Hyperthreading & VC7.1

[Andrew Maclean]

Does VC7.1 support hyperthreading?

I looked in the documentation but I could find no reference at all.

 

[Carl Daniel [VC++ MVP]]

Does VC7.1 support hyperthreading?

I looked in the documentation but I could find no reference at all.

Hyperthreading is supported by Windows XP, Windows 2000 SP3 (I believe) and
Windows Server 2003. There's no compiler support necessary. The
OS-provided support simply exposes multiple "virtual processors", so a
machine with a single HT-capable CPU will appear to have two processors.

-cd

 

[John Smith]

Hyperthreading is supported by Windows XP, Windows 2000 SP3 (I believe)
and

Windows Server 2003. There's no compiler support necessary. The
OS-provided support simply exposes multiple "virtual processors", so a
machine with a single HT-capable CPU will appear to have two processors.

Intel recommends that Hyper-Threading should be disabled for *all* Win2000
systems (at least it's still posted at their site)

 

[Tomas Restrepo \(MVP\)]

Hi John,

Intel recommends that Hyper-Threading should be disabled for *all* Win2000
systems (at least it's still posted at their site)

Weird. Anyway, for anyone interested in this, there's a good article (albeit
a little bit old) from MS regardgin HT support for windows here:
http://www.microsoft.com/windows2000/server/evaluation/performance/reports/hyperthread.asp

 

[John Smith]

Weird

http://www.intel.com/support/platform/ht/os.htm?iid=ipp_htm+os

 

[Tomas Restrepo \(MVP\)]

Hi John,

http://www.intel.com/support/platform/ht/os.htm?iid=ipp_htm+os

Sounds like a little FUD to me. The page clearly mentions at the begining
that it references OSs with HT optimizations, which indeed Win2k doesn't
have. In particular, it has two problems:
1- It cannot distinguish between logical and physical processors, which has
licensing issues
2- because of 1, the OS doesn't take into account the difference when
scheduling threads, leading to sub-optimal scheduling (because of possible
extra context changes when moving threads between physical processors)

It works, though, from what little I've seen...

 

[Stephan Schaem]

Intel recommends that Hyper-Threading should be disabled for *all* Win2000
systems (at least it's still posted at their site)

I use XP and disabled HT because it seem compile time goes up with it enabled.
I turn it on for testing/bebugging purposes (short of using a dual processor system)

So I'm not sure VC7.1 is multithreaded optimized. (compile process)
What I noticed also in the compile process is bunch of dead CPU time,
so I'm guessing the the compiler dont build an async queue of file to read/write ?

Stephan

 

[Carl Daniel [VC++ MVP]]

I use XP and disabled HT because it seem compile time goes up with it
enabled.
I turn it on for testing/bebugging purposes (short of using a dual
processor system)

So I'm not sure VC7.1 is multithreaded optimized. (compile process)
What I noticed also in the compile process is bunch of dead CPU time,
so I'm guessing the the compiler dont build an async queue of file to
read/write ?

Correct. The compiler is a completely synchronous, single-threaded process.

I use XP and leave hyperthreading enabled because I feel that it improves
overall system performance, especially when running a compute-intensive
application. The performance of a single-threaded application will likely
be lower with HT enabled, but the impact that compute-intensive process has
on system responsiveness (mouse responsiveness, for example) is greatly
reduced over the non-HT case.

-cd

 

[Pent]

Whidbey alpha has support for maximum concurrent builds. And you can specify
how many. It's pretty nice for HT and more than one CPUs.