Is this TBB roadmap still accurate?

[robert_jay_gould]

I was just looking for TBB's current roadmap and found this article:

http://software.intel.com/en-us/blogs/2007/10/12/threading-building-blocks-product-roadmap/

It says roadmap on it, but it's from 2007, so I was wondering if things have changed.

By taking a quick look at it and the current situation it seems IO has gotten postponed, which

in my case is ok, since I use ASIO for IO, but was wondering if TBB is still considering IO, or if

it was deemed to be outside of the problem-domain?

Thanks

[robert-reed]

Quoting - robert.jay.gould

I was just looking for TBB's current roadmap and found this article:

http://software.intel.com/en-us/blogs/2007/10/12/threading-building-blocks-product-roadmap/

It says roadmap on it, but it's from 2007, so I was wondering if things have changed.

By taking a quick look at it and the current situation it seems IO has gotten postponed, which

in my case is ok, since I use ASIO for IO, but was wondering if TBB is still considering IO, or if

it was deemed to be outside of the problem-domain?

Thanks

Well, that post being over a year old, you can easily guess that things have changed. You'll note that most of the items listed in that plan havebecome part of TBB. I/O and blocking were issues then and remain so now because of the difficulty of the problem. The tbb_thread implementation of the C++0X (at the time) proposed standard was released asa safety valve, permitting those needing access to asynchronous tasks in the context of a TBB application the ability to do so in a platform independent manner. The more general problem of keeping all the HW threads busy in a graph of task order dependencies with bounded memory while avoiding oversubscription remains a difficult problem.

[robert_jay_gould]

Quoting - Robert Reed (Intel)

Well, that post being over a year old, you can easily guess that things have changed. You'll note that most of the items listed in that plan havebecome part of TBB. I/O and blocking were issues then and remain so now because of the difficulty of the problem. The tbb_thread implementation of the C++0X (at the time) proposed standard was released asa safety valve, permitting those needing access to asynchronous tasks in the context of a TBB application the ability to do so in a platform independent manner. The more general problem of keeping all the HW threads busy in a graph of task order dependencies with bounded memory while avoiding oversubscription remains a difficult problem.

Yes, that's why I was wondering, the roadmap seemed 80% complete (most features there were already implemented). Also thanks for clearing up a bit on the current status of the other features. Right now I feel TBB has become well rounded and very good at its problem domain (Modern C++ threading "blocks").

[Dmitry_Vyukov]

Quoting - robert.jay.gould

Yes, that's why I was wondering, the roadmap seemed 80% complete (most features there were already implemented). Also thanks for clearing up a bit on the current status of the other features. Right now I feel TBB has become well rounded and very good at its problem domain (Modern C++ threading "blocks").

Since TBB is open-source project with community contributions, but developed and maintained by Intel development group, I think it will be useful if there will be up-to-date roadmap. Maybe some periodically updated document, or just periodic blog/forum posts (probably once per release). I would take a look at it.