kev009

It's basically MIPS all over again

The conclusion is honest, and you can of course brute force any ISA into any role. I used to loathe x86 for that reason, but now that I'm older I respect the game.

I've assumed it worked fine on Linux but your comment brought me to this after a quick search: https://bugzilla.kernel.org/show_bug.cgi?id=118721. I've not seen any major issues on Linux e1000e defaults but never pushed it harder than laptop usage. I've occasionally tried to make em(4) TSO work on FreeBSD (it works fine on igb(4)) but gave up due to stalls. Linux has GSO (maybe invented partially due to those issues?) so the benefit is a lot smaller there.

There's a comical change where Intel downclocked DMA to fix some issue and it makes TSO max out at ~700mbps: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin... but they later disable TSO there with https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...

Most realtek is F tier, they basically repeat the bad part of this NE2000 saga, and the simplicity comes at the cost of CPU cycles. In that sense the Intel PCH controllers were great.

In this era 3Com was still the gold standard, but intel eventually dominated the market because they integrated 100mbit and then gigabit on the PCH relegating any margin in the expansion card market to servers. As someone that has worked on a lot of Intel drivers I always find Intel being put on a pedestal to be a little amusing, they have a tendency to make a lot of gaffes in silicon and sometimes these reappear in followup designs. It's certainly not bad, but I would place Intel networking about mid-tier historically and today.

They did have someone from Intel on when covering Intel's switches and he brings up his involvement in the Pro/100.

I wish they had tracked down someone from Nat Semi for this. I imagine they would have been somewhat dismayed by some of this.

One of the issues of the perceived low quality is the massive variance in NE2000 silicon. Basically everyone cloned/emulated the chip, but they have subtle variation, features, bugs. So a generic driver like Linux has to deal with a lot of damage.

Even things built directly on underlying malloc and free typically have some form of "garbage collection" in the malloc implementation for efficiency and performance (geometric sizing, thread caching, etc).

It's best to think about lifetimes and lifecycles where possible. Immutability where sensible and things like pool allocation are examples of this.

GC languages can result in quite pessimistic code because they encourage people to NOT think about what is going on. But people have also built functional HFT engines on things like the JVM by thinking about lifetimes and lifecycles.

Because Go is kind of a Steampunk design. The creators collectively ignored decades of PL developments. What that nets is kind of a C without sharp edges, but can't be used where C can.

Rust is definitely jarring to look at, in the same way that decoding some strange C declaration can be, in ye olde days when you had to float all this context in your mind while doing work. But with modern tooling who cares: "explain this lifetime to me"

This seems like a cope, if you aren't writing the syntax who cares and everything here is even better with a stronger type system like Rust, F#, Scala, TypeScript.

One interesting thing with wider addresses isn't necessarily increasing itself, but features built upon the expansion. CHERI is one example of that.

"If the database size is smaller than the mmap_size, the entire database is mapped into memory, turning disk reads into simple pointer arithmetic." hah so that's how it works?

If it's on the house (and they can probably turn up the token speed) why use anything other than the most powerful models?

The same thing is happening in software, I see the ladder pulled up and don't feel vulnerable as senior staff. If anything, we face a massive and increasing competency crisis in computing because there is a cult dumb enough to believe acceleration and doomer cases for LLMs.

The GPT ones are strange. The $25 fable one to me is subjectively better than the others. The $100 fable one is too literal and robotic.

The jevons paradox is you need auteurs to curate vignettes or effects and cut or mask them in etc. That's not really different philosophically when software entered art in other ways. I could see errors/glitches lowering in time but I doubt there will be much acceleration.

This is silly or just AI slop post? Because using a Go quote as an example of doing something right in the arena is laughable at best where it has the same problems, more magic, and worse observability.

The punchline seems to be something like the LMAX disruptor style which is genuinely good for some things, but if you have I/O loops like the illustration shows you can easily block that loop with some long running function.. so you have the same cognitive load as managing thread pools or async pools or disruptors..

It's chronoception effect, not placebo effect. Standard literature claims humans can readily visually perceive 60-90Hz, and perhaps somewhere 500-1000Hz for ghosting. Note that all this is somewhat tangential to the article which focuses on _input latency_, I think the author is driving the high refresh just to minimize _output latency_ as much as possible from the test.

More like AI psychosis 2040. There is zero chance of the 2031 doomer timeline presented, and I would stake a large sum of money on it.

This sounds somewhat similar to IBM's Centaur used on the POWER9/10/11. Some of the POWER9 hardware people ended up at Meta so might be related.

Those two companies spend all their effort patting themselves on the back about "peer reviewed studies" and posturing immeasurables like security theatre and "trust and safety". It's really not hard to believe that Grok or Chinese models can show there is no moat.

Grok build already punched above its weight and is the nicest TUI, claude and codex are clearly vibecoded by web developers that don't understand systems (eating SSDs, spaghetti logic, extinguishing kernel watch limits, etc). I think Anthropic and OpenAI are both engaged in their own theatre, trying to define and redefine what game they are even playing, trying to shift to immeasurables like safety or security or exclusivity. There's definitely room at the top.

I hear a thread of envy throughout that is basically "why isn't my magic money machine as hyped as their magic money machine"

Sure, humans are attracted to laziness by default and at its best technology must be used to lower toil and suffering (unfortunately there are unintended consequences etc for philosophical debates).

The bigger issue with brain rot is attention spans. Social media has ruined attention spans.

I came into the US tech industry in 2011 from a CS program (programming on my own longer), which was basically a pipeline into Java or web dev at that time (the classes on other things like OS, compilers, and FPGA were a thing, but seen by most as a hurdle not a promising or practical career pathways).

Most of the people I dealt with for the decade following that (i.e. pre-LLM) had little clue how any part of computers worked, and I don't intend that to be particularly pejorative, it's just that it's been somewhat easy to carve out a niche doing something to do with programming while having little understanding of the small or big picture.. all heuristics and applying working patterns.. for a few decades.

You can still understand whatever you want, today, with added tools. It's just a choice to turn your brain on or off. I think LLMs are perfectly fine as a learning tool to interrogate a subject, do comparatives, and then formalize your understanding by reading the sources. At a macro level brain rot is real, but it cuts across all generations and it started long before LLMs.

This hasn't been true for quite a long time, because it's domain of other products for both.

i.e. MySQL/Maria's MM is one of the worse options compared to anything else.. like TiDB or Aurora on AWS for MySQL compat and Yugabyte or CockroachDB for Postgres compat (or Aurora on AWS or AlloyDB on GCP).

Versus letting a singular entity snoop everything? If you actually open a connection to the result what is the difference? The only way to fully deal with all that is an overlay or mixnets.

I've never felt this. Most large services run or delegate to anycast DNS services.

If you have knowledge of TCP, you know you will occasionally get stalls much greater than that beyond control.

Maybe.. and "a few" was pulled out of thin air, but these machines were national treasures with immense R&D expense. Think rockets, heavy aircraft, and lithography.. not commodities or software. Having stuff in these categories is kind of "you have to be this tall to ride the ride"

The history is complicated and a side quest to the conversation here, but a free market did play out in the US and became fever pitched after AT&T was hand tied in the 1980s.

A free market means if you can do something better/cheaper/faster, or at least convince other people of the promise, you have a shot. The more that come (think of a gold rush), the potential returns diminish so it will eventually be hard to either acquire revenue/customers or funding for speculative approaches if they are capital intense.

Also, ironic to the conversation, the "best" machine vis a vis when and what, came out of the monopoly: the 5ESS. The Bell System was a reflection of a different culture and values system than what the US has today post leveraged buyout conglomerates and software company monopolies.

Scroll down the list and look how many there are and where they were manufactured. Economic efficiency would mean only a few vendors.

I think a national security argument is much more sound than an economic one, although costs are externalized in a way that isn't obvious, i.e. ecological disaster that shipping everything around the world and back (components, assemblies) is, and hollowing out a local supply chain takes virtually no time while the impact or limits of it are hidden until abrupt breakage (i.e. covid-era shortages on basic supplies, wars, or heavy handed statesmen dictating preferential access to silicon or whatever today). That is, every nation has to maintain some stake in not hollowing out completely while still participating in global commerce.

Once upon a time nations understood the issues better: https://en.wikipedia.org/wiki/List_of_telephone_switches

I always just set up root recursors at my home and other locations. I've never noticed any downside.