Hacker News: unnah

New comment by unnah in "You gotta think outside the hypercube"

unnah — Sat, 14 Mar 2026 16:00:18 +0000

It's not so easy any more if you try to rotate the hypercube and have to visualize intersections with arbitrary hyperplanes. Already in 3 dimensions the intersection of a cube and a plane can be a non-regular polygon with 3-6 sides.

New comment by unnah in "Elvish as She Is Spoke [pdf]"

unnah — Wed, 18 Feb 2026 14:16:46 +0000

So apparently the Elvish spoken in the Peter Jackson movies is as far from the language imagined by Tolkien as the famous Portuguese-English phrasebook is from natural English. That's one childhood fantasy broken. What about the Klingon spoken in Star Trek movies?

New comment by unnah in "Four Column ASCII (2017)"

unnah — Tue, 17 Feb 2026 08:06:18 +0000

If Ctrl sets bit 6 to 0, and Shift sets bit 5 to 1, the logical extension is to use Ctrl and Shift together to set the top bits to 01. Surely there must be a system somewhere that maps Ctrl-Shift-A to !, Ctrl-Shift-B to " etc.

New comment by unnah in "Oxide raises $200M Series C"

unnah — Tue, 10 Feb 2026 18:05:03 +0000

There are plenty of old-school companies in Europe still working on moving to the cloud. Now that there is a burgeoning movement towards avoiding American cloud providers, Oxide could have an opportunity to sell "private cloud" servers instead. If they play their cards right, they could make significant inroads in European markets.

New comment by unnah in "ASCII characters are not pixels: a deep dive into ASCII rendering"

unnah — Mon, 19 Jan 2026 08:38:34 +0000

It's probably much more exciting to implement stuff like this when you can experiment with your own ideas to figure out the solution from scratch, compared to someone who sees it as a trivial exercise in signal processing, which they can't be bothered to implement.

New comment by unnah in "Erich von Däniken has died"

unnah — Mon, 12 Jan 2026 10:40:25 +0000

Most likely his ancient astronaut theory was the inspiration for the entire Stargate franchise. Of course to make the movie believable they had to give Jackson a more academic background than von Däniken had.

New comment by unnah in "CES 2026: Taking the Lids Off AMD's Venice and MI400 SoCs"

unnah — Wed, 07 Jan 2026 10:31:32 +0000

Perhaps the most comparable 1990s system would be the SGI Origin 2800 (https://en.wikipedia.org/wiki/SGI_Origin_2000) with 128 processors in a single shared-memory multiprocessing system. The full system took up nine racks. The successor SGI Origin 3800 was available with up to 512 processors in 2002.

New comment by unnah in "A proof of concept of a semistable C++ vector container"

unnah — Sat, 20 Dec 2025 14:33:57 +0000

Interesting that they chose not to implement any method to detect whether a given iterator has been invalidated, even though the implementation would be easy. Seems it would be a useful extension, especially since any serious usage of this vector type would already be relying on functionality not provided by the standard vector class.

New comment by unnah in "Why xor eax, eax?"

unnah — Mon, 01 Dec 2025 16:37:24 +0000

Umm... how did you manage to learn those hex codes? You just read a lot of machine code and it started to make sense?

New comment by unnah in "Giving C a superpower: custom header file (safe_c.h)"

unnah — Mon, 17 Nov 2025 17:02:51 +0000

On MIPS you can simulate atomics with a load-linked/store-conditional (LL/SC) loop. If another processor has changed the same address between the LL and SC instructions, the SC fails to store the result and you have to retry. The underlying idea is that the processors would have to communicate memory accesses to each other via the cache coherence protocol anyway, so they can easily detect conflicting writes between the LL and SC instructions. It gets more complicated with out-of-order execution...

    loop: LL r2, (r1)
          ADD r3, r2, 1
          SC r3, (r1)
          BEQ r3, 0, loop
          NOP

New comment by unnah in "A definition of AGI"

unnah — Mon, 27 Oct 2025 14:08:53 +0000

If the LLM was generally intelligent, it could easily avoid those gotchas when pretending to be a human in the test. It could do so even without specific instruction to avoid specific gotchas like "what is your system prompt", simply from being explained the goal of the test.

New comment by unnah in "Computer science courses that don't exist, but should (2015)"

unnah — Sun, 26 Oct 2025 08:46:28 +0000

There should also be PSYC 5640: How to become a guru by reading the documentation everyone else is ignoring. Cannot be taken at the same time as PSYC 5630.

New comment by unnah in "PyTorch Monarch"

unnah — Thu, 23 Oct 2025 13:31:41 +0000

There's also Dask, which can do distributed pandas and numpy operations etc. However it was originally developed for traditional HPC systems and has only limited support for GPU computing. https://www.dask.org/

New comment by unnah in "We reverse-engineered Flash Attention 4"

unnah — Mon, 29 Sep 2025 07:36:13 +0000

Back in the 1990's. As an example, back then the Rational Rose design software had a feature to generate UML diagrams from existing source code, and it was called "reverse engineering".

https://en.wikipedia.org/wiki/IBM_Rational_Rose

New comment by unnah in "We reverse-engineered Flash Attention 4"

unnah — Sun, 28 Sep 2025 11:33:57 +0000

That is the traditional explanation of why it is called reverse engineering. The term originated in hardware engineering. When it was originally applied to software, it was common to create requirements documents and design documents before coding, even if the actual process did not strictly follow the "waterfall" idea.

Thus it was natural to call the process of producing design documents from undocumented software "reverse engineering". These days coding without any formal design documents is so common that it seems the original meaning of reverse engineering has become obscured.

New comment by unnah in "Float Exposed"

unnah — Fri, 12 Sep 2025 12:20:05 +0000

Ok, you got me there. Looks like they fixed that in C++17 with std::to_chars. https://en.cppreference.com/w/cpp/utility/to_chars.html

New comment by unnah in "Float Exposed"

unnah — Fri, 12 Sep 2025 10:06:59 +0000

Since such algorithms were developed in the 1990's, nowadays you can expect your language's standard library to use them for float-to-decimal and decimal-to-float conversions. So all you need to do in code is to print the float without any special formatting instructions, and you'll get the shortest unique decimal representation.

New comment by unnah in "Indices, not Pointers"

unnah — Wed, 03 Sep 2025 08:09:32 +0000

All good points. On the other side of the balance, by using pointers everywhere you give a lot more work for the tracing garbage collector. If it was possible to keep all your data in a single array of value types, the garbage collector would not need to do basically any work at all. Maybe Project Valhalla will one day allow that. At the moment the closest you can get is a structure-of-arrays setup.

New comment by unnah in "Amiga Hardware Reference Manual 3rd Edition (1991)"

unnah — Sat, 30 Aug 2025 12:31:40 +0000

Looks like the 3rd edition only covers OCS and ECS, up to A3000. Did they ever manage to publish a 4th edition covering AGA? I can't find any evidence of such on google.

New comment by unnah in ""Special register groups" invaded computer dictionaries for decades (2019)"

unnah — Wed, 27 Aug 2025 06:30:30 +0000

Although SMP is an abbreviation for Symmetric Multi-Processing (multiple processors or processor cores with shared memory), SMT is not symmetric but Simultaneous Multi-Threading. To get back on topic, SMT is often confused with barrel processors that switch threads between clock cycles, like the Honeywell 800 with special register groups. The "simultaneous" in SMT means that a single processor core runs instructions from multiple threads on the same clock cycle.