2015-08-12 19:02:50	ototo	Ok, let's start the meeting anyway.
2015-08-12 19:02:53	mikelbbn	ok
2015-08-12 19:03:01	ototo	Good, Mike is here.
2015-08-12 19:03:08	ototo	Tor is missing though.
2015-08-12 19:03:55	ototo	mikelbbn: I wanted to thank you for sharing your code before your vacation. This is a very good first step.
2015-08-12 19:04:04	mikelbbn	no problem
2015-08-12 19:04:16	ototo	I haven't had a chance to look at it properly yet so have no feedback.
2015-08-12 19:04:45	ototo	broonie: do you have anything to share in that regard?
2015-08-12 19:05:25	broonie	Yes, I think the main thing was the thinness of the C API.
2015-08-12 19:05:26	ototo	I know Mathieu is diving into perf and had no time for review.
2015-08-12 19:06:22	broonie	It seemed like all the API offered was the ability to convert a byte stream into a tree, I was expecting a bit more help doing things like iterating the tree and finding things in it.
2015-08-12 19:07:27	mikelbbn	the user of the library needs to know the hardware structure of the coresight system and build a tree to suit
2015-08-12 19:07:30	broonie	But we should probably work out what the best way of providing feedback is here.
2015-08-12 19:08:26	mikelbbn	the snapshot reader test code takes that file format and builds a suitable tree
2015-08-12 19:08:57	broonie	That was the other thing actually... I couldn't see any way to free anything.
2015-08-12 19:09:10	 *	broonie didn't find the reader test code.
2015-08-12 19:09:36	 *	ototo was thinking about SAX-like API (hooks based) as a first quick-and-dirty parser - hooks are called when certain tags/records are parsed. He's not sure if that would be a good idea in CoreSight world though.
2015-08-12 19:10:59	mikelbbn	the /tests directory has the /snapshot_parser_lib
2015-08-12 19:11:08	broonie	That seems to be C++
2015-08-12 19:11:51	mikelbbn	very much so at present - taken pretty much verbatim from anther ARM project
2015-08-12 19:12:26	ototo	mikelbbn: do I understand correctly that the reasoning for using C++ was to reuse existing code?
2015-08-12 19:12:36	broonie	Right, so I was primarily reviewing this from the point of view of trying to use the C API.
2015-08-12 19:13:34	mikelbbn	one of the main reasons - refactored C++ and a little Java from various sources. Plus one e-mail I saw seemed to favour C++
2015-08-12 19:14:39	mikelbbn	Additionally the C++ infrastucture is design for reuse across the multiple decoder types we will eventually need to cover all the CS hardware
2015-08-12 19:16:04	broonie	OTOH as was raised originally C++ is quite hard to bind to other languages which is important for integration with existing tooling.
2015-08-12 19:16:31	 *	ototo nods
2015-08-12 19:16:51	mikelbbn	which is why we will have a C-API too
2015-08-12 19:17:35	broonie	Sure, I'm just saying that my review was mostly confined to that.
2015-08-12 19:18:19	ototo	mikelbbn: did you get any feedback from Tor since you've published your code?
2015-08-12 19:18:35	ototo	He was going to review the code.
2015-08-12 19:18:44	mikelbbn	I am currently working on gettign the ETMv4 packet processor going. once this is done I am happy for flesh out the C-API and provide a similar test for that.
2015-08-12 19:18:55	mikelbbn	no feed back from Tor
2015-08-12 19:19:03	mpoirier	mikelbbn: one thing
2015-08-12 19:19:20	mpoirier	mikelbbn: the perf integration is going ahead with etmv3.
2015-08-12 19:19:27	mpoirier	mikelbbn: simply because it was easier.
2015-08-12 19:19:42	mpoirier	mikelbbn: because of HW availability
2015-08-12 19:20:49	mpoirier	EOF
2015-08-12 19:21:12	ototo	mpoirier: which HW are you using ATM?
2015-08-12 19:21:21	mpoirier	TC2
2015-08-12 19:21:40	ototo	Oh, so it still works for you, good
2015-08-12 19:21:46	mikelbbn	I am testing a V8A trace snapshot pulled off Juno
2015-08-12 19:22:08	mpoirier	mikelbbn: right, Juno is difficult to use for me...
2015-08-12 19:24:30	ototo	I'd like to create a list of work items with statuses. To define the scope of the project so to say.
2015-08-12 19:24:42	ototo	mikelbbn: is Google Spreadsheets accessible for you?
2015-08-12 19:25:02	mikelbbn	no idea - can you supply a link?
2015-08-12 19:25:05	ototo	drive.google.com that is
2015-08-12 19:26:11	mikelbbn	OK - don't currently have a google account.
2015-08-12 19:26:12	ototo	AFAIK using google services is prohibited in ARM in general, but for our open project it should be OK I suppose
2015-08-12 19:26:40	mikelbbn	Not sure - may have to check with legal / IT about accessing this.
2015-08-12 19:26:41	ototo	mikelbbn: oh, ok, let's then use a wiki page on github
2015-08-12 19:26:55	ototo	mikelbbn: no need, let's stick with github
2015-08-12 19:27:19	mikelbbn	ok - I'll check up about google drive and we can go back there if needed
2015-08-12 19:28:01	mikelbbn	OK
2015-08-12 19:28:08	ototo	So, based on the list you've provided earlier, Mike, I'll create a wiki page with the work items and will ask you and Tor (as well as others) to check/amend the list and reflect the current status of those.
2015-08-12 19:28:51	mikelbbn	Sounds good
2015-08-12 19:29:10	ototo	As soon as the initial list os fine - I'll create github issues per work item (https://github.com/Linaro/OpenCSD/issues)
2015-08-12 19:29:17	ototo	s/os/is/
2015-08-12 19:30:07	ototo	I'll also try to get some feedback from Tor (will send the log for this meeting to him).
2015-08-12 19:30:18	ototo	...for the code Mike has shared.
2015-08-12 19:30:39	broonie	Can I suggest that we add a work item to have some demo/test code for the C API too?
2015-08-12 19:31:46	ototo	broonie: definitely! having a decent C API is vital for attracting contributors
2015-08-12 19:32:25	ototo	mikelbbn: you said you're working on ETMv4 parser ATM, right? can you share some more details on your plan in that regard?
2015-08-12 19:34:03	mikelbbn	yes, the packet processor is being debugger - packet processor takes raw trace and converts to discrete packets.
2015-08-12 19:34:41	mikelbbn	then the packet decoder will be written - this takes sequences of packets and gneerates trace of teh program flow
2015-08-12 19:34:59	mikelbbn	also - specifically ETMV4 instruction trace, not data yet
2015-08-12 19:35:48	mikelbbn	Additionaly, as the ETM is configurable, the first version will not support some of the more exotic elements
2015-08-12 19:36:19	mikelbbn	just basic instruction trace - e.g. no ld/st tracking, no full branch broadcast options
2015-08-12 19:36:39	mikelbbn	This will get us a good base to develop from 
2015-08-12 19:37:51	 *	ototo definitely needs to read CS manual or ask lots of questions at SFO15 :)
2015-08-12 19:38:50	ototo	mikelbbn: so, your plan till the next meeting is to get the packet processor working and, potentially, start the decoder, right?
2015-08-12 19:39:43	mikelbbn	Yes. The decoder will need some additional infrastructure - memory access and instruction decode, but that is essentially the plan
2015-08-12 19:40:52	ototo	mikelbbn: good. what is the planned cadence for updating your branch on github with the WIP code?
2015-08-12 19:41:16	mikelbbn	Pretty much at the end of each day.
2015-08-12 19:41:33	mikelbbn	Assuming I actually did anything!
2015-08-12 19:42:23	ototo	mikelbbn: great, thanks
2015-08-12 19:42:59	 *	ototo is at EOF for this meeting
2015-08-12 19:43:09	ototo	mikelbbn: do you have any questions/topics for discussion?
2015-08-12 19:43:18	mikelbbn	Not today
2015-08-12 19:43:26	ototo	mpoirier: do you have any questions/topics for discussion?
2015-08-12 19:43:37	mpoirier	ototo: not at this time.
2015-08-12 19:43:38	ototo	broonie: do you have any questions/topics for discussion?
2015-08-12 19:43:46	broonie	nothing from me
2015-08-12 19:44:32	ototo	Good. Thanks everyone for joining and have a nice localtime. I'll put the log for this meeting to the wiki.
2015-08-12 19:44:35	mikelbbn	broonie: did you get my mail regarding the connect session - I got a bounce from somewhere?
2015-08-12 19:44:59	broonie	Yes, I was going to reply - main thing is can you provide a title and abstract please?
2015-08-12 19:45:07	mikelbbn	no problem
2015-08-12 19:45:52	ototo	Good. The meeting is officially over. :)
2015-08-12 19:46:02	mikelbbn	bye
2015-08-12 19:46:30	broonie	mikelbbn: Great, thanks.