GCC Compiler Tests At A Variety Of Optimization Levels Using Clear Linux

[indepe]

...
For my normal machines I guess I'll stay with -O2 wich is said to be stable, while -O3 is suspected to occasionally produce hiccups on complex code.

I said it elsewhere but I'll repeat it here: as far as I could find out so far, the "suspicions" about -O3 are a 10-year old myth and not true anymore. (Except maybe for occasional things the same as for any optimization level), (And perhaps except for some old OS packages in Gentoo, but that isn't really clear since that info is always reported as vague hearsay).

This was recently discussed elsewhere, and I only repeat that here because you are the second person re-iterating this suspicion without actually knowing anything.

On the other hand, I am always interested in any "real" specific information about any actual problems.

[M@yeulC]

What is -O6? Isn't it supposed to be equivalent to -O3?
I've seen it in a some (flawed) benchmarks that proclaimed that Javascript was faster than C, though.

[cj.wijtmans]

I said it elsewhere but I'll repeat it here: as far as I could find out so far, the "suspicions" about -O3 are a 10-year old myth and not true anymore. (Except maybe for occasional things the same as for any optimization level), (And perhaps except for some old OS packages in Gentoo, but that isn't really clear since that info is always reported as vague hearsay).

This was recently discussed elsewhere, and I only repeat that here because you are the second person re-iterating this suspicion without actually knowing anything.

On the other hand, I am always interested in any "real" specific information about any actual problems.

there are several modern packages on gentoo that dont compile with -O3

[indepe]

there are several modern packages on gentoo that dont compile with -O3

What happens?

[duby229]

there are several modern packages on gentoo that dont compile with -O3

There -are- in fact also packages which produce undefined behaviour. But without some mechansim for automatic detection of undefined behaviour it was never worth figuring out.

[indepe]

There -are- in fact also packages which produce undefined behaviour. But without some mechansim for automatic detection of undefined behaviour it was never worth figuring out.

How do you know it is a fact? It sounds like there are no bug reports or anything that you could reference.

[cj.wijtmans]

How do you know it is a fact? It sounds like there are no bug reports or anything that you could reference.

Oh gee perhaps its not a bug with any package but abuse of compiler flags which is generally known.

[duby229]

Because Gentoo's developers never provided a mechanism to do anything about it. I would have to step through the system one package at a time and tst each individual dependency for correct behaviour. It's not worth it. But what I do know for absolutely for sure is if I build the whole system with -O3 it results in a bunch of random strange behaviour, that then goes away if I rebuild it with -O2.

I don't understand why that's hard to understand.

[indepe]

Oh gee perhaps its not a bug with any package but abuse of compiler flags which is generally known.

Is there a reason to keep insisting there are problems, but not provide any meaningful information?

[duby229]

Is there a reason to keep insisting there are problems, but not provide any meaningful information?

Why do you think experience from multiple people isn't meaningful information? I mean really, if you only choose to believe what you experience and not what others experience and then never try to duplicate the other experiences then very few things will ever be meaningful to you.