minipci1321

Another one:

> the exact order that C++ evaluates expressions is extremely complicated and not always defined, ... "

"(Un)defined" and "unspecified" (which order of evaluation of fun-args really is) are two distinct terms in the C++ spec meaning different things, they cannot be used interchangeably.

But I guess it's just yet another thing of "Pedantic assembly-code analysts"...

> NATO is happy to have war against Russia or Russian allies to the last Ukrainian solder and civilian.

I think this devalues your all other arguments. Where it comes from?

I'd guess it looks under-appreciated on the surface, because it is hard to run a business on such a premise.

Admitting it "officially" ultimately boils down to retaining by all means the engineers who already acquired the knowledge. Some of those will ask for more benefits, others will play games of power, others still will simply refuse to stay once they stopped learning.

(In my industry and geo) businesses chose the opposite -- they slice and dice the knowledge into tiny bins so that a total newcomer can learn up quickly enough. This drives down wages but now requires many more people where a few should suffice.

Solution to that? Total subcontracting of sw development. Which is of course a slow death for the business, but a) at least it looks like a business model, b) death is slow enough so that managers have time to move on, c) this has a merit of being implementable. The alternative -- finding people who already know or learn very fast, and then keeping them -- is not for every budget.

lawmakers != government

I used to work on drivers and firmware for consumer electronics remote controls, and the empirical threshold we used at the time, was that an average human cannot press a button two times in a row much faster than with a 200 ms delay. (So, if the driver was getting same key presses spaced by less than that, it could safely assume those come from auto-repeat in the RC.)

Conveniently omits any mention of hardware initialization and related fallouts in HVM.

Find it easy? Try to push all components -- interconnect, DDR and flash controllers, accelerators, etc -- to max performance and get the phone ready to hear from the factory about how the boards fail testing.

Not all of the SoC blocks may be reconfigured later after the boot completed, some have to be done right in the very early stages of it.

Didn't seem to have helped much in Belarus. OSCE/ODIHR were present since 2001 (at least), noted some things, and here we are:

https://www.statista.com/chart/22512/belarus-presidential-el...

Then they were absent this year, and the official result was the exact same 80%.

From official source:

"The IT sector in Belarus. The total production and sales revenue of the IT sector was $3.1 billion in 2018. In 2018 IT accounted for 5.7% of Belarus's GDP, which is comparable with agriculture and forestry (6.4%), construction (5.4%), and transport (5.8%). By 2022, IT's share of GDP will have increased to 10%."

URL: https://minskdialogue.by/en/research/memorable-notes/how-can...

> It might be better for Intel to split into two.

How exactly would it be better -- again, given the volumes?

-- The "design" split-off would be instantly lacking modern-node fab capacities (TSMC already said they wont scale up for Intel),

-- the manufacturing part would find itself in competition with those already having better node tech ==> not enough orders to fill the pipe (customers with comparable volumes are already all set WRT fabbing).

... and the CPU has no direct support of in memory stack needed when calling a subroutine. Rumor had it, the guy responsible for this decision (to not include it), was later mutated to the IBM's internal equivalent of Siberia

Great list.

+ When stating the issue, avoid using "You are/were", prefer "You did/didn't do".

EDIT: and especially, absolutely avoid saying "You will never"

IBM S/370 Macro Assembler also was very feature-rich. I needed to have modules be loaded into the kernel in a certain order, so I implemented sorting of symbolic names in macro-asm (had to use loops at the macro level if I recall), and then assembly code would be emitted clean, already in the right sequence.

DCB was the same magic incantation as #include <stdio.h>

I was 16 and learning FORTRAN at school. I was a lazy bum at that time (well, that stayed ;-), didn't really prepare, and at the final exam, the teacher asked me in what order the program is executed. Not knowing any better I replied "a value is calculated when the rest of the code needs it" (like in regular math that we did a lot at the time too). She said "not correct, it executes in the order it is written".

-- I doubt she knew anything about functional programming back then.

-- It took me 40+ years to get to the universe where this answer was not completely laughable.

-- And meanwhile what 16 y.o. kids were doing at the time, 6 y.o ones are capable of doing now.

But aren't some choices, while arguably no more safe, more ethical than others? If you know that your action will bring certain death to some category of people, shouldn't you choose to refrain from doing it?

The OP, dated Apr 4th, sounds like it is from 3 months ago. We learned a lot more since then about how reasonable people are in their behaviors and how inclined they are to understand sensible explanations.

> To me, this implies a further disadvantage: extremely high latency when compared with in-person collaboration.

This is not an implication by all means.

Low-skilled not very well incentivized junior team (let's call it so in absence of better terms) needs more/most "in-person collaboration". Team of experts to whom the goals and the overall vision have been conveyed, who knows how to put it to practice, will give the shortest latency in async flows.

Don't take it as a personal attack, but people valuing high live discussions, are frequently those on the receiving end of it.

> We need to make it cool to work in a semiconductor company.

I am not optimistic. "Semiconductor company" ==> high capitalization (expensive processes) ==> fertile ground for lawsuits and attacks.

As long as a slightest mistake has a potential to turn into multi-million-loss lawsuits, it will never be cool. Is it easy to be cool walking a tightrope between two skyscrapers? Not for many I'd guess.

BTW it is the same for non-semi companies (like OEMs), except that volumes of a given single product are generally lower, so mistakes are more bounded.

IANAL, but you seem to be trying to apply law as a scientist. Has anyone complained (so massively that it has become noticed) that their tablets or laptops run slower after some time?

(BTW, laptops would report bad-battery condition usually, this is how you know you need a replacement, which is an easy operation with them.)

Of course you don't have to. But, to be honest, waking-up at 7am spells "no commute" to me.

Because when one works from home, there is no need to allot time for commute. People with short commute times (==> no need to adjust sleeping schedule), are much keener to go work from the office.

Remote work is a skill, it takes time to acquire, just like office work and barely less. (Have you forgotten how hard it was to stay put for 8+ hours in the same place? I still remember.)

My personal estimations -- it takes 18 month to work out an established routine, so any RW "decreed" on the worker before that time, is indeed effectively vacations.

The worst scheme is working from home 1-2 days a week, and coming to office for the rest. Beside the established routine it also requires adapting the sleep schedule (which is twice as hard if the RW days are floating, not fixed).

> Governments in some countries tried limiting imports from that country, and they have been punished by the EU.

Do you have any references to the above two affirmations?

Not direct answer to your question, but Berstel's publications:

http://www-igm.univ-mlv.fr/~berstel/PubsJeanBerstel.html

You can try to contact him, he might be able to help you.

(This missed my question. Once the compiler is updated to the uArch change, we don't rebuild all binaries out there that were built by the old versions, and their speed is possibly compromised by the uArch change. Should we then even care about emitting binary code optimized for a given uArch, if we can guarantee it is stable in time?)

Your logic would also apply to the compiler back-end to the same measure. Should it emit optimized code then?

> What I want is for it to have one cable type

Cables have conflicting requirements:

-- high current requires thick wires and heat dissipation (if long cable). Possibly other safety-related requirements

-- high frequency wants more wire surface because of skin effect

-- flexibility and cost don't like either of the above.

I personally prefer to have a choice of a thin flexible cable when situation allows, and am ready to pay for it.

To be fair to the modern BIOS, there is a whole lot more hardware to initialize than in 80s and 90s, the added ROM space isn't consumed only by bloated compiler output. Some motherboards store a second pristine copy of the BIOS image for fail-safe upgrades.

> and elegantly clever Asm

I personally miss returning boolean result in C (carry) flag.

Also, while asking people that sort of question, ask yourself whether improving the situation is within your reach, whatever is the cause? If you think you could -- do you think people you are interviewing, also believe you can?

I wouldn't tell anything to a person who has no power or no incentive to improve things, and sadly, this covers 100% of people one usually has exit interviews with. So I always declined them, politely.

You wouldn't believe how much respect and good contacts this attitude earned me over the years.

Like this one for example?

https://www.fda.gov/drugs/drug-safety-and-availability/fda-u...

Kolmogorov complexity?

The title should have been "I've picked getting unjustfully condemned people out of prison, over my kids". It would make 2/3 of the comments irrelevant.

Prioritizing a more "normal" job over raising own kids is a very different subject.