Ive heard this for thirty years and it is as wrong now as it was then.
Yes you need to be good at assembly (especially data oriented architecture), yes it takes forever, but that is no excuse to spew falsehoods.
Ive heard this for thirty years and it is as wrong now as it was then.
Yes you need to be good at assembly (especially data oriented architecture), yes it takes forever, but that is no excuse to spew falsehoods.
correct, understanding the instruction set architecture you are working with is required for reasoning about the performance of a given algorithm in detail.
you will likely not be writing a lot of assembly by hand, however steering the compiler codegen in the right direction requires an understanding of what the compiler produces.
even outside of enhancing performance, knowledge of instruction sets is instrumental for security research and reverse engineering. for some fun but practical demonstrations, see work by Nathan Baggs on YouTube - it involves staring at a lot of disassembly.
i don't know where this misguided notion that assembly language is "1975" comes from. it's not like Cobol where a few large but important systems keep it alive. this is something that lies at the core of every interaction with computers that you have daily.
What? Did I just give you an example that it is true?
Natural language is the highest of course. We will return to assembly programming eventually, because those intermediate "highlevel" languages are not needed anymore between you and machine.