remexre

Also, I'd imagine you'd want the Ztso extension to port PC games, assuming you mean Rosetta-style instruction translation rather than "somehow get the source and port the engine and all the middleware" -- I don't think the P550 has that extension.

It seems very logical to me that there'd be orders of magnitude more training data for some domains than others, and that existing models' skill is not evenly distributed cross-domain.

The list of fused instructions is kinda weird; besides the ones that correspond to SH{1,2,3,4}ADD anyway, I don't think I expected any of the others.

I don't see a mention of Sifive-style short jump predication there either, though I'm using Google Translate.

if you look at this from the riscv angle, moving from "u-mode only vm that doesn't use paging under the hood" to "u+s-mode vm with sv39" isn't an enormous jump in complexity imo

i think i might teach it starting as like, "sv21" (page tables aren't nested), then pose real sv39 and the tree structure as the solution to making a sparse mapping over 512GiB

then moving on to the idea of having a TLB is simple, especially if students have already been introduced to hashtables

however, my university has a mandatory class taught in ocaml, which i've ta'd for a few times; this is the _number one_ "the undergrad ta couldn't figure out my syntax error" issue students have

    (5GiB / 1TiB) * 4KiB to bits

      ((5 gibibytes) / (1 tebibyte)) × (4 kibibytes) = 160 bits

    loopType : Int -> Type
    loopType x = loopType x

    foo : List (loopType 3) -> Int
    foo _ = 42

    bar : List (loopType 4)
    bar = []

    baz : Int
    baz = foo bar

If I recall correctly, SVN was also more popular than Git at the time, so migrating hosts was a lot more painful than now...

    $ ./result/bin/llama-cli -m models/Meta-Llama-3-8B-Instruct/ggml-model-q8_0.gguf -f <(printf 'User: Summarize the following:\n\n'; w3m -dump https://www.jeremykun.com/2024/09/02/shift-networks/; printf '\n\nAgent: ') -fa -ngl 28 -ts 1,0

Summary:

The article discusses the concept of packing in homomorphic encryption schemes, which involves placing multiple plaintext data into RLWE ciphertexts in a way that minimizes the number of extra multiplications and rotations required. The author explains the problem of converting between packing strategies and presents two methods: the naive approach and the Vos-Vos-Erkin g raph-coloring approach. The article also mentions the importance of minimizing the multiplicative depth and the number of rotations in the shift network.

Harry Potter Fanfiction:

As Harry and Ron walked through the Hogwarts grounds, they stumbled upon a mysterious door hidden behind a tapestry. The door was adorned with strange symbols and markings that seemed to shift and change as they looked at it.

"This looks like some kind of puzzle," Ron said, his brow furrowed in concentration.

[...]

I wonder which ones he considers those to be -- the interview doesn't say, sadly...