The mathematics here are far beyond me, but it's interesting that Gemini more or less concurs with chatgpt with respect to "load bearing reliance on other preprints".
Gemini's summary (there's much more in the link above that builds up to this):
The mathematics is largely robust but relies on heavy "black box" theorems (Iritani's blowup formula, Givental's equivariant mirror symmetry) to bypass difficult geometric identifications in the non-archimedean setting. The primary instability is the lack of a non-archimedean Riemann-Hilbert correspondence, which limits the "enhanced" atoms theory. The core results on cubic fourfolds hold because the specific spectral decomposition (eigenvalues $0, 9, 9\zeta, 9\zeta^2$) is robust and distinct22, but the framework's extension to finer invariants (integral structures) is currently obstructed.
For other people, they're going to be thinking "some other company is going to get one of the most effective and impactful performance engineers on the planet".
Aah, that's much better and more realistic than my previous assumption that they were "government instructions", something used in the military and similar more secretive contexts, but I suppose they didn't use off-the-shelves components perhaps like today.
These instructions were not intentionally designed and put in there in secret. They're simply an unintended consequence of the "don't care" states of the instruction decoding logic.
The decoder is the part of the CPU that maps instruction opcodes to a set of control signals. For example "LDA absolute" (opcode 0xA5) would activate the "put the result in A" signal on its last cycle while "LDX absolute" (opcode 0xA6) would activate the "put the result in X" signal. The undocumented "LAX absolute" (opcode 0xA7) simply activates both because of the decoder logic's internal wiring, causing the result to be put in both registers. For other undocumented opcodes, the "do both of these things" logic is less recognizable but it's always there. Specifically disallowing these illegal states (to make them NOPs or raise an exception, for instance) would require more die space and push the price up.
See here[1] for example to get a sense of how opcode bits form certain patterns when arranged in a specific way.
I don't think they were "intended" for anything - it's just that was the state of the control lines after it decoded that instruction byte, and combination might do something somewhat sane.
Wiring all the "illegal" instructions to a NOP would have taken a fair bit of extra logic, and that would have been a noticeable chunk of the transistor budget at the time.
"Scrape" intervals (and the plumbing through to analysis intervals) are chosen precisely because of the denoising function aggregation provides.
Most scaled analysis systems provide precise control over the type of aggregation used within the analyzed time slices. There are many possibilities, and different purposes for each.
High frequency events are often collected into distributions and the individual timestamps are thrown away.
Let's say you have 10 million servers. No matter what you are deploying, some of those servers are going to be in a bad state. Some are going to be in hw ops/repair. Some are going to be mid-deployment for something else. A regional issue may be happening. You may have inconsistencies in network performance. Bad workloads somewhere/anywhere can be causing a constant level of error traffic.
At scale there's no such thing as "instant". There is distribution of progress over time.
The failure is an event. Collection of events takes times (at scale, going through store and forward layers). Your "monitorable failure rate" is over an interval. You must measure for that interval. And then you are going to emit another event.
Global config systems are a tradeoff. They're not inherently bad; they have both strengths and weaknesses. Really bad: non-zero possibility for system collapse. Bad: Can progress very quickly to towards global outages. Good: Faults are detected quickly, response decision making is easy, and mitigation is fast.
Hyperscale is not just "a very large number of small simple systems".
Denoising alerts is a fact of life for SRE...and is a survival skill.
Exact same here. I watched that video about ten times in a row.
I learned it's more important to know how the big sound pieces fit together and what you can do by tweaking them. Have many, many different versions of the big pieces doesn't really matter.
I also came away wishing that Bitwig had a strudel mode. Every time anyone does anything in the grid, they'd be better off with a strudel equivalent. I think.
Trucks think only trucks can tow.
I tow a 24 foot boat with an Audi Q7. Reasonably frequently, truck guys say something like "You tow that, with THAT?"
Uh, yeah. 7700 pound tow capacity (nearly as much as a base F150). Tows really well.
reply