Serious optimizer bug in current Intel Fortran compilers

[Christofer]

I think that there is a serious optimizer bug in the Intel Fortran compiler, the bug is related to the moving line of code without any of the semantic alteration or inserting the statements to find out the bugs. Because of this bugs are not appearing. It will be right to say that to work in this way is really difficult, since it is not possible to find the bug in code. So I just need a solution for this serious optimizer bug in the Fortran compiler.

[Algonkin]

I totally agree with this, I have also seen this bug in Intel’s current Fortan compilers. I will like to share my experience with this kind of bug, normally this thing happens when you are running Fortan with optimizer and if we are disabling the optimizer everything works fine. So if you are having problem with this then the only solution for this problem is to disable the optimizer.

[Illinois]

It is true that there is bug with the Fortan compiler but it can be cured and there is no need to disable the complier that you are using. You just need to download the new compiler that is available on Intel’s official site. Go to Intel Fortran compiler, and do not forget to provide your user name password there for getting the latest compiler. In short you just need to log in with the premier Support Loing ID and Password and the you have to select the “Product Downloads�.

[Iroquois]

The priorities do not go from first to last, but very dependent user. So when I put the performance in front of all, why should not be the first benchmark. For example I normally affects the speed of filling it directly affects my productivity as a programmer, though, when everything works, I get a faster executable and / or smaller size to build final, I'm willing to lose a bit ' more time.

[Bangari]

According to me the size becomes important depending on the target chosen for distribution. Cannot be a problem if the vehicle used was a disc Blu-ray, for example, because the space that you can use is measured in the tens of GB (but you can saturate even that). Becomes critical if the product you must download, instead, on a mobile phone via WAP, which can be much slower, free or paid.

[Kshaunish]

Support multiple architectures / platforms may be useless, value added, or a prerequisite. Again, needless to say, it all depends on what we want to achieve. For example, if I want to make a PC game, will probably not m'interesserÃ* this feature, because usually develops the architecture for Windows and x86 (which dominates) and / or AMD64 (which represents the near future). If, however, my target is represented by mobile phones, at least it will be essential to support ARM.

[Cherkasov]

One fifth entry that I mentioned is reliable, but is a characteristic that should not even be in this debate by a compiler, like another very important element which can be a filesystem, it is expected that solid and produce the original source code that makes faith with regard to calculations, and that obviously has problems that do not relate to the same source.

[Hajra]

It is practically impossible to combine all four quality is a classical optimization problem on the related variables that cannot provide the best results for all. Just to give some examples, if I want the best performance it is clear that you must spend more time to optimize the final code, and this inevitably affects the speed of compilation. If, however, interest me include the size of the executable, I'll give up some performance because, for instance, I could not make an aggressive inline function.