[HN Gopher] Bringing the Instructions to the Data
___________________________________________________________________
Bringing the Instructions to the Data
Author : mattpope
Score : 41 points
Date : 2024-12-04 03:06 UTC (5 days ago)
(HTM) web link (mattpo.pe)
(TXT) w3m dump (mattpo.pe)
| deathanatos wrote:
| Regarding the thesis of the article of bringing the instructions
| to the data, I'd love to see the core of that applied much more
| broadly in software / aaSes. Random services probably won't JIT
| LLVM IR for you, given the security of running untrusted LLVM IR
| ... but there's also WASM out there. I'd _love_ to see interfaces
| start having the capability of accepting WASM instead of an
| enumeration of the 2 use cases a PM thought I might have.
|
| E.g., we use PagerDuty, and there's several places, such as
| routing pages, where I'd just like programmatic control, and
| _code_ would express much more succinctly the needs of what I
| want to do than trying to express it through some UI-based
| "routing" editor thing.
|
| Artifact storage aaS's often come with "cleanup" policies that
| let you choose between 2-3 different modes of cleanup, mostly
| wrong one way or another.
|
| In all cases an enumeration of the names of 2-3 fixed functions,
| when I'd rather _pass the function_.
|
| ---
|
| Since this is posted by the author... you set `.content a { word-
| break: break-all; }`? Like the name sort of implies, it permits
| line breaks _anywhere_ , which means that the opening line
| renders as, One paper that has been in the back
| of my mind is for a few years is Efficiently Compil ing
| Efficient Query Plans for Modern Hardware1 which details how
| Tableau's internal
|
| (I.e., it splits mid-syllable, and without a hyphenation (which
| is a different CSS property).)
|
| The layout is basically fixed at the containing element's max-
| width for nearly any width someone might reasonably be using, so
| outside of font variations, it should basically always render
| like that (and I don't think font variations would make enough of
| a difference). (It steps up to a wider width at some point ...
| but that width _also_ lops a word on the opening line.)
| mattpope wrote:
| A great idea but the last time I worked with WASM, using
| something like Wasmtime's shared host functions
| https://github.com/bytecodealliance/rfcs/blob/main/accepted/...
| they had quite the overhead. Maybe things have changed and have
| become less heavy.
|
| Thank you for the styling tip, I'll see what I can do it does
| look wonky.
| xnx wrote:
| Thought this was going to be about computational RAM:
| https://en.wikipedia.org/wiki/Computational_RAM
| pjot wrote:
| Apache Substrait is doing exactly this. Paired with Arrow makes
| for powerful composable data architectures
|
| https://substrait.io/
| mattpope wrote:
| This is the first time I've seen this. It looks quite promising
| for taking care of a logical planning phase. But you still need
| an execution engine like Calcite or Datafusion (or an LLVM IR
| based executor)
| mzs wrote:
| links to his useful blogpost: A Gentle Introduction to LLVM IR
|
| https://mcyoung.xyz/2023/08/01/llvm-ir/
| mattpope wrote:
| Yes! All of the articles by this author are insightful and well
| written.
| jonlong wrote:
| > The astute reader will notice that float operations are not
| communicative
|
| Presumably this was meant to read "commutative". IEEE 754
| addition and multiplication _are_ commutative (ignoring NaN
| values), but not associative.
___________________________________________________________________
(page generated 2024-12-09 23:01 UTC)