08:34:33 <yspan> hi
08:34:55 <Qian__> hi al
08:35:02 <Qian__> hi everyone
08:36:13 <bero> Qian__: do you want to go first?
08:36:33 <Qian__> Qian Chen <qian.chen@linaro.org>
08:36:34 <Qian__> What did you do?
08:36:34 <Qian__> Compile and run-test  SPEC CPU2k6 on x86 successfully
08:36:35 <Qian__> Cross-compile CPU2k6 successfully using the toolchain(5.1) provided by Bogden and using the toolchain bero compiled with the error caused by sed:
08:36:38 <Qian__> “sed: -e expression #1, char 2449: unknown option to `s'”,
08:36:39 <Qian__> the root case should not be caused by toolchain.
08:36:41 <Qian__> Download the CPU2K6 on Nexus5 for test, have some problems when runspec.
08:36:43 <Qian__> What are you going to do?
08:36:45 <Qian__> test perl on Nexus5 if it can work or not?, it’s very important for the next steps.
08:36:47 <Qian__> try runspec and fixed the errors
08:36:49 <Qian__> find why “sed: -e expression #1, char 2449: unknown option to `s'” happens
08:36:51 <Qian__> What problems do you have?
08:36:53 <Qian__> i don’t know how the SPEC tools(under tools/src/, like perl, bzip2...) were been built, use the bin now(ask bogden for help)
08:37:47 <bero> The sed thing is probably caused by a variable containing a special character or something
08:37:54 <Qian__> it seems i will download per-version-cross-xxx and cross-compile it ,that will work on arm
08:38:26 <bero> sed -e "s/a/$b/" can do odd thing if b contains a / etc.
08:38:30 <Qian__> bero, yes, i don't care it now, go on my work now
08:39:15 <bero> ok, sunao: want to go next?
08:39:58 <sunao> What did you do?
08:40:00 <sunao> 1. chage the parameter of jemalloc, test on the emulator, but it is hard to see the difference.
08:40:01 <sunao> The default value of the parameters are very conservative already.
08:40:03 <sunao> what are you going to do?
08:40:04 <sunao> 1. continue the investiagate on the jemalloc.
08:40:06 <sunao> 2. Investigate how android cache fonts and rendered graphics.
08:40:07 <sunao> 
08:40:09 <sunao> what problems do you have?
08:40:10 <sunao> 1.  when we change the parameter of jemalloc, it is hard to see the impact.
08:41:06 <bero> yes, sometimes the mallocs don't have much of an impact -- it's perfectly ok if the result of the investigation is that it doesn't really matter either way
08:41:17 <sunao> Do we need to investigate the dlmalloc? seems jemalloc is the default malloc allocator
08:42:09 <bero> yes, there's a possibility jemalloc works faster on higher end devices but dlmalloc uses up less memory unnecessarily
08:42:09 <bero> that's what one article I've read on mallocs implied
08:42:09 <sunao> only if you defined the MALLOC_IMPL := dlmalloc, the system will use dlmalloc
08:42:22 <bero> if that's true, we may want to use MALLOC_IMPL := dlmalloc on low memory devices
08:42:35 <sunao> OK, I will test the dlmalloc
08:44:13 <bero> yspan: do you want to go next?
08:44:41 <yspan> What did you do?
08:44:41 <yspan> Fixed a bug. Booting to Android home screen OK! Use pngtest to test 3 pgn files OK! attend ARM Tech Day in MediaTek.
08:44:54 <yspan> What are you going to do?
08:44:54 <yspan> Generate patch. Measure improvement.
08:44:54 <yspan> What problems do you have?
08:44:54 <yspan> none.
08:45:24 <bero> good... anything interesting from ARM Tech Day?
08:46:08 <yspan> I attend low power section EAS, IPA :p
08:46:34 <bero> nice, sounds like fun stuff to do
08:46:53 <bero> liuyq: do you want to go next?
08:47:07 <liuyq> What did you do?
08:47:07 <liuyq> submitted patches for Bug 795
08:47:07 <liuyq> Investigated about the utilization of �ifeq ($(TARGET_ARCH),arm64)� in the android source
08:47:07 <liuyq> Investigated about the LCR build errors.
08:47:07 <liuyq> What are you going to do?
08:47:07 <liuyq> submit fix for LCR build errors, and trigger build
08:47:11 <liuyq> check jobs on LAVA and close the Bug 795
08:47:13 <liuyq> finish check about the utilization of �ifeq ($(TARGET_ARCH),arm64)�
08:47:15 <liuyq> What problems do you have?
08:47:17 <liuyq> none.
08:47:53 <liuyq> bero, there are some changes on libstagefright  with android-6.0.0_r26 tag
08:47:53 <bero> good... zhizhoutian_: next?
08:48:07 * zhizhoutian_ What did you do? 1. Investigate Juno toolchain issues with C-Reduce tools a)resolved llvm’s low version issue: install llvm with source code type, not apt-get install. b) another issue: errors while execute ./configure checking can compile with and link with LLVM(engine)... no configure: error: in
08:48:07 * zhizhoutian_ `/home/likewise-open/SPREADTRUM/zhizhou.tian/workspace/creduce': configure: error: cannot compile and link test program with selected LLVM See `config.log' for more details /me What are you going to do? 1. Investigate Juno toolchain issues with C-Reduce tools /me What problems do you have? none
08:48:19 <zhizhoutian_> What did you do?
08:48:20 <zhizhoutian_> 1. Investigate Juno toolchain issues with C-Reduce tools
08:48:20 <zhizhoutian_> a)resolved llvm’s low version issue: install llvm with source code type, not apt-get install.
08:48:20 <zhizhoutian_> b) another issue:
08:48:20 <zhizhoutian_> errors while execute ./configure
08:48:20 <zhizhoutian_> checking can compile with and link with LLVM(engine)... no
08:48:20 <zhizhoutian_> configure: error: in `/home/likewise-open/SPREADTRUM/zhizhou.tian/workspace/creduce':
08:48:21 <zhizhoutian_> configure: error: cannot compile and link test program with selected LLVM
08:48:21 <zhizhoutian_> See `config.log' for more details
08:48:22 <zhizhoutian_> What are you going to do?
08:48:22 <zhizhoutian_> 1. Investigate Juno toolchain issues with C-Reduce tools
08:48:23 <zhizhoutian_> What problems do you have?
08:48:23 <zhizhoutian_> none
08:48:44 <bero> hmm... missing development headers or missing -I flag for LLVM headers in a nonstandard location maybe?
08:49:07 <bero> luther: want to go next?
08:49:33 <zhizhoutian_> I just executed ./configure
08:50:29 <bero> yes, I think it'll only look for llvm headers in /usr if you don't tell it otherwise. Not sure where you installed your newer version of llvm. If it's any place other than /usr, you probably have to tell creduce where it can find things
08:51:13 <bero> Not sure if luther is around today... so in the mean time here's mine:
08:51:17 <bero> What did you do?
08:51:17 <bero> Ran test builds to see the effect of hardfloat toggles and FPU adjustments
08:51:17 <bero> Updated to android-6.0.0_r26 tag
08:51:17 <bero> Added TARGET_OPTIMIZE_FOR_SIZE switch to BoardConfig.mk (for now, changes default optimization level to -Os and adds -Oz for clang)
08:51:17 <bero> Started analyzing the “Nexus [457] camera doesn’t work if OS was built with gcc 5.2” bug
08:51:17 <bero> Sent fix for libc++ overly pessimistic assumptions about gcc upstream
08:51:17 <bero> What are you going to do?
08:51:18 <bero> Analyze effects of -Os and -Oz on code size/memory use with both gcc and clang
08:51:18 <bero> Hopefully fix/work around Nexus [457] camera crash
08:51:19 <bero> Backport fix for libc++ gcc assumptions to AOSP
08:51:19 <bero> What problems do you have?
08:51:20 <bero> none
08:53:14 <bero> ok, luther doesn't seem to be here and xavierhsu isn't here... did I forget anyone else?
08:54:44 <bero> probably not... so have a great day and let's talk tomorrow at latest
08:54:47 <bero> #endmeeting