08:34:44 <bero2> #startmeeting
08:34:44 <linarobot> Meeting started Tue Dec 15 08:34:44 2015 UTC.  The chair is bero2. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:34:44 <linarobot> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:35:12 <bero2> #topic sprint
08:36:22 <bero2> Looks like we're getting too many "can't attend" replies from members
08:36:58 <bero2> If anyone hasn't asked yet, please do so today
08:37:17 <bero2> No other news yet...
08:37:32 <bero2> #topic updates
08:38:04 <bero2> sunao: do you want to go first?
08:38:11 <sunao> What did you do?
08:38:12 <sunao> 1.   static linked (libc_nomalloc + libptmalloc )test process works normall.  But the whole system can't bootup. too many process crashed. still trying.
08:38:14 <sunao> what are you going to do?
08:38:15 <sunao> 1. have a linaro workshop tomorrow, maybe unaviable for tomorrow's daily meeting.
08:38:17 <sunao> 2.  continue porting ptmalloc2.
08:38:18 <sunao> what problems do you have?.
08:38:20 <sunao> None.
08:38:46 <bero2> Ok
08:39:15 <bero2> Qian__: next?
08:39:22 <Qian__> What did you do?
08:39:22 <Qian__> Test bionic performance, and found out the latest modify of string_copy.S
08:39:22 <Qian__> may cause strcpy and strcat regression on mx5, this patch has been merged into andorid-M
08:39:24 <Qian__> see https://bugs.linaro.org/show_bug.cgi?id=1944
08:39:26 <Qian__> Investigate lmg-966 for saving memory of specific process, ProgressList.java and ActiviytyManagerService.java seem good now
08:39:28 <Qian__> What are you going to do?
08:39:30 <Qian__> Go on investigate how it works
08:39:32 <Qian__> Test bionic patch of string_copy.S on PRO 5
08:39:34 <Qian__> What problems do you have?
08:39:36 <Qian__> Shall we test string_copy.S on other platforms
08:39:50 <xruxa> sunao, looks like you will not have time to finish the LMG-933 (font caching) work this month. bero2 I will move it to next sprint?
08:41:19 <Qian__> the 64 bit strcpy has a regression about 5%, and the 64 bit strcat has a regression about 10%
08:41:23 <bero2> Hmm, that may be worth looking at,  we may have ported a reimplementation there that isn't better than the new upstream code
08:41:35 <xruxa> Qian__, are we done with the LMG-912 (statuc kunkage, copied code) work?
08:42:03 <Qian__> xruxa: I think bero2 has checked it yet
08:42:03 <bero2> sunao: do you think you'll have time for 933?
08:42:36 <sunao> yes, I will look at LMG933 this month
08:42:50 <Qian__> bero2: I have tested on android-M, the same result too
08:43:12 <Qian__> bero2: I have tested on pro 5, it has the same problem
08:43:42 <xruxa> sunao, ok, so let me keep it there till Thursday at least :)
08:43:45 <Qian__> next, i will test on pro 5 without the patch to have a try
08:44:40 <xruxa> #action bero2 to check findings on LMG-914 (static linkage, copied code) analysis
08:44:52 <Qian__> Directory:  android/bionic
08:44:52 <Qian__> libc: arm64: add optimized strcpy implementation
08:44:52 <Qian__> 
08:44:52 <Qian__> This optimization is extracted from cortex-strings and bionic-ized.
08:44:54 <Qian__> stringbench result:
08:44:56 <Qian__> |---------------+--------+--------+-------------+
08:44:56 * xruxa wonders if he can make actions if not chair
08:44:58 <Qian__> |   item        |  aosp  | cortex | cortex gain |
08:45:00 <Qian__> |---------------+--------+--------+-------------+
08:45:02 <Qian__> | strcpy 15000b | 5.23 s |  5.20s |       0.46% |
08:45:04 <bero2> List on 912 looks good - now we need to start getting rid of static linkage and friends
08:45:07 <Qian__> |---------------+--------+--------+-------------+
08:45:07 <Qian__> | strcpy 15b    | 3.64 s |  3.28s |       9.91% |
08:45:08 <Qian__> |---------------+--------+--------+-------------+
08:45:10 <Qian__> 
08:45:16 <Qian__> Change-Id: If04ebf677f8f41b4555b00709d41cc09b11cc927
08:45:16 <Qian__> Signed-off-by: Yingshiuan Pan <yingshiuan.pan@linaro.org>
08:45:22 <bero2> #chair xruxa
08:45:22 <linarobot> Current chairs: bero2 xruxa
08:45:52 <xruxa> #info made new JIRA item for Qian__ to work on the Bionic perf regression investigation - https://projects.linaro.org/browse/LMG-1004
08:46:35 <bero2> xavierhsu: next?
08:47:48 <xavierhsu> ok
08:47:54 <xavierhsu> What did you do?
08:47:55 <xavierhsu> I modify the source code(jpegenc_stub.c).
08:48:10 <xavierhsu> What are you going to do?
08:48:10 <xavierhsu> I only add the "dlopen" in function (jpeg_CreateCompress) and the "dlclose" in function (jpeg_destroy_compress).
08:48:37 <xavierhsu> What problems do you have?
08:48:37 <xavierhsu> Testing now.
08:49:07 <bero2> OK,  I hope I'll have some time to look at it today too
08:49:30 <bero2> liuyq: next?
08:49:48 <xavierhsu> bero, thanks.
08:50:42 <liuyq> What did you do?
08:50:42 <liuyq> Investigated about the difference on the size of libc.a between Nexus7 and Nexus9(Nexus9 uses dlmalloc)
08:50:42 <liuyq> Checked jobs for Hikey On LAVA(still no cts part2 jobs for member build, and part1 and part2 jobs for hikey reference build)
08:50:42 <liuyq> Started to prepare for 15.12 release notes
08:50:44 <liuyq> What are you going to do?
08:50:46 <liuyq> Finish the draft version of 15.12 release notes
08:50:48 <liuyq> Finish the comparison of bootchart between Lollipop and Marshmallow build
08:50:50 <liuyq> What problems do you have?
08:50:52 <liuyq> None
08:51:14 <liuyq> bero2, mainly release thing this week I think
08:51:23 <bero2> I looked at that HSPACE thing btw,  it has no effect at all
08:51:45 <bero2> Probably had some effect during the K beta phase or so
08:52:20 <bero2> eric___: next?
08:52:33 <eric____> What did you do?
08:52:33 <eric____> Investigate GYP, on how to generate makefile; It can work for single lib now, but found there’s no Android.mk, and there’re some issues when running all.gyp, try to fix it;
08:52:33 <eric____> Investigate Chrome, on how to replace libjpeg, libpng, and zlib with system libraries;
08:52:33 <eric____> What are you going to do?
08:52:33 <eric____> Try to build with system lib with GYP;
08:52:33 <eric____> What problems do you have?
08:52:34 <eric____> GYP doesn’t generate Android.mk;
08:52:34 <eric____> there’re some issues when running all.gyp, try to find a way to fix it;
08:52:52 <liuyq> bero2, thanks, good to know that about HSPACE
08:54:06 <eric____> I'll focus on build with gyp, not on generate makefile
08:54:29 <xruxa> bero2, eric____, how about LMG0977 (libwebp) ?
08:54:38 <bero2> eric___: AFAIK gyp isn't supposed to generate Android.mk files.  I think it generates ninja files that can be used to build the APK
08:54:54 <eric____> It seems that only need modify "all.gyp"
08:55:06 <bero2> Then you just drop the APK into the AOSP tree
08:55:14 <eric____> LMG0977 , is there any problem?
08:55:35 <eric____> You confirmed that this issue was changed by upstream
08:55:38 <bero2> xruxa: 977 was done upstream before we could get to it
08:55:58 <xruxa> bero2, ok, with that confirmations I will close it. Thanks.
08:55:59 <bero2> xruxa: so it must be cancelled
08:57:20 <bero2> Did I forget anyone who is here?
08:58:11 <eric____> bero2: but I didn't the ninja file after running it
08:58:24 <xruxa> bero2, I wonder how the KSM work is going, is the engineer still with us?
08:58:47 <liuyq> bero2, wuhai and wangjian are back to Shenzhen today, and they will have some days for leave, so will not show up for this meeting this week I think
08:59:40 <bero2> There should be a build.ninja file,  but I haven't tried to do this myself,  so there may be something wrong
08:59:50 <xruxa> bero2, https://projects.linaro.org/browse/LMG-963 system services optimizations - looks like no one is working on that? Or not worth doing?
09:00:07 <bero2> liuyq: ah, good to know
09:03:53 <xruxa> bero2, looks like no other engineers here - lets wrap the report round
09:04:38 <bero2> xruxa: I still think there's a few kbs to cut off in 963,  but it may be more effort than it's worth
09:04:48 <bero2> #endmeeting