- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
We are observing very poor performance in our floating point intensive commercial engineering application after upgrading from the latest classic compiler version (19.2) to oneAPI 2024 for both x86 and x64 platforms in the Windows environment. As a matter of fact, the performance is much worse than code generated by Visual C++ 2022. In the past, the classic compiler application slightly outperformed the Microsoft code, but the 2024 compiler code is much slower. In fact, it is on the order of clang 18 in performance (both much slower than VC++). The application only uses the native processor (no GPU).
Anyone else observing this or have ideas how to solve it? We are contemplating dropping the Intel compiler because the builds are much slower than VC++ and the application is much slower as well. With the classic compiler, we could take the slower build times because the application performance was better or similar.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Yes, it is known that ICX can underperform ICC - often by large margins. I've had some pretty bad experiences as well that are still ongoing.
But from my experience, if you're willing to take the time to isolate some examples and submit them to Intel they may be willing to diagnose the problem and fix it in a future version of ICX.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The application is very large. That will take considerable effort. Not certain it's worth the effort when VC++ is starting to give comparable performance out of the box without this effort. To me, this compiler is of beta quality at this point if they are selling it as generating the best Intel code of all compilers. We waited as long as we could to make the conversion from the classic compiler because we were afraid this may happen as we have had a long history with Intel compilers.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Same situation with icx on a Linux Debian GNU/Linux 12 (bookworm) x86_64 server! What's the reason for migrating to a compiler that's a nightmare to use with poorer performance? Pragmas unusable, vectorisation and parallelization inefficient (apparently assumed by the developers of Intel). The answers to the countless questions asked by users about these problems are never precisely answered, why?
How can I return to the latest version of icc , even if you no longer maintain it (and we wonder why?)...
Thank you for your reply.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi, may I ask if you would mind sharing some test code which can reproduce the issue you mentioned in your post ?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I have nothing simple I can supply. It's a very large application, but it doesn't look like it is isolated to me.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
And part of the problem is that the compiler crashes so much in both the latest 2024 and 2025 releases that we can no longer build the application. We are mostly moving away from using it since the Microsoft compiler usually gives better application performance (and builds much faster). clang does not crash building the same application. In the earlier releases where we could actually build the application, the application performance was similar to native clang.
We are still using MKL, but I hope it is not being built with the poor performing compiler.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
As you are still using the MKL, i guess you can benchmark it (post VS pre one API) and know if there was a drop of performance, right ? (Personally I did not remark such a drop. But I did, of course, remark a huge drop of performance of the compiler ...)
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Are you saying that your company's compiler team is not even aware of such a blatant shortcoming of its brand new compiler ? Or that perhaps it is but that it doesn't already have valid use cases to come up with the crucial improvements needed, more than one year after that compiler's release? I'm in the quantitative finance industry, where there was a time you mattered (75% mkl and 25% compiler). People are turning away of you. And we aren't the only ones, according to intel's share . Wake up.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
the Intel Corporation (NASDAQ:INTC) share price dropped 59% over five years. That is extremely sub-optimal, to say the least. We also note that the stock has performed poorly over the last year, with the share price down 32%. Even worse, it's down 13% in about a month, which isn't fun at all.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
BTW, what do mean with "apparently assumed by the developers of Intel" regarding "vectorisation and parallelization inefficient" ?
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page