Performance difference Intel Compiler Classic vs Intel One Api Compiler

[GunnarB]

This is about the performance of the compiled program, not the compilation time.

I am running a program on a NP-hard problem with billions of iterations.

I am using the Intel Compiler Classic 19.3 as the generated output performs 30% faster than any other compiler (VS, Clang, Intel One Api, Intel Classic 2022).

Since VS Studio 17.4 does not compile I tried switching to One Api, but am a bit disappointed on the performance.

I would expect subtle differences but >30% is a lot.

The code in question is pure C.

I am seeking some help how I can analyze this in more detail. In brackets my personal benchmark per minute / threaded 12 CPUs (6 of those real).

The log shows me these compiler options for Classic 19.2: (135 billion calculations)

/c /Qm64 /IC:\Development\C++\FastMaster /Zi /nologo /W3 /O2 /Oi /Qipo /D __INTEL_COMPILER=2021 /D NDEBUG /D _CONSOLE /D _UNICODE /D UNICODE /EHsc /MD /GS /Gy /permissive- /Zc:wchar_t /Zc:forScope /Zc:inline /std:c++14 /Fo"X64\RELEASE\\" /Fd"X64\RELEASE\VC143.PDB"

For the Intel Api: (90 billion calculations)

/c /Qm64 /IC:\Development\C++\FastMaster /Zi /nologo /W3 /O2 /Oi /Qipo /D __INTEL_LLVM_COMPILER=20220200 /D NDEBUG /D _CONSOLE /D _UNICODE /D UNICODE /EHsc /MD /GS /Gy /Zc:forScope /Fo"X64\RELEASE\\" /TC

I also tried these settings for Intel Api: (64 billion calculations)

/c /Qm64 /IC:\Development\C++\FastMaster /Zi /nologo /W3 /O3 /Oi /Ot /Qipo /D __INTEL_LLVM_COMPILER=20220200 /D NDEBUG /D _CONSOLE /D _UNICODE /D UNICODE /EHsc /MD /GS- /Gy /fp:fast /Qxcoffeelake /Zc:forScope /Fo"X64\RELEASE\\" /TC

 Where Qxcoffeelake really puts the brakes on. Everything else no measurable difference.

Best

Gunnar

[SeshaP_Intel]

Hi,

Thank you for posting in Intel Communities.

Could you please provide the following details you were using so that we can reproduce the issue on our end?

1. Compiler versions you are using.

2. Operating system you are using.

3. Hardware details.

4. Complete reproducer code with the steps you have followed.

Thanks and Regards,

Pendyala Sesha Srinivas

[SeshaP_Intel]

Hi,

We haven't heard back from you. Could you please provide an update on your issue?

Thanks and Regards,

Pendyala Sesha Srinivas

[GunnarB]

Hi,

sorry for not coming back on this yet.

My code is rather classified. I will have to create a version without the classified part to share with you.

I understand that a 30% speed increase would be of high interest for you and that you need code to verify the compiler optimizations.

This has to do with compiler optimization, other parts of the code run 50% faster compiled with the One Api compiler, unfortunately in less critical parts.

Bear with me, I will provide the code in time.

Best,

Gunnar

[SeshaP_Intel]

Hi,

Could you please provide an update on your issue?

Thanks and Regards,

Pendyala Sesha Srinivas

[SeshaP_Intel]

Hi,

We have not heard back from you. This thread will no longer be monitored by Intel. If you need further assistance, please post a new question. 

Thanks and Regards,

Pendyala Sesha Srinivas