GCC 9 libstdc++ Lands The C++17 Parallel Algorithms Implementation From Intel

[phoronix]

Phoronix: GCC 9 libstdc++ Lands The C++17 Parallel Algorithms Implementation From Intel

While the release of GCC 9 (v9.1) is just a few weeks ago, a late addition to this annual compiler collection update is its C++ standard library now having a C++17 parallel algorithms implementation thanks to Intel developers...

http://www.phoronix.com/scan.php?pag...lel-Algorithms

[staalmannen]

Are they also aiming to add this to llvm libc++?

[ldesnogu]

Are they also aiming to add this to llvm libc++?

Michael's last link contains this:

This is the Intel implementation of the C++17 parallel
algorithms, which has been donated to both GCC and LLVM. The upstream
project is at -

https://reviews.llvm.org/source/pstl/

So I guess the answer to your question is yes

And in fact when looking with the project name, Michael already published this: https://www.phoronix.com/scan.php?pa...lel-STL-Commit

[trifud]

PSTL is disabled by default in both LLVM and GCC since it requires TBB (which is not part of the compilers).

[babali]

If it requires tbb, that kind of sucks.

[AdrianBc]

If it requires tbb, that kind of sucks.

I see that now TBB is open source:

https://github.com/01org/tbb


Therefore I do not think that there is any reason for PSTL to suck.

Moreover, looking at the source of TBB, I see that TBB not only supports Intel processors, but also IBM POWER processors and it appears that it would be easy to port it to other architectures, e.g. ARM.

[xorbe]

> While the release of GCC 9 (v9.1) is just a few weeks ago

Away?

[ypnos]

If it requires tbb, that kind of sucks.

TBB is great, it worked very well for me so far. PSTL supports several backends, and ships with a TBB backend. So other backends are not that hard to add. OpenMP might be sensible as it also ships with the compiler.