posted by felix9

GCLargeObjectHeapCompactionMode is next time.

posted by felix9

posted by felix9

what's next for Windows, Windows Server, Windows Azure, Visual Studio ...

what about XBox, inside the 'and more' ?

posted by felix9

posted by felix9

http://wesnerm.blogs.com/net_undocumented/2013/03/immutable-collections-critique.html

posted by felix9

If the MSR work you talked about is this
http://research.microsoft.com/apps/pubs/default.aspx?id=170528

then its very interesting to think about it, IIRC (could be wrong) the 'immutable' reference applies to the whole object graph can be reached from it, meaning a 'immutable' collection can't have mutable objects as elements !? so the pattern here is actually quite different.

That said, as far as I understand it, the algorithm itself should be perfectly suitable for immutable references, you just create an 'isolated' node, mutate it as you like, including let it point to the previous immuable tree nodes etc, then return it as an immutable reference, its exactly the 'builder' pattern and 'freeze'/consume pattern, and I think this kind of ability is exactly why the approach in that paper is better than previous works.

posted by felix9

posted by felix9

posted by felix9

posted by felix9

posted by felix9

The bad thing is, now I fear writing any multithreaded code since I'm not using a C++11 compiler! coding without a memory model is so scary ! How many bugs are there in VC10 ? VC9 ?

posted by felix9

But, apparently the captions are encoded into the video ?

If captions are in seperate files, it will make translations/multi-language support much easier.

posted by felix9

posted by felix9

posted by felix9

posted by felix9

posted by felix9

http://www.microsoft-careers.com/job/Redmond-Development-Manager%2C-Principal-Startup-Business-Group-Job-WA-98052/2312102/

Introduction - Beyond Kinect

Do you want to part of a team that builds the next big, game-changing entertainment and communication experiences? We are developing a cutting-edge new video and depth-sensor based technology that goes beyond anything on the market today.

This position will lead development of a V.1 game-changing technology currently under wraps at Microsoft. This is a newly created, high-visibility, high-impact role within Microsoft's Startup Business Group (SBG) specifically developed for this incubation project. This team is developing multiple products, based on this new technology that will be shipped with the XBOX and Lync/Skype divisions. A critical part of this role is the transitioning of complex, cutting-edge R&D into shipping products.

In particular, the SBG team seeks demonstrated world-class technical expertise and accomplishments in at least one or more of the following areas:
• Computer Vision, Computer Graphics, 3D Reconstruction, Multi-view Video, Image Interpolation, Field Programmable Gate Arrays (FPGAs)
• Digital Media and Entertainment, particularly video conferencing and video compression and decompression, and transmission/transport of video across large networks
• Game Development, specifically game engine development, complex game systems

posted by felix9

posted by felix9

posted by felix9

But actually, async timeout notification is useless in XP anyway since there is no CancelIoEx, and CancelIo can only be called in the same thread that started it. So I kinda understand why MS teams all left XP support behind, because XP stinks. but I'm not Microsoft so I still have to.

Now I still want to ask, if I use those new Vista APIs, can I start a socket (TCP/UDP) overlapped io and let the threadpool notify me if its not yet completed after a timeout ?? I mean, I know maybe I CAN do that, but maybe I need both ...Wait APIs and ...Io APIs, and set 2 callbacks, but IMHO thats too much for such a common task, can I get the notification with just ...Io APIs ?

posted by felix9

Thread pool is my working horse for everything not in the UI. I use QueueUserWorkItem to do async work, RegisterWaitForSingleObject to do notification/timeout/timer, and I use it to do (overlapped) IO too, I wish I can use BindIoCompleionRoutine to do IO but that dont have a async timeout notification as far as I can see, so I use RegisterWaitForSingleObject with the event in OVERLAPPED instead, which works, but its kinda wierd that BindIoCompletionRoutine has such a flaw, and I wonder if using RegisterWaitForSingleObject for IO completion has any performance hit, since its TP_WAIT while BindIoCompletionRoutine uses TP_IO ? on the other hand I wish there is RegisterWaitForMultipleObjects which would be super useful.

posted by felix9

posted by felix9

The exported funtions in pgodb110.dll

PogoDbIsLocalSymbol
PogoDbAllocEntryProbeId
PogoDbAllocSimpleProbeId
PogoDbAllocValueProbeId
PogoDbClose
PogoDbCloseCallGraph
PogoDbCloseFunction
PogoDbCloseModule
PogoDbCreateGlobalStreams
PogoDbEnableThreadSafetyChecks
PogoDbFGNodeType
PogoDbFinishFunction

The modules in pgodb110.dll and pgort110.dll

See ...Src\Tools\Pogo\....

0002 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\pogodb\objr\i386\pogodb.obj
0003 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\pogodb\objr\i386\pogodbmt.obj
0004 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\pogodb\objr\i386\loadui.obj
0006 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\pogodb\objr\i386\pgodb.rso
0002 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\irt\pgort\objr\i386\irtdll.obj
0003 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\irt\pgort\objr\i386\pogowin32interface.obj
0004 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\irt\pgort\objr\i386\pgort-loadui.obj
0005 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\irt\pgort\objr\i386\pgc.obj
0007 F:\dd\vctools\Compiler\UTC\Src\Tools\Pogo\irt\pgort\objr\i386\pgort.rsoj

posted by felix9

Ah, bad joke.

posted by felix9