If your development cost per piece exceeds the finished product cost per piece, then you're in an ultra-low volume area.

Generally speaking, if you increase the hardware cost as you might by using an HLL rather than staying within the resources specified by the system designer, you'll be fired, and probably should take up something at which you're competent, like selling shoes. That doesn't happen with most development efforts, however.

The key is in careful analysis and design, including proper systems engineering and specification before the hardware and firmware are designed and coding begins. It's up to the coder to choose his tools, though they're often chosen from among what his employer can make available, but the coder has to be sure he can make his work product fit the application requirements and the limitations imposed by hardware.

Sadly, analysis seldom precedes design, and design seldom precedes coding. Coding often precedes understanding of the system requirements. That's why cellphones are marginal, and most PC peripherals work "most of the time." It's also why "software engineering" is so commonly described as an oxymoron, as there's seldom any engineering discipline applied to software/firmware generation.

I disagree with your assertion that it's more important to get something, even if it's a piece of crap, to market rather than to do it right.

All you have to do is to look at the world-wide cellular telephone system. I've been using the telephone system for well over 50 years, yet I've never had a "dropped call." With cellular service, it's unusual to go even a few minutes without losing a syllable or two, and I don't believe I've heard of anyone going a whole day without losing a call. This is because the system was inadequately designed, poorly implemented, and has become so thoroughly entrenched so it will never be "fixed." To this they added "features" that were equally poorly designed and implemented, so they can charge more per phone. Though they cost about $15 in the stores, they're only worth about $0.25, when one considers the quality.

If it weren't for the fact that so many youngsters would rather have what's "kewl" rather than what works, they'd have been gone long ago, in favor of something that actually works.

The language that's used only has adverse impact on the product cost if (a) it causes the object code to exceed the design-size or (b) if it causes the object code to execute so slowly that it fails to meet specified limits.

My assertion is that, while HLL's sometimes make it possible to meet properly engineered specifications, ASM always makes it possible. If that's not the case, then it's the systems engineering that's at fault, and not the language choice.

RE