Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.

Different results on different PC.

eliopoulos
新手
11,343 檢視

Hi,

I have installed Windows 11, Visual Studio 2019 Community and Intel oneAPI 2021 Base and HPC Toolkit. When I run my FORTRAN program on a different PC with Windows 10, I get different results. I have tried not only the same compiler, but the same executable file as well. Is there anything I can do to solve this issue?

Best regards,

Dr. Elias N. Eliopoulos

55 回應
JohnNichols
傑出貢獻者 III
2,608 檢視

Andrew:  My humble apologies.  

I can offer no mitigating factors, except for the two academic friends who were both roundly criticized in the last few days, one has to listen but can do nothing, so you received some of the fallout.  Unjustly, but you are a good stick and a real brick of a friend. 

Fabaceae family is the nice crowd I stayed with in the summer for the holidays in Italy, as for Americans, they as well do not understand Lingua Franca (Australiano). The Fabaceae family soon learned that calling some one an old bean in Australia is a term of endearment, although one told me his father was not a baker, when I asked what his old man did for a crust.  It is easier not to explain.  

Whenever I see rate someone posts, I think of my old professor Neville Betts, who taught me statics and mechanics of materials at the same time. He was brilliant and a bit mathematical so I loved him and 100 people in the class hated him, today he would be castigated in his reviews.  Civil and structural engineers are not big on mathematics.  Anyone who rates people on a Likert Scale is not a nice person. 

This morning I explained to someone, somewhere in the world, that their 1948 bridge was designed by a team that knew their math, as good as this lot and Neville, but as they had done it by hand they made two small design mistakes, if you want to bridge to last 200 years, for their 50 yr design life it is fine, that problem I found with 3 months work on FEM.  I had to tell them it was fixable but tough. 

I also stumbled onto this picture this morning from the programming guy with the XYZ address who is very funny, sometimes. 

circuit_diagram.png

We all live in the arena and pray to the gods, personally I prefer Thor, because it is past midweek, to reach the holy water.  

The only mistake is the 3 liters,  if you use Webster's abomination then center becomes centeral. 

PS You are all not wimps, so you would leave in the link and blow everything up with the 240 Volts. 

 

andrew_4619
榮譽貢獻者 III
2,598 檢視

The spam is dead (removed)! Yes my comment is out of context now.  I was in the shops earlier and bought some Spam! I like Spam but not spam.

AlHill
超級使用者
2,594 檢視

I got an e-mail saying 'At Google Earth, we can even read maps backwards', and I thought... 'That's just spam...'

 

Note:  I am not the original source of the above.

 

Doc (not an Intel employee or contractor)
[Maybe Windows 12 will be better]

Steve_Lionel
榮譽貢獻者 III
2,582 檢視

The lesson here is to not comment on spam - just report it.

AlHill
超級使用者
2,563 檢視

I think a spam discussion is preferrable to a John Nichols story (which could be considered spam).<G>

 

Doc (not an Intel employee or contractor)
[Maybe Windows 12 will be better]

JohnNichols
傑出貢獻者 III
2,537 檢視

My Dearest Al:

I laughed at your comment this morning, thank you.  It made me think of young children playing marbles in a dusty school yard, so I sought a picture of same and found

children-playing-marbles-in-residential-street-AWNBXE-1292064730.jpg

Which was annotated on the WWW as young boys playing marbles, which immediately made me again think of you and smile.  

I am sure you can spot the humour in the annotation.  The tag with the file is correct, but the WWW page is not. 

After all this is just a big game of grownup marbles. 

I have asked the pleasant ladies from Intel on this site if they wish me to avoid stories and the two returns I have had so no.  

But there is one important question,  why is a dusty playground better than tar field?  Excellent question to model in Fortran. 

You realize that our banter will end with the death of one of us, sorry for the survivor. 

JohnNichols_0-1702134150524.jpeg

 

Warm regards from a former Father XMAS.  

Back to the Fortran. 

John

 

 

Brocklenser
初學者
2,543 檢視

It's a complex process. Firstly, the loop generates millions of matrices, and I must choose the one to post. Secondly, understanding how to print the matrix requires some time.

eliopoulos
新手
2,520 檢視

I think I found the guilty. It's the Visual Studio. I had exactly the same output from two different systems. All they have in common is the Visual Studio version (2019 Community). When we used different Visual studio versions, namely 2019 Professional and 2022 preview, the output was different.

Steve_Lionel
榮譽貢獻者 III
2,485 檢視

More precisely, a different Visual C++ math library.

JohnNichols
傑出貢獻者 III
2,427 檢視

I have the code and I have the two different VS Studios,  how to I find the differences between the C++ libraries?

This code appears to be a laminated product that is degrading with time, it is a very difficult problem to start with.  

 

Steve_Lionel
榮譽貢獻者 III
2,407 檢視

You can't reasonably "find the differences", other than to demonstrate that there is a difference. If as I suggested earlier, you instrument the code and see which operation causes things to diverge, that can give you a better clue. This assumes, of course, that the VS difference is the ONLY difference, which is unlikely. A different CPU can cause math library differences.

Brocklenser
初學者
2,242 檢視

It's a complex process. Firstly, the loop generates millions of matrices, and I must choose the one to post. Secondly, understanding how to print the matrix requires some time.

suncropgruop
初學者
2,123 檢視

Dr. Eliopoulos.

The discrepancies in the results of your FORTRAN program when running on different operating systems, despite using the same compiler and executable, can be attributed to a variety of factors. These include potential differences in hardware architecture, which might cause slight variations in floating-point arithmetic, or nuances in the operating systems themselves that could affect program execution. It's also important to consider environmental variables and system settings, as well as any differences in the runtime libraries associated with the compiler on each OS. To address this issue, you should first ensure that your program is free from non-deterministic elements like uninitialized variables. Investigate the sensitivity of your program to floating-point rounding differences, and consider using compiler options that enforce floating-point consistency. Running the program with a controlled dataset can help isolate the problem, and implementing detailed logging might reveal where the results start to diverge. Additionally, review the compiler options and check for any dependencies on external libraries that may behave differently across systems. Consulting Intel's documentation or user forums, and possibly reaching out to their support, could also provide more specific guidance tailored to your situation. A systematic approach, starting with simple test cases and gradually moving to the full complexity of your application, is often crucial in resolving such complex issues.

AlHill
超級使用者
2,111 檢視

@suncropgruop    "Consulting Intel's documentation or user forums, and possibly reaching out to their support, could also provide more specific guidance tailored to your situation. "

 

Is that not what the OP did by posting in this forum, or do you need ChatGPT to answer that for you as well?

 

Doc (not an Intel employee or contractor)
[Maybe Windows 12 will be better]

回覆