What's New in the .NET Framework 4.5 Developer Preview

Standard Template Library

As part of the added support in Visual Studio 11 Developer Preview for the C++11 specification, the Standard Template Library (STL) support in Visual Studio is extended to provide the additional programming constructs that specification requires. Highlights include the following:

Support for new headers <atomic>, <chrono>, <condition_variable>, <filesystem>, <future>, <mutex>, <ratio>, and <thread>.

Auto-vectorizer. One of the biggest potential performance gains is doing things in parallel instead of sequentially. Visual Studio 11 Developer Preview provides not only parallelism on the task level, but also at the loop level, automatically. The C++ compiler will automatically vectorize loops if it is possible. Vectorization reorganizes a loop, for example, a summation loop, so that the CPU can execute multiple iterations at the same time. By using auto-vectorization, loops can be up to 8 times faster when executed on CPUs that support SIMD instructions. For example, most modern processors support SSE2 instructions, which allow the compiler to instruct the processor to do math operations on 4 numbers at a time.

Because the compiler recognizes loops that can be vectorized when you compile your code, this occurs automatically. Note that this differs from parallelization, which is described in the next list item.

Auto-parallelizer. Visual Studio 11 Developer Preview is improving the way you can take advantage of multiprocessor and multi-core hardware. With the the auto-parallelizer, a loop is reorganized so that it can be executed on multiple threads at the same time. This means that your application can take advantage of multi-core CPUs and multiprocessors to distribute chunks of the work to all available processors. The parallelizer automatically reorganizes loops and executes multiple tasks, all without the need for you to add any new code to your application. You can tell the compiler which loops to parallelize with the #pragma parallelize directive.

There are some key differences between auto-vectorization and auto-parallelization. First, auto-vectorization is always on and requires no user interaction, while auto-parallelization requires the programmer to decide which loops will be parallelized. Also, vectorization improves the performance of loops on single-core CPUs that support SIMD instructions, while parallelization improves the performance of the loop on multiple CPUs and multi-core CPUs. The two features can work together so that a vectorized loop is then parallelized across multiple processors.

Mamad (moveax1ret) (4) skrev:

Fandt den, men den kræver login

Windcape (5) skrev:

Indtil i morgen :)

Mamad (moveax1ret) (3) skrev:

Auto-vectorizer. One of the biggest potential performance gains is doing things in parallel instead of sequentially. Visual Studio 11 Developer Preview provides not only parallelism on the task level, but also at the loop level, automatically. The C++ compiler will automatically vectorize loops if it is possible. Vectorization reorganizes a loop, for example, a summation loop, so that the CPU can execute multiple iterations at the same time. By using auto-vectorization, loops can be up to 8 times faster when executed on CPUs that support SIMD instructions. For example, most modern processors support SSE2 instructions, which allow the compiler to instruct the processor to do math operations on 4 numbers at a time.

Because the compiler recognizes loops that can be vectorized when you compile your code, this occurs automatically. Note that this differs from parallelization, which is described in the next list item.

Auto-parallelizer. Visual Studio 11 Developer Preview is improving the way you can take advantage of multiprocessor and multi-core hardware. With the the auto-parallelizer, a loop is reorganized so that it can be executed on multiple threads at the same time. This means that your application can take advantage of multi-core CPUs and multiprocessors to distribute chunks of the work to all available processors. The parallelizer automatically reorganizes loops and executes multiple tasks, all without the need for you to add any new code to your application. You can tell the compiler which loops to parallelize with the #pragma parallelize directive.

There are some key differences between auto-vectorization and auto-parallelization. First, auto-vectorization is always on and requires no user interaction, while auto-parallelization requires the programmer to decide which loops will be parallelized. Also, vectorization improves the performance of loops on single-core CPUs that support SIMD instructions, while parallelization improves the performance of the loop on multiple CPUs and multi-core CPUs. The two features can work together so that a vectorized loop is then parallelized across multiple processors.

Det lyder godt nok smart, det vil jeg gerne læse mere om

Mamad (moveax1ret) (2) skrev:

Standard Template Library

As part of the added support in Visual Studio 11 Developer Preview for the C++11 specification, the Standard Template Library (STL) support in Visual Studio is extended to provide the additional programming constructs that specification requires. Highlights include the following:

Support for new headers <atomic>, <chrono>, <condition_variable>, <filesystem>, <future>, <mutex>, <ratio>, and <thread>.

<mutex> og <thread> er sweet.

Så slipper man for boost :)

lorenzen skrev:

Hvorfor er det fedt at slippe for boost?

markjensen (6) skrev:

Det der "auto-parallellisering" er da set før. Eller kan dette her noget nyt?

Windcape (7) skrev:

http://www.microsoft.com/download/en/details.aspx?id=27543