Somasegar's WebLog

I assume one of the specifics is PLINQ. Another, perhaps, are some of the things borrowed from COmega.

I'm really looking forward to PLINQ. That will be huge.

To what extent is Software Transactional Memory (STM) still being considered? I note the STM C# library over at Microsoft Research hasn't been refreshed since June 2005.  I'm hoping this is because the right place for STM is to be implemented is in the runtime rather than in a library and as such that's where the current action is. In as much as I understand it I don't think STM is a silver bullet but I do think it has its place and it would be great to know what developments have taken place since SPJ et al where last interviewed on Channel 9.

According to Soma .net will push for parallelism. This is hardly a surprise as multicore CPUs and multi

[PL] Tak jak wspomniałem na konferencji Intela, podejście Microsoftu przy dostarczaniu narzędzi i technologii

Hi Tom,

We believe Software Transactional Memory (STM) is a very promising technology, particularly now that multi-core systems are becoming commonplace. STM has the potential to make writing concurrent code easier than using traditional lock-based mechanisms for controlling shared state within a program. We are continuing to do research and early development work in this area to understand how we may integrate STM into our products. We also have awarded a grant for research in STM compiler support (see: http://research.microsoft.com/ur/us/fundingopps/RFPs/Phoenix_SSCLI_RFP_Awards.aspx#Compiler_Support_for_Software_Transactional_Memory">http://research.microsoft.com/ur/us/fundingopps/RFPs/Phoenix_SSCLI_RFP_Awards.aspx#Compiler_Support_for_Software_Transactional_Memory).  

You can find recent Microsoft STM-related papers at: http://research.microsoft.com

•         Beautiful concurrency by Simon Peyton Jones – May 1, 2007 (http://research.microsoft.com/Users/simonpj/papers/stm/#beautiful)

•         Transactional memory with data invariants by Tim Harris and Simon Peyton Jones – June 2006 (http://research.microsoft.com/Users/simonpj/papers/stm/stm-invariants.pdf)

• Optimizing Memory Transactions by Tim Harris, Mark Plesko, Avraham Shinnar, and David Tarditi – June 2006 (http://research.microsoft.com/~tharris/papers/2006-pldi.pdf)

Speaking of cores

PLEASE fix the naming of .Net Framework libraries.

We have System.Core.dll in version 3.5. So how are you going to name 4.0 version? System.Core.Duo.dll?

I really thought that the awfull "windowsbase.dll" and WinFX->.Net 3.0 would be the last bad namings...

This job posting below is related to this article.

"Are you excited about the opportunity to enable the masses of developers to deliver the next generation of application experiences? Industry analysts forecast that by 2012, the client hardware platform will be massively parallel. The day of the many core machine is fast approaching, and exploiting those cycles via concurrency is the only way to leverage the utility and power of these platforms. To enable the next generation application experience, Microsoft must develop simpler and more composable ways of expressing parallelism and the tools to make developers successful at doing so."

"Microsoft Concurrency Runtime Group: Our team will focus on a set of innovative new technologies-including Transactional Memory (TM), parallel libraries, a Concurrency Runtime, and Parallel LINQ-which we believe we will provide the concurrency framework developers need to deliver the next generation of application experiences which will compel Microsoft customers."

Somasegar, you are morron! Now I understand, why MS releases worse products year by year.

The week of OOPSLA is now over and life has now returned to its regular path. Some notes from last week's...