topic Re: oneAPI minor release version number in Intel® Fortran Compiler

oneAPI minor release version number

jimdempseyatthecove — Mon, 21 Feb 2022 18:26:04 GMT

On the oneAPI download pages (basic, HPC, ...)

The verson number is listed as (for example) 2022.1.2

However, what is downloaded is 2022.1.2.nnn

Where nnn is a minor version number

How do we know when (if) the minor release (nnn) gets updated (without downloading it)?

Jim Dempsey

Re: oneAPI minor release version number

Steve_Lionel — Mon, 21 Feb 2022 22:56:32 GMT

Typically, the nnn is a "build number" and only one release per version is offered. So, after 2022.1.2 would be 2022.1.3 or 2022.2.0. Then the compiler version may be completely different (as it is now.)

Re: oneAPI minor release version number

Ron_Green — Tue, 22 Feb 2022 16:53:32 GMT

Steve is right, versioning is <major>.<minor>.<patch> as the primary version customers need to be concerned about. That is what we release, we won't "patch" a .<patch> with a release.

If you are curious ... this is unnecessary detail but ...

BEHIND THE CURTAINS as we prepare a release for a <major>.<minor>.<patch> release, we may INTERNALLY have multiple builds. For example right now we're preparing for the 2022.2.0 release. Internally we'll have an early "Preview" build with a build number, Let's say

2022.2.0.<nnn>. We test that build, find problems, fix those, build 2022.2.0.<nnn+1>. Then daily we continue to build/test so the builds increase nnn+2, nnn+3, etc. until we finally get everything through testing and certify the entire oneAPI Tookit package for customer release. At that point the release 2022.2.0.<nnn+whatever> package is staged to the IRC and repos and container repos. Then on the planned release date we "Publish" or make public this package.

There's a lot more ugly details in component versioning that confuses things further - like the Classic compilers still on the 2021.X branch whilst IFX/ICX are on a 2022.X branch ( separate git code branches, separate versions therefore).

The takeaway is what Steve said - you can ignore the build number, it's an internal artifact of our build/test process.

Re: oneAPI minor release version number

JohnNichols — Tue, 22 Feb 2022 18:46:17 GMT

You see I think something similar in the Windows Preview Builds, the numbers are not sequential in steps of unity, but appear somewhat randomly increasing. Interestingly very occasionally you get a midpoint release that looks like 22547.1000 that appears to be fixing a bug they found on release.

Re: oneAPI minor release version number

Steve_Lionel — Tue, 22 Feb 2022 22:04:31 GMT

You'll find that this general scheme is near-universal in software products.

Re: oneAPI minor release version number

JohnNichols — Tue, 22 Feb 2022 23:07:13 GMT

If you randomly selected two programmers from the set of all programmers, they would I believe agree on nothing except that reading SCI FI is a good way to relax.

Re: oneAPI minor release version number

AlHill — Tue, 22 Feb 2022 23:15:17 GMT

No, one would say that Star Trek is sci-fi and the other would say that Lost In Space is sci-fi.

Doc (not an Intel employee or contractor)
[Waiting for Windows 12]

Re: oneAPI minor release version number

JohnNichols — Wed, 23 Feb 2022 18:58:31 GMT

I missed you, where have you been.

There was a major article in MAX PC about Windows 11 being slow, it appears that MS managed to overcome the problem.

Neither are real programmers, as the real answer is Heinlein or Asimov.

I wonder if they are running it on Fortran. I have never looked into war models, I assume they are discussed somewhere. I assume they are running on Crays. Can you run Intel Fortran on a CRAY?

[no politics please]

If you look long term, the simple reason that the USA needs Ukraine to stay intact is NVIDIA and INTEL. Mainly NVIDIA.

Re: oneAPI minor release version number

Ron_Green — Wed, 23 Feb 2022 16:53:57 GMT

John, you asked "Can you run Intel Fortran on a CRAY?" Absolutely. Cray has their own Fortran compiler and it's quite good. Crays FTN is part of their default Cray Programming Environment (Cray PE), which includes compilers, libraries, MPI (MPICH based), performance analyzers. Cray will also include Intel Fortran or all of oneAPI if a customer asks for it, and integrate it with the Cray PE. So you can load ifort/ifx with a 'module load' command, and they have a Cray MPI wrapper that can use Intel Fortran instead of Cray Fortran.

It's my experience that the 2 compilers are fairly close in performance on Cray systems with Intel processors. It varies by application on which compiler creates "faster" code. But since you can't run Cray Fortran anywhere but on a Cray, a lot of users use Intel since they can do dev work on workstations and laptops and save cycles on the big iron. I haven't used a Cray with AMD processors so I can't comment on performance for those platforms.