Visual Studio 2010 C++ Express Beta

[keinrath]

Anyone tried this? Does it produce faster code than Visual Studio 2008 C++ Express ?

thanks

[jdart]

The Express versions don't do PGO, which you need for really good performance.

I heard VC++ 2010 betas were buggy so I haven't tried it yet.

--Jon

[pijl]

I heard VC++ 2010 betas were buggy so I haven't tried it yet.

Beta 1 was very buggy.
Beta2 works fine and is a big improvement over VS2005 (I skipped VS2008).
I'm using the 'premium' beta, not express, which does do PGO and has a profiler included.
Richard

[mjlef]

Anyone tried this? Does it produce faster code than Visual Studio 2008 C++ Express ?

thanks

I installed Visual Strudio 2010 C++ Express, and in testing my program, it ran about 1.4% faster. Note my program is mostly 32 bi, this was a Win32 compile on Windows XP, and running on an Atom processor notebook.

Install was the typical annoying Microsoft. After downloading 200 mbytes of stuff, it instal some, then required a reboot, then installed more, then required a second reboot then finished the install. Expect to take about 45 minutes or so to do this.

Your mileage will vary

Mark

[jarkkop]

Why not try the Microsoft Visual Studio 2010 Premium Beta 2 version


http://www.microsoft.com/downloads/deta ... laylang=en

[mjlef]

Why not try the Microsoft Visual Studio 2010 Premium Beta 2 version


http://www.microsoft.com/downloads/deta ... laylang=en

OK, I tried that. Took nearly three hours to install. I get just under a 2% speedup using PGO. But I am not expert on how to refine PGO (how long should the instrument run be?).

[Teemu Pudas]

Why not try the Microsoft Visual Studio 2010 Premium Beta 2

Why Premium when there's Ultimate?

http://www.microsoft.com/downloads/deta ... laylang=en

[Kempelen]

Anyone tried this? Does it produce faster code than Visual Studio 2008 C++ Express ?

thanks

And does anybody knows pros and cons of using MS vs. MINGW?

[jdart]

The instrument run has be long enough to hit a representative sample of your code. I use 3 test positions run for 3 minutes each. That gets enough depth that the multithreading and other depth-dependent code is starting to run. Not sure that's optimal but it works well for me.

--Jon