Please help me look at my submission is successful no, thank you.

            


I just submitted my work. 7 after the decimal point on the Core 2 E6600 measured performance is 0.25 seconds.
Please help me look at my submission is successful no, thank you.
My mail is flyingdog88@gmail.com

Reply:
Strong ah!
Reply:
I was impressed with the double-precision single computing x ^ 2 + y ^ 2 + z ^ 2 will be more than 0.25 seconds time strongman ah!
Reply:
flyingdog88@gmail.com


Hello, I have received your masterpiece
Reply:
I am sorry, because I uploaded report is relatively large. Compressed file has 1M. Can you help me confirm, compressed file you received is complete? Thank you!
Reply:
flyingdog (flyingdog)

Your works have: version1, version2, version3 (final), optimization report, potential_serial (final) .exe
Reply:
Thank you. All the.
Three versions are sequential optimization, parallel optimization, optimization of the different stages of compilation.
Reply:
I do not know compilation, suffer a great deal, only to 0.57s
Reply:
lanphaday (Love admiral)

So I like to take this opportunity to learn about
Reply:
My experience: If the code is the same, with vc2005 compiled code may be slower than handwritten compilation of about 10% -15% (feeling mainly in the planning capacity of people worse than register),
ICC compiled code and many handwritten compilation speed considerably;
Reply:
If the code rule. ICC compiler efficiency is very high. Handwritten compilation is barely benefit of. Instead, the ICC is to see the generated code can improve their abilities.

But for complex logic might not be. In particular, the compiler can only be understood in the perspective of the C language program, while the user can understand the overall logic of the program's point of view.

I do not have anything to use the assembly. Only a snippet of code probably use about 20 lines of assembly. And because time is relatively short, and there is no good assembly code optimization, there is still much room for optimization. If the time to loose points, together with the other code compilation, and estimated time to break 0.2s.

Originally intended synchronization function written in assembler done in user mode, but too late, do not write.

Over time, I have the source code announced, we look at, certainly can further increase speed.

Another point I was very depressed. When I developed using Xeno 5130's CPU, and the same Core 2 Duo-based core microarchitecture. However, there seems to compatibility problems when using VTune. Unable to identify the CPU. I do not know other people use Core 2 or Xeno 5100 series do not have this problem.

Reply:
You too, seemingly to ensure the accuracy and answers under the right circumstances, you are the only one I know of who break 0.4s.

Well tomorrow or the day we expect to optimize public report, when you can learn something ~~~

I will open, because I'm calling on everyone, quack ~~

Drifting ~~
Then write the document to ~
Reply:
flyingdog with the E6600, envy ing;
The CPU is very strong, especially SSEx instruction
lanphaday (Love admiral) code on the E6600 can also be set to within 0.4s

Waiting to see the report: D



Reply:
housisong said it right. E6600 too strong.
In fact, Core 2 Duo or a core microarchitecture strong.

Reply:
Finally finished documents, and submit up, you can go home ,,, it is still firm. .

Troublesome staff have not submitted a look at the success of my email: lanphaday@126.com, just submitted.
Thank you. . . .