JohnnyCode said:
Such an incorrect statement, of course knowing ASM gives you ability to optimize compiled binary in lots/most of cases!

Nope, thats total horseshit. Compilers are better at producing ASM than any human in 99.9% of all cases PERIOD. Its not 1990 anymore. I'm sick of this nonesense being being spewed every now and then in this forum. I'm still waiting for anyone to show some real-world examples of ASM performing better in a common-real world example (which shouldn't be hard to come by if its better in “MOST” of cases).

If on the other hand you don't want to base your knowledge on makebelief and antiquated facts, here's a few talks that prove quite the opposite:

What has my compiler done for me lately
(from the guy who runs compiler-explorer https://godbolt.org/, which lets you inspect the generated assembly across multiple different compilers easily)
Speed is in the mind of people
(you can take pretty much any talk by alexandrescu. that guy is a beast. he's optimizing algorithms that were thought to be unbeatable left and right, and do you know how much assembly his optimization contains? Zero!)

@Juliean If we get stuck in the end of Moore's Law. And there are no quantum computers yet, and not faster silicon either, ASM will become a pop-star again. Then, you will have to update or be deprecated.

https://www.theguardian.com/commentisfree/2020/jan/11/we-are-approaching-the-limits-of-computer-power-we-need-new-programmers-n-ow

Automatization will always help you as a coder regardless the language you use. But if we get stuck in the end of Moore's Law you will have to do most of your code in ASM. If you are afraid, of ASM, you should not. ASM is more writing(sometimes), but at the same time, it gives you such a good understanding of what are you doing, that it is easier to develop. You don't spend 10 hours googling in internet in order to understand how a new syntax coming from C++ works. You just code it yourself. Sometimes you read about something new fancy extra somebody introduced in HLLs, and you find out that you already did that in ASM. It is just natural. No more “C++ lacks this. C can not do that. How i fake it?”. You just do it.

And you can copy paste in ASM too!!!!

Doubt it. ASM pop stars will all work on better compilers, tuned to each hardware, respecting all the hardware limits, strengths and weaknesses, bugs, pipelining abilities etc. a human programmer can't and shouldn't need to know.
Modern software is too complex to be hold back with boring implementation details. We would never get done with anything if we had to care about register allocation or instruction latencies all the time.

If we get back to ASM programming, this would not show how we care about performance - it would only proof our inability and inefficiency.
Not going to happen. Because we are geniuses, pulling AGI right now out of our pockets and replacing ourselves with superior machinery which will spread over milky way in no time. And timetravel too. So why should we fail on improving compilers further.

JoeJ said:
it would only proof our inability and inefficiency.

In a way this is the reason. We will try to pull competitive advantage from other places than the stuck hardware.

I wonder why people think one can not use OOP and Functions in ASM?
One DO use libraries and OOP and Functions and all in ASM. You call the OS from inside a program to do lot of things for you. Once you get used to it, it is great!

Right now it is not needed, because for the boost it gives us, it is not worth the time. But if Moore's Law is completely stuck, companies will start to hire ASM programmers to can keep competing on the market.

@Calin
“While I dissagre with how you chose to treat people, on some issues you`re right on the nose.”

or

“While I dissagre with how you chose to treat people on some issues, you`re right on the nose.”

?

NikiTo said:
Once you get used to it, it is great!

No. I assure you i could not finish my current work if i had to use ASM again.

And most importantly: ASM does NOT give me an advantage on optimization at all. I try to reduce the time complexity of algorithms to get speed ups of orders of magnitude. Low level gives me only a factor of two. Nice but makes no real difference during 99% of dev time.
Optimizing for time complexity means i need a lot of flexibility to do big changes in short time. That's not possible with ASM.

The other point is that modern CPUs and compilers have both grown in complexity. Compilers are not perfect, but ASM only makes sense in case they totally fail at something, which they do but ‘hopefully will be fixed soon’.

What were your reasons to abandon ASM years ago? I think you mentioned you did, in other posts, right?

Moore's Law is alive and kicking. Wasn't that stated just recently by one of the hardware eggheads from intel or amd ? When i read about the massive increase in for instance parallel computing power just in the last 2 years or look at the development of the supercomputers i must say yep, nobody ist stuck right now. It may even be that in contrary competition has come alive again.

So, maybe the curve has flattened a bit since 2010 or so (i doubt it), but the claim that hardware is stuck is not correct, or at least heavily in need of proof.