News Intel's Haswell brings transactional memory tech

[brumgrunt]

Supercomputing tech reaches the desktop.

http://www.bit-tech.net/news/hardwar...-haswell-tsx/1

[r3loaded]

Another clever thing about this tech is that it's backwards compatible with existing software that uses locks for safe concurrency. The thread will believe it has an exclusive lock on the data, but the system will actually allow another thread access to the data in a transactional manner. This'll allow a speed improvement in cases where thread #1 is modifying the data, but thread #2 is only reading it or modifying a non conflicting part of it. Of course, software written with transactions in mind will run even better.

Sent from my GT-I9000 using Tapatalk

[DbD]

From what I can tell this is just a very low level somewhat restrictive read/write lock. Except instead of waiting for a write lock everything just runs at once and the one that commits the finished operation back to memory first gets to do it, the others then finish, get told they failed and have to start again.

Hence it seems very computensively expensive as before if 6 processes all wanted a write lock 5 would wait and one would do it's work. Now all 6 go for it, but 5 have to abort and start all over again.

[schmidtbag]

if we're going to have features that are incompatible with current software, personally i'd rather go with amd's route of using a fused cpu and gpu rather than TSX. i'd like to know how much of a difference tsx makes, because i feel like certain tasks won't benefit from this at all.

[azrael-]

I'd like to refer you to Andreas Stiller's article from December 2011. In a nutshell, with future many core CPUs transactional memory isn't a "nice-to-have" feature but more of a "must-have" feature.

PS: You can read the original German version here.

[b1candy]

Originally Posted by schmidtbag

if we're going to have features that are incompatible with current software, personally i'd rather go with amd's route of using a fused cpu and gpu rather than TSX. i'd like to know how much of a difference tsx makes, because i feel like certain tasks won't benefit from this at all.

How can you program for extensions you don't even know about yet?
Like SSE/SSE2, it made a difference when programmers started coding for it. It didn't help straight away on legacy software. And it only helps if the situation warrants it. You can't just go around implementing everything on everything - these functions have specific tasks, not a catch all situation.

[Buddidge]

if it does to cpu's as it did databases then this could be epic.

sent from my keyboard using my fingers! :-P

[Gradius]

There is any "demo" (simulated is ok) around?

[r3loaded]

Originally Posted by Gradius

There is any "demo" (simulated is ok) around?

I'm sure Intel will demo it at an engineering conference once their Haswell chips are ready to be shown to the world.

[Risky]

Originally Posted by schmidtbag

if we're going to have features that are incompatible with current software, personally i'd rather go with amd's route of using a fused cpu and gpu rather than TSX. i'd like to know how much of a difference tsx makes, because i feel like certain tasks won't benefit from this at all.

I'm not sure I see the point here. This doesn't preclude a CPU/GPU at all.

And yes certain talks won't benefit from it but that appliest to almost and particular element of processor technologyn

[fluxtatic]

Originally Posted by schmidtbag

if we're going to have features that are incompatible with current software, personally i'd rather go with amd's route of using a fused cpu and gpu rather than TSX. i'd like to know how much of a difference tsx makes, because i feel like certain tasks won't benefit from this at all.

You missed it - it isn't incompatible. There is a still a fallback to the current lock mechanisms. The question is, with as much software that remains unupdated as far as threading goes around, how long will we have to wait for software that can take full advantage of this? If the procs aren't due to drop 'til 2013, it'll be 2020 before we see mainstream software that can take advantage of this (assuming it isn't deemed a failure and dropped in the arch that follows Haswell.)