Hi all,
I've written a simple class that encapsulates the cpuid instruction. It could probably be improved and polished, but I'm fairly happy with it.

One thing I'm considering is if a static only class would be better, it's not like the info can change dynamically...

Get it here: http://famolsen.dk/d/cpuid.zip

The zip contains the cpuid module and a simple test app. Running the test app I get personally get this:

Vendor string:    AuthenticAMD
Processor string: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
Features:         MMX SSE SSE2 SSE3 3DNow! 3DNow!+ MMX+ AMD64 HTT
Multithreading:   2 threads / 2 cores

The threading info should work on both Intel and AMD cpus, but I have no way to test the Intel path. Any feedback would be much appreciated.

Also, I'm considering creating a library with sse/2/3 optimised batch operations for vector operations normalisation etc. Would such a library be of interest to the D community?

Tomas Lindquist Olsen wrote:
> Hi all,
> I've written a simple class that encapsulates the cpuid instruction. It could probably be improved and polished, but I'm fairly happy with it.
> 
> One thing I'm considering is if a static only class would be better, it's not like the info can change dynamically...
> 
> Get it here: http://famolsen.dk/d/cpuid.zip
> 
> The zip contains the cpuid module and a simple test app. Running the test app I get personally get this:
> 
> Vendor string:    AuthenticAMD
> Processor string: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
> Features:         MMX SSE SSE2 SSE3 3DNow! 3DNow!+ MMX+ AMD64 HTT
> Multithreading:   2 threads / 2 cores
> 
> The threading info should work on both Intel and AMD cpus, but I have no way to test the Intel path. Any feedback would be much appreciated.
> 
> Also, I'm considering creating a library with sse/2/3 optimised batch operations for vector operations normalisation etc. Would such a library be of interest to the D community?

Very yes.

On my machine...

D:\Users\invernizzi\tmp>build test.d -exec
Vendor string:    GenuineIntel
Processor string:               Intel(R) Pentium(R) D CPU 3.00GHz
Features:         MMX SSE SSE2 SSE3 AMD64 HTT
Multithreading:   2 threads / 2 cores

---
Paolo Invernizzi

Tomas Lindquist Olsen wrote:

> Vendor string:    AuthenticAMD
> Processor string: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
> Features:         MMX SSE SSE2 SSE3 3DNow! 3DNow!+ MMX+ AMD64 HTT
> Multithreading:   2 threads / 2 cores
> 
> The threading info should work on both Intel and AMD cpus, but I have no way to test the Intel path. Any feedback would be much appreciated.

Paolo Invernizzi wrote:
> On my machine...
> 
> D:\Users\invernizzi\tmp>build test.d -exec
> Vendor string:    GenuineIntel
> Processor string:               Intel(R) Pentium(R) D CPU 3.00GHz
> Features:         MMX SSE SSE2 SSE3 AMD64 HTT
> Multithreading:   2 threads / 2 cores
> 
> ---
> Paolo Invernizzi
> 
Hmm. Is that information correct? I thought all recent Pentium CPU's had two hardware threads per core.

Paolo Invernizzi wrote:
> On my machine...
> 
> D:\Users\invernizzi\tmp>build test.d -exec
> Vendor string:    GenuineIntel
> Processor string:               Intel(R) Pentium(R) D CPU 3.00GHz
> Features:         MMX SSE SSE2 SSE3 AMD64 HTT
> Multithreading:   2 threads / 2 cores

And here's a single-thread result:

>bud test.d -exec
Vendor string:    GenuineIntel
Processor string:         Intel(R) Pentium(R) M processor 1.86GHz
Features:         MMX SSE SSE2
Multithreading:   1 threads / 1 cores

> 
> ---
> Paolo Invernizzi
> 
> Tomas Lindquist Olsen wrote:
> 
>> Vendor string:    AuthenticAMD
>> Processor string: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
>> Features:         MMX SSE SSE2 SSE3 3DNow! 3DNow!+ MMX+ AMD64 HTT
>> Multithreading:   2 threads / 2 cores
>>
>> The threading info should work on both Intel and AMD cpus, but I have no way to test the Intel path. Any feedback would be much appreciated.


> Also, I'm considering creating a library with sse/2/3 optimised batch
> operations for vector operations normalisation etc. Would such a library
> be of interest to the D community?

Yes. But, make sure you're not just reinventing the wheel, since it's such a well-explored problem. I think it would make sense for D to come with a simple BLAS implementation for general use.

I just found one little BLAS1 implementation that's under a BSD license:

http://iridia.ulb.ac.be/~fvandenb/miniSSEL1BLAS/miniSSEL1BLAS.html

which looks as though it is D-compatible and easy to port. No doubt there are many others on the net.

Tomas Lindquist Olsen wrote:
> Hi all,
> I've written a simple class that encapsulates the cpuid instruction. It could probably be improved and polished, but I'm fairly happy with it.
> 
> One thing I'm considering is if a static only class would be better, it's not like the info can change dynamically...
> 
> Get it here: http://famolsen.dk/d/cpuid.zip
> 
> The zip contains the cpuid module and a simple test app. Running the test app I get personally get this:
> 
> Vendor string:    AuthenticAMD
> Processor string: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
> Features:         MMX SSE SSE2 SSE3 3DNow! 3DNow!+ MMX+ AMD64 HTT
> Multithreading:   2 threads / 2 cores
> 
> The threading info should work on both Intel and AMD cpus, but I have no way to test the Intel path. Any feedback would be much appreciated.
> 
> Also, I'm considering creating a library with sse/2/3 optimised batch operations for vector operations normalisation etc. Would such a library be of interest to the D community?

Thank you and yes, of course.

On Tue, 24 Oct 2006 06:11:42 -0400, Tomas Lindquist Olsen <tomas@famolsen.dk> wrote:

> Hi all,
> I've written a simple class that encapsulates the cpuid instruction. It could probably be improved and polished, but I'm fairly happy with it.
>

Cool! I was just thinking about detecting the CPU recently.


Vendor string:    GenuineIntel
Processor string:               Intel(R) Pentium(R) 4 CPU 2.66GHz
Features:         MMX SSE SSE2 HTT
Multithreading:   1 threads / 1 cores

HTT with 1 thread/core?

Vendor string:    GenuineIntel
Processor string:               Intel(R) Pentium(R) 4 CPU 2.80GHz
Features:         MMX SSE SSE2 HTT
Multithreading:   2 threads / 1 cores

>
> Vendor string:    GenuineIntel
> Processor string:               Intel(R) Pentium(R) 4 CPU 2.66GHz
> Features:         MMX SSE SSE2 HTT
> Multithreading:   1 threads / 1 cores
>
> HTT with 1 thread/core?

HTT can be disabled/enabled in bios, maybe you have it disabled?

On Tue, 24 Oct 2006 13:11:42 +0300, Tomas Lindquist Olsen <tomas@famolsen.dk> wrote:

> Hi all,
> I've written a simple class that encapsulates the cpuid instruction. It could probably be improved and polished, but I'm fairly happy with it.
>
> One thing I'm considering is if a static only class would be better, it's not like the info can change dynamically...
>
> Get it here: http://famolsen.dk/d/cpuid.zip
>
> The zip contains the cpuid module and a simple test app. Running the test app I get personally get this:
>
> Vendor string:    AuthenticAMD
> Processor string: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
> Features:         MMX SSE SSE2 SSE3 3DNow! 3DNow!+ MMX+ AMD64 HTT
> Multithreading:   2 threads / 2 cores
>
> The threading info should work on both Intel and AMD cpus, but I have no way to test the Intel path. Any feedback would be much appreciated.
>
> Also, I'm considering creating a library with sse/2/3 optimised batch operations for vector operations normalisation etc. Would such a library be of interest to the D community?

c:\>build test.d -exec
Vendor string:    AuthenticAMD
Processor string: AMD Athlon(tm) XP 2000+
Features:         MMX SSE 3DNow! 3DNow!+ MMX+
Multithreading:   1 threads / 1 cores

it works fine

-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/

Tomas Lindquist Olsen wrote:
> Get it here: http://famolsen.dk/d/cpuid.zip

I'd like to put this into Phobos. Can you explicitly make it public domain?