No, you won't learn anything more from the ASM than from the HLL.
It "doesn't matter" if performance isn't an issue. Just like choice of a linear vs switching regulator might not matter if efficiency and thermal issues aren't an issue.
Note that most code does NOT run on/in a desktop. *And* that portable devices are becoming increasingly more complex and capable (e.g., you wouldn't think of cache -- beyond the tiny prefetch buffer -- in a handheld device 10 years ago; but, I suspect a good many devices today incorporate a fair bit of cache and *rely* on it for performance).
The same is true (to a lesser degree) with VM, XIP, etc.
And, power management issues further suggest a role for this level of familiarity with these sorts of techniques.
No, if the object won't fit in a cache-line (or VM page), you will see a difference in performance. An object 10 times larger will see a proportionately larger difference. Etc.
And the *chemist* has that sort of intimate knowledge of the differences? Counted strings vs. ASCII-Z strings? The differences between Call by Name vs. Call by Value semantics, etc.?
See above. I use it, for example, to run code out of serial flash devices. The speedup (with resource constraints) is
*huge*.
And exactly where do the physics/chemistry majors have those classes in *their* curriculae?
The portion of my original comment that preceded this example (which you neglected to include) stated: 'A (good) "software" curriculum (avoiding the use of "CS") exposes people to issues that are extremely DIFFICULT TO STUMBLE ONTO IN A CASUAL (i.e., taking whichever jobs come your way) CAREER.' [emphasis mine] The "chemist" has to rely on someone more knowledgeable of the technology to bring this to his/her attention. Perhaps a physicist?
How many successfully master multithreaded programming techniques?
How many are aware of the different techniques (and shortcomings) for synchronizing competing actors?
How many can exploit a multiprocessor environment? How efficiently do they learn these techniques (i.e., what is it costing their employer for this OJT -- in terms of increased development times and/or reduced reliability)?
[I chuckle thinking of all the shows Weir did "trying" to learn how to play slide guitar. Who bore the cost of that learning experience?]