/srv/irclogs.linaro.org/2011/09/23/#linaro-multimedia.txt

=== fangbao_ is now known as fangbao
=== fangbao_ is now known as fangbao
=== fangbao_ is now known as fangbao
ronynandyhi ibiris08:39
ibirisronynandy: hi08:40
ronynandyibiris: any idea why http://status.linaro.org/11.11/  is showing a lower perstange for  http://status.linaro.org/11.11/group/tr-multimedia-neon-optimization-forum.html08:41
ronynandythere are no TODO items but the burndown chart is showing some TO workitesm08:42
ibirischecking08:42
ronynandyI had seen this last month also but had thought that it will change later08:43
ibirisronynandy: my best guess is that some work items had a line break in them08:44
ibirisronynandy: so part of the line was without any designation (DONE) as the line was broken08:44
ronynandyoh ok,i will chage it..08:44
ronynandythx08:44
ibirisI have removed the line breaks  let's see of that fixes it08:44
ibirispls have a look as well08:45
ronynandyibiris:thx,no change in http://status.linaro.org/11.11/08:47
ibirisronynandy: it will take some time08:47
ibirisslo does not update instantaneously08:47
ronynandyI think the scripts in the server run at predetermined intervals08:47
ronynandyyes correct08:47
ibirisronynandy: y08:48
ibirisbest bet is to check again tomorrow08:48
ibirisronynandy: from the log ("update log" link at the bottom of slo page) it seems it runs every hour on the half hour.08:50
ibirisso check again  in 40 mins or so08:50
ibiris(or whatever time for your TZ :-)08:50
ronynandysure,good find..I was just thinking of asking someone on the linaro channel :)08:51
ibiriskan_hu: ping10:37
ibiriskan_hu: so in your last email there was all the information regarding formats and source for the sample media as well as the codecs used right? was there anything else you'd like to point out?10:38
ibiriskan_hu: I have:10:38
ibiriswe used MPEG4+MP3 in AVI container10:38
ibirisMPEG2+MP3layer2 in MPG file10:38
ibirisH264+AAC LC in MP4 file10:38
ibirisVC1+WMA3 in ASF file10:38
ibirisVP8+Vorbis in WEBM file10:38
av500you miss H264+DTS in MKV :)10:39
av500aka what the real world uses10:39
ibirisav500: not sure if that combination was covered in the samples kan_hu is releasing in 1109 :-)10:40
ibirisbut sure kan_hu ^^^ is that combination covered as well ? Do I need to add it to the email I am sending to the Linaro TSC ?10:41
av500also, MPEG2 + layer2 is more likely10:41
av500ah sorry, it is layer210:41
av500ignore10:41
ibirisok10:41
av500MP3layer2 is a bit funny :)10:41
ibirisbbiab10:41
kan_huibiris, yes10:43
kan_husome other audio only format, you can check the csv file on the samplemedia.linaro.org10:43
kan_husorry, should be mpeg1 layer210:44
ibiriskan_hu: that's great, but I doubt the TSC will open the CSV file to check in such detail.10:45
ibiriskan_hu: can you summarize it for me the way you did for Loic in your email10:45
ibirisyou can send me a summary via email10:46
kan_huibiris, I have send you mail including all codec combinations including audio only for samplemedia10:52
kan_huI have go home now10:52
krtaylorronynandy, ping11:38
ronynandyhi11:41
ronynandykrtaylor:hi11:41
krtaylorhi ronynandy11:41
ronynandyhi krtaylor: does the test detail suffice or should i put more details?11:42
benjiGrobclark: hi, can you explain me for what we have attach/detach functions in dma-buf ?12:18
benjiGkrtaylor, robclark: in UMM/dma-buf roadmap I hope that dev/test/example of scatter list is coming soon12:23
robclarkbenjiG, attach/detatch is basically "I'm going to start using buffers 1 or more times", and "I'm done for now"...12:55
robclarkthat might correspond to, for example, v4l2 STREAMON/STREAMOFF12:55
robclarkwhere {get,put}_scatterlist corresponds to individual DMA's, ie individual frames of video or frames from camera..12:56
robclarkthe idea.. the exporter can cache the mapping until detatch12:56
robclarkso each get_scatterlist returns the same..12:56
robclarkbut if there is memory pressure and he needs to evict some buffer, because of the put_scatterlist, he knows which buffers are not currently being DMA'd to and so are safe to evict12:57
benjiGrobclark: why dma_buf_{get,put} aren't enough to know when buffer is still in use ?13:01
robclarkthey are enough to know when it is not in use13:02
robclarkbut not enough to give a hint that it won't be used again with that device for a while13:02
benjiGis it because you want to manage a buffer split in multiple non-contiguous pages ?13:03
robclarkso makes it tricky if you want to try and cache the mappings13:03
robclarkwell, it is for devices w/ iommu or something like that13:03
ibiriskrtaylor: please check the email I sent to the TSC (with input from kan_hu on the list of formats and codec combinations)13:04
* ibiris will bbiab13:04
krtayloribiris, already did, looks good - but I still dont understand the codec link13:05
krtaylorwe are just distributing samples13:05
benjiGrobclark: to be honest I not yet understand well :-) maybe with a sample code it will be more clear13:10
robclarkwell, assume for example your video decoder had an MMU that the buffer needed to be mapped in to13:11
robclarkthe first get_sglist would map it in13:12
robclarkbut put_sglist would mark it as candidtate to unmap.. but you might not want to immediately unmap if there is still free space in the mmu/iommu13:12
benjiGI had imaging that sglist was about userspace mmaping of buffer not for hw devices13:16
robclarkno, for hw devices13:26
benjiGfor example a android video decoder can request a big buffer of 64MB and use offset to split in multiple N sub-buffer13:27
benjiGI was thinking not allocate a big chunk of 64MB but N small chunks and use sglist to mmap them in one virtual buffer13:27
ibiriskrtaylor: for our 1-1 I am having some trouble to chat now. is it possible to move it to Tuesday instead?13:56
ibiriskrtaylor: sorry for the very short notice :-[13:57
krtayloribiris, sure, thats fine14:01
krtaylorrobclark, ping14:09
benjiGkrtaylor: do you got my mail about khronos presentation ?14:11
robclarkkrtaylor, pong14:12
krtaylorbenjiG, yes, thats ok, we've done what we can do then14:13
krtaylorrobclark, I need to get 30 mins with you today sometime14:13
krtaylorirc is fine or mumble or phone14:13
robclarkyeah, ok14:14
krtaylorrobclark, when is a good time?14:14
robclarkumm, I guess in the afternoon/evening?14:16
krtaylorrobclark, ok, we need to close out the eval review, we can fill it out at the same time if you prefer14:16
robclarkJesseBarker, I don't suppose you could create https://blueprints.launchpad.net/linaro-mm-sig/+spec/engr-mm-kms-2011.10 ?21:20
JesseBarkerrobclark, Done.  But I need 2 small things from you now that it's done.21:24
robclarkuhh ohh21:24
JesseBarker1) Move your backlog from the 2011.09 blueprint into this one (or into the general backlog if you're not doing them right away).21:24
JesseBarker2) Update the title/description with something more descriptive than "KMS work for 2011.10".21:25
robclarkyup.. that is why I needed 2011.1021:25
robclarkoh, ok21:25
JesseBarkerI figured as much (sorry if my statement sounded ominous :-)21:26
robclarkno worries21:26
ibirisJesseBarker: robclark: I was about to check if the 1110 blueprint needed any help from me, seems you guys have it all under control :-)21:28
robclarkyup21:29
ibirisokdok then I will remove it from the todo list I was making - time to call this week over for me21:30
JesseBarkeribiris,  I've been updating the backlog blueprints so that they have the backlog milestone targets, let me know if you see any i've missed.21:30
ibirisJesseBarker: yup this is in my list for Monday check :-D21:31
ibirisJesseBarker: btw was there any issue with the blueprint naming scheme (alf mentioned this earlier to me)21:31
ibiris- not so much a mmedia issue - but since you are here ... :-)21:31
JesseBarkeribiris, given the shift in the way we are doing things, we thought the actual names could get shorter and stay descriptive.21:32
ibirisJesseBarker: there is no official recommendation for the names so we can change them if they need a trim21:32
JesseBarkere.g., glmark-dpm rather than engr-benchmark-glmark2-2011.10 to reflect the addition of a deep parallax map test scenario to glmark221:33
JesseBarkerthis seemed especially relevant if we're moving in the direction of a blueprint per feature (which is something Alexandros is piloting) rather than a blueprint per component per cycle.21:34
JesseBarkerthis idea seems to fit better with launchpad and our workflow, but it would really be great if we had a better interface for creating/managing blueprints.21:34
ibiris+1 for me (on both statements - name and interface)21:35
JesseBarkeribiris, we can talk about it next week (I'm supposed to start an email thread).  Go have yourself a good weekend!!!21:35
ibirisok signing off, let's talk next week... thanks JesseBarker21:35

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!