08:32:59 <bero> #startmeeting
08:32:59 <linarobot`> Meeting started Tue Dec  1 08:32:59 2015 UTC.  The chair is bero. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:32:59 <linarobot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
08:33:04 <bero> #chair xruxa
08:33:04 <linarobot`> Current chairs: bero xruxa
08:33:18 <xavierhsu> Hi all
08:33:29 <yuxing> Hi all
08:33:43 <bero> xruxa: would you mind running the meeting today? I'm on mobile right now (but watching), will be back at the computer in 5min or so (the dogs took a bit longer outside than usual)
08:34:07 <xruxa> bero, ack, let me call the status round
08:34:13 <xruxa> #topic Status round
08:34:21 <xruxa> sunao, hello, please go first
08:34:26 <sunao> ok
08:34:31 <sunao> What did you do?
08:34:33 <sunao> 1. tuning the parameters of dlmalloc. test both (memory_alignment=8) and (memory_alignment=8 & default_trim_threshold=1MB), but both of them didn't save more memory than default dlmalloc. update the result to LMG-942. https://projects.linaro.org/browse/LMG-942
08:34:34 <sunao> the previous conclusion of change memory alignment can save 1~2MB is wrong. Sorry for that.
08:34:36 <sunao> what are you going to do?
08:34:39 <sunao> 1.  continue work on dlmalloc or investigate on ptmalloc?
08:34:40 <sunao> what problems do you have?.
08:34:41 <sunao> 1.  None.
08:35:01 <xruxa> sunao, does bero have some patches from you for dmalloc already?
08:35:27 <sunao> bero said he has submit the patch, but I have not checked it.
08:35:49 <xruxa> sunao, ack. Then I guess bero has the patch and will want you to look into ptmalloc
08:35:56 <sunao> OK
08:36:00 <bero2> yes
08:36:33 <xruxa> I will make item in JIRA for ptmalloc now
08:36:38 <xruxa> sunao, thanks.
08:36:39 <bero2> dlmalloc activation is in
08:36:44 <xruxa> Qian_, please go next
08:36:51 <bero2> tweaking not yet
08:36:58 <Qian_> What did you do?
08:36:59 <Qian_> 1. Test the bionic performance on Nexus5 and MX5, Both have a remarkable improvement.
08:36:59 <Qian_> nexus5 before
08:37:01 <Qian_> 10000000_*_memchr_finding_match_at_5000:_57.255937_seconds 10000000_*_memchr_not_finding_anything_in_15000_chars:_181.078651_seconds
08:37:03 <Qian_> after
08:37:05 <Qian_> 32Bit_10000000_memchr_finding_match_at_5000=24.10,seconds
08:37:06 <Qian_> 32Bit_10000000_memchr_not_finding_anything_in_15000_chars=81.31,seconds
08:37:09 <Qian_> About 2 times improvements
08:37:11 <Qian_> MX5 before
08:37:13 <Qian_> 10000000 * memchr not finding anything in 15000 chars: 334.668085 seconds
08:37:15 <Qian_> 10000000 * memchr finding match at 5000: 110.170890 seconds
08:37:16 <Qian_> after:
08:37:18 <Qian_> 10000000 * memchr not finding anything in 15000 chars: 87.666744 seconds
08:37:20 <Qian_> 10000000 * memchr finding match at 5000: 29.528417 seconds
08:37:23 <Qian_> About 4 times improvements
08:37:25 <Qian_> What are you going to do?
08:37:27 <Qian_> Continue to investigate LMG-914
08:37:29 <Qian_> What problems do you have?
08:37:31 <Qian_> none
08:38:08 <xruxa> Qian_, thanks, I have seen your comments in the JIRA item, looks like good progress.
08:38:20 <xruxa> luther, are you here or still busy with company work?
08:38:25 <Qian_> xruxa: thanks
08:39:37 <xruxa> guess not luther
08:39:43 <xruxa> xavierhsu, please go next
08:39:44 <luther> xruxa: i am here
08:39:47 <xavierhsu> ok
08:39:51 <luther> xruxa: sorry
08:40:15 <xruxa> luther, no problem, you still busy with company work?
08:40:19 <luther> xruxa: yes, still work on company work, so many news come
08:40:23 <xruxa> luther, ack
08:40:27 <xavierhsu> What did you do? 1. If I want to test original jpeg (Hikey), libjpeg-turbo (Modify by YongQin) or mine libjpeg-turbo that need to compile all source code of Android or it will happen some link error (display show black).  2. The build-in app (Gallery) can be used by libjpeg-turbo (modify by YongQin).  3. Other android apps can normally be used (they will call libjpeg-enc.so).
08:40:28 <xruxa> xavierhsu, please go next
08:40:42 <xavierhsu> What are you going to do?I dump the error message (using logcat) for finding the bug.
08:40:51 <xavierhsu> What problems do you have?Debug now.
08:41:15 <xruxa> xavierhsu, you say that you need to rebuild all android to test your libjpeg-turbo? that does not sound right. bero2 and otehrs - any suggestions?
08:41:52 <xruxa> #info luther still busy with company work
08:42:20 <Qian_> xavierhsu: xruxa, I think except the first time
08:42:34 <xavierhsu> xruxa, If I did not rebuild the source code, some code will happen some link error
08:43:00 <xruxa> let me give xavierhsu action to resolve this on IRC with you guys during today.
08:43:09 <xruxa> xavierhsu, thanks for the report, moving on
08:43:17 <xruxa> yspan, can you go next?
08:43:19 <Qian_> I have built on MX5 and PRO 5, Seems good
08:43:26 <yspan> sure
08:43:32 <yspan> What did you do?
08:43:33 <yspan> Fixing an error reported by bero. But have no idea where to put
08:43:33 <yspan> PRODUCT_PACKAGES += libpng-enc
08:43:33 <yspan> I grep-ed device/ and build/ but did not find where PRODUCT_PACKAGES is.
08:43:33 <yspan> What are you going to do?
08:43:33 <yspan> Add �PRODUCT_PACKAGES += libpng-enc� to generate libpng-enc.so.
08:43:35 <yspan> What problems do you have?
08:43:37 <yspan> Where is the suitable place for �PRODUCT_PACKAGES += libpng-enc"
08:43:52 <xruxa> #action xavierhsu to ask around and resolve his issues with deploying new libjpeg-turbo on his HiKey without the need to rebuild whole android every time
08:44:22 <yspan> I grep-ed device/ and build/ but did not find where PRODUCT_PACKAGES+=libpng is.  I think libpng and libpng-enc should be put together
08:45:11 <bero2> I think libpng is pulled in implicitly because stuff links to it
08:45:30 <yspan> I guess so
08:45:43 <xavierhsu> xruxa, I mean if I need to change jpeg, jpeg-turbo (modify by YongQin) and mine jpeg-turbo for testing my task.
08:46:07 <bero2> rebuilding the OS after replacing libjpeg is the safest thing to do
08:46:56 <xruxa> yspan, thanks.
08:47:03 <xruxa> liuyq, please go next
08:47:27 <xavierhsu> bero2, If I did't rebuild. Maybe the source code can be build(no error), but it will happen some error at run time.
08:47:52 <Qian_> yspan, try device directory
08:48:22 <xruxa> liuyq, it looks like the jobs with new Ethernet dongles failed just the same as always - I see only two benchmarks passing :-/
08:49:35 <xruxa> lets skil liuyq for a bit, zhizhoutian__ you are next
08:50:12 <xruxa> zhizhoutian__, are you here?
08:51:11 <xruxa> no luck with zhizhoutian__  either, and bero2  on phone, so let me just copy&paste here for bero bero2  :
08:51:13 <xruxa> What did you do?
08:51:13 <xruxa> Created reference builds with and without memory patchset for Nexus 7
08:51:13 <xruxa> Ran initial tests
08:51:13 <xruxa> Fixed up the Nexus 7 kernel to build with current toolchains
08:51:13 <xruxa> Cleaned up the strict-aliasing and visibility=hidden patchsets and added them to the ram patchset
08:51:15 <xruxa> Modified the build system to automatically use dlmalloc if TARGET_OPTIMIZE_FOR_SIZE is set
08:51:16 <xruxa> Modified the build system to add ro.config.low_ram=true etc. if TARGET_OPTIMIZE_FOR_SIZE is set
08:51:18 <xruxa> Tried to build things with external clang 3.7
08:51:20 <xruxa> What are you going to do?
08:51:22 <xruxa> Check if there�s anything we can do to shrink the kernel
08:51:26 <xruxa> Clean up and merge more local patches that reduce memory usage
08:51:28 <xruxa> Fix problems with external clang 3.7
08:51:30 <xruxa> Investigate the version of clang on AOSP master, it seems to have LTO fixes
08:51:32 <xruxa> What problems do you have?
08:51:34 <xruxa> Nexus 7�s kernel is too old -- 3.4 doesn�t have KSM or ZRAMFS, so we can�t measure their effectiveness on the reference build. Kernel team will look into it
08:51:37 <xruxa> bero2, anything you wanted to add?
08:52:28 <xruxa> ok, moving on to eric____ - please eric____ go next with your status
08:53:47 <eric____> yes
08:53:48 <xruxa> tough luck with attendance today :-P liuyq, zhizhoutian__ and eric____
08:53:55 <xruxa> eric____, please go ahead
08:54:29 <eric____> I found libwebp has already sparated into enc and dec on android M
08:54:48 <xruxa> eric____, can you point us to the files so that bero2 could check quickly?
08:55:40 <eric____> which file do you want? I got the message from README.android in webp directory
08:56:19 <xruxa> eric____, update the JIRA task to point to the file(s) and bero2 will have a look. Thanks.
08:56:43 <xruxa> #action eric____ update the JIRA task to point to the file(s) and bero2 will have a look at the alrady split libwebp
08:56:46 <xruxa> eric____, thanks.
08:56:57 <eric____> :)
08:57:09 <eric____> already done
08:57:37 <Qian_> eric____: which file? libwebviewchromium_loader.so?
08:57:46 <xruxa> wuhai, please go next
08:58:12 <eric____> Qian_: no
08:58:23 <xruxa> eric____, point us to the exact files in your sources, and add those on LMG-977
08:58:54 <xruxa> wuhai, ?
08:58:55 <eric____> xruxa: OK
08:59:08 <wuhai> xruxa:ok
08:59:13 <wuhai> What did you do?
08:59:13 <wuhai> 1.Continue to analyzes the structure of code LocationManagerService.
08:59:13 <wuhai> 2.Discuss low memory project status with liuyongqin and sprd engineer。
08:59:13 <wuhai> 3.Learning linaro code management process from liuyongqin
08:59:13 <wuhai> 4.Learning  apply the linaro patch to T-shark baseline from liuyongqin
08:59:13 <wuhai> What are you going to do?
08:59:13 <wuhai> 1.Verify the linaro patch effect on T-Shark baseline.
08:59:14 <wuhai> What problems do you have?
08:59:14 <wuhai> I still need to continue to work in LMG-963 ? because the android-M has done a lot in this optimized,
08:59:15 <wuhai> I think I can do someting that linaro the patch port to t-shark baseline and verify patch effect
09:00:27 <xruxa> wuhai, unless you agree with bero2 please continue on the system service code optimization work
09:01:05 <xruxa> wuhai, I do not see any comment on the LMG-963 with notes - add those that will help discussions with bero2
09:01:39 <xruxa> wuhai, thanks.
09:01:45 <xruxa> wangjian, please go next
09:01:52 <wuhai> ok,i will do this
09:02:47 <wangjian> What did you do?
09:03:07 <wangjian> 1. Analyze launcher resource loading process.
09:03:07 <wangjian> 2. Discuss with Liu Yongqin about save memory projects, learning patch download from linaro, merger and other operations.
09:03:13 <wangjian> What are you going to do?
09:03:32 <wangjian> 1. Spreadtrum has done a lot of optimization for the launcher, but does not merger into T-SHARK code now. I need to communicate what they do Spreadtrum engineers to optimize, avoid duplication of work.
09:03:32 <wangjian> 2. Merger and verify bero the patch set On T-SHARK with Liu yongqin.
09:03:49 <wangjian> What problems do you have?
09:04:12 <wangjian> I found that launcher optimized work is details to modify code, which requires a lot of time to debug, and for a specific platform, such as T-SHARK, the code differences very large to AOSP, so this  task of reducing memory optimization project is useless. I've communicated with Liu Yongqin, maybe I can get other task.
09:04:53 <xruxa> wangjian, I see, we do not want you to take or leak exsting spreadrum code unless spreadrum is willing to donate those optimizations. Check with bero2 before going forward.
09:05:33 <xruxa> #action wangjian to check with bero2 about launcher work - he says Spreadrum made many optimizations already and we need to make sure we do not just take something that was not contributed properly
09:05:34 <wangjian> OK,I will discuss with him
09:06:09 <xruxa> #action bero2 to check if we can give wangjian another task in the meanwhile
09:06:43 <xruxa> wangjian, thanks. if you find some tasks in the backlog that you like ping bero2 , otherwise he will decide himself
09:06:53 <xruxa> yuxing, please go next
09:08:13 <xruxa> liuyq, zhizhoutian__ , yuxing - any of you here to report status?
09:09:15 <yuxing> xruxa: sorry
09:09:30 <xruxa> yuxing,yes please, go ahead with your status
09:09:30 <yuxing> What did you do?
09:09:43 <yuxing> 1、 Investigate how to enable ksm on mobile.
09:09:53 <yuxing> 2、 Investigate the ksm related parts in android system (not kernel driver)
09:10:04 <yuxing> 3、 Investigate ksminfo tool(system/extras/ksmutils/), through the tool, analysis the ksm
09:10:11 <yuxing> What are you going to do?
09:10:26 <yuxing> 1、Finish above 1 and 2       2、continue Investigate ksminfo
09:10:39 <yuxing> What problems do you have?
09:10:42 <yuxing> none
09:10:47 <xruxa> yuxing, thanks, looks like progressing well.
09:10:59 <xruxa> alright, end of status round
09:11:28 <xruxa> #topic AOB
09:11:48 <xruxa> any other topics to discuss? Otherwise we close the meeting in a minute
09:12:26 <xruxa> last call for topics
09:12:47 <xruxa> ok, thanks all for attending and letting us know how you are doing, closing the meeting
09:12:51 <xruxa> talk to you tomorrow
09:12:59 <xruxa> #endmeeting