A comment for a 10-year-old article! That speaks for a long-term influence.
Quite recently, (last couple of years, I think), there was an advertisement near here for a PDP-11 Assembly programmer. I checked to see that it hadn't slipped through some wormhole, and it was quite genuine. A nuclear power plant wanted someone to maintain their control software. The plant was certified with a PDP-11 in charge, running code written in Assembly, so that's what they had to use.
Aside from the question of how do you maintain the hardware, (do they have a huge cache of parts and spare processors?), all sorts of issues come up. How do you avoid terminal boredom and stay current with the language? (Skills decay if you don't exercise them.) The opportunity to make changes to a functioning plant must be quite limited. Random experimentation isn't an option.
There's also the matter of career cul-de-sacs. A gig like that on the resume' is going to be a total career-ending repellent, so the incumbent would need a contract guaranteeing a lifetime income, (barring episodes of mass homicide in the workplace).
Seriously, that's a big issue for anyone thinking of admitting to Assembly eperience. It's not interpreted as "This dude is capable of some really skilled, detailed work, so should be adaptable", it's "Limited vision, doesn't know $FLAVOUR_OF_MONTH, pass".
While waiting for an IBM Metal C compile job to finish the compile of several source parts which target 3 different ARCH levels (soon to be 4) on IBM System z hardware, got urge to Google ASSEMBLER code.
Not sure why the ASSEMBLER Google search started, assume must have wanted something to fill the time while waiting for the IBM Metal C compiler to finish.
The goal of the IBM Metal C compiler is to generate ASSEMBLER code which targets the new System z hardware by using hardware instructions new to the ARCH level and optimizing the code based on the targeted platform.
The developers here often drop into ASSEMBLER code either when the compiler does not support a new instruction, or as this article points out the compiler just does not generate the code well enough.
One of the technology leaders at Rocket raves about using the Metal C compiler as it allows the code to take advantage of hardware advancements as soon as they are avilable to the market.
Ten years later and yet another spin on why there are advantages to Assembler, the IBM Metal C seems to cross the boundary with it's hybrid design.
My opinion is that coding everything once done in ASSEMBLER only is now best done with a Compiler which allows you to take advantage of newer faster hardware instructions without the need to rewrite pure ASSEMBLER routines.
Of course any pure ASSEMBLER routines used will need to be reviewed at each new ARCH level, but it is obvious the advantages this bit of extra effort provides for System Software Solutions.
My Compile, ASM and Link are done - so back to work!