kolbe

11 ms·

91.
by kolbe 2y ago
I don't think Gates has much influence on Microsoft these days.
92.
by kolbe 2y ago
I think you're right that IBM is considering in particular the most effective cost savings strategy. And it happens that older people are more effective in cost reduction. But that's not ageism.
93.
by kolbe 2y ago
FWIW, I didn't flag you. Just wanted to say you're accusing me of a straw man, but when I give a particular example that is a concrete example within a general statement, that's not a straw man. Between the two of us, I am lo
94.
by kolbe 2y ago
Ok. Asian Americans are consistently over represented in US higher education, especially in math-heavy fields. What discriminatory practices by these institutions are causing this?
95.
by kolbe 2y ago
So how does this work? Asian Americans are vastly over represented relative to their population in higher education. This "implies" they've been receiving special treatment or favoritism?
96.
by kolbe 2y ago
"Might" is a delightfully cop out word. Anything "might" be true.
97.
by kolbe 2y ago
Disparate outcomes do not imply disparate treatment
98.
by kolbe 2y ago
Runners like the dual band GPS in the Ultra 2.
99.
by kolbe 2y ago
Ultimately it doesn't matter. Your "blue" is just a translation of that frequency to some distinguishable impression to allow you to see. But it's a good bet that the same wiring that went into your brain making that tra
100.
by kolbe 2y ago
Nano is claiming insanely better performance over spdlog, which confuses me
101.
by kolbe 2y ago
I could swear I did a deep dive into Spdlog vs Nanolog six months ago, and the performance differences weren't nearly this stark
102.
by kolbe 2y ago
So, bad CEOs are bad? What a revelation. We're not isolating this analysis to bad ones (who generally don't last very long). We're talking about a blanket, overarching, ill-informed and bitter "analysis" that conten
103.
by kolbe 2y ago
Would you be willing to consider that some half baked theory about why CEOs are the manifestation of luck is actually just an unhealthy way of coping with failures?
104.
by kolbe 2y ago
Chipotle's stock dropped 10% and Starbucks's rose 25% when the CEO switched from one to the other this past week. If this guy is just a meritless lottery winner, then you can make a lot of money betting against the market and the
105.
by kolbe 2y ago
Mine has been a little bit more along the lines of helping them understand how everything is linked. They can't really even understand the power of JOIN, which greatly reduces the power of what I set up. Basically, I'm sort of stu
106.
by kolbe 2y ago
Basic question. Is there a service out there where I can easily link my database to an LLM to do this exact same type of analysis, except on one of my own Postgres DBs instead of one backed by PGlite? My org has several non-technical people
107.
by kolbe 2y ago
I'll put it this way. He clearly loves teaching. He's better at it than almost any teacher I ever had. He's clearly very intelligent. He loves doing work that interests him, as opposed to lucrative professional tasks. I'
108.
by kolbe 2y ago
In the grand hierarchy of college admissions committees ranking people, "getting into northwestern" means roughly "the 20,000th best student in his class year."
109.
by kolbe 2y ago
I think you need to wait one year before being able to post an exact link again. You can ask deng to reset it if you want, but maybe take some initiative for yourself rather than giving other people attitude, then asking them to do you favo
110.
by kolbe 2y ago
It's in his bio. "However, ever since grade school, I've always sucked in terms of grades and standardized tests. I graduated from Palo Alto High School in the bottom quartile among all the college-bound students. My GPA was
111.
by kolbe 2y ago
Mystical (the author) does such fantastic work for the CS community. I really like that guy. A compilation of his stack overflow answers on SIMD would be better than any available book. His intelligence and openness, despite no one paying h
112.
by kolbe 2y ago
Is there a limit to how much the OoO execution buffer can handle? I don't want to dig up the benchmarks again, but I remember for very long calculations (like computing a bunch of exp(x[i]), which boil down to about 40 AVX512 instructi
113.
by kolbe 2y ago
What do you mean? It's an intensely competitive process to get approved by the VC mafia, who then get to dictate which space you get to monopolize
114.
by kolbe 2y ago
> The store will still take up throughput even if nothing depends on it right now - there is limited hardware available for copying data from the register file to the cache, and its limit is two 32-byte stores per cycle, which you'l
115.
by kolbe 2y ago
> I don't understand your point about pipelining - OoO should mean that, as long as there's enough decode bandwidth and per-iteration scalar overhead doesn't overwhelm scalar execution resources, all SIMD ops can run at fu
116.
by kolbe 2y ago
L2 speeds are ~180GB/s on Zen 4. That's also a part of my confusion. This should be line speed. I do not have the same experience as you with unrolling AVX512 loops on Zen 4. I recall even with double pumping, you can do 2.5 per c
117.
by kolbe 2y ago
There is something to be said for using masking load/stores instead of downgrading the SIMD register types. But to my eye, something is clearly wrong with your code in such a way that the "naive" autovectorized version should
118.
by kolbe 2y ago
Well, the conundrum sits with the fact that this is the disassembly of the main loop: vmovdqu64 zmm3, zmmword ptr [rdi + rcx] vmovdqu64 zmm4, zmmword ptr [rdi + rcx + 64] vmovdqu64 zmm5, zmmword p
119.
by kolbe 2y ago
Did you try something like this? The autovectorizer looks pretty clean to me. https://godbolt.org/z/1c5joKK5n
120.
by kolbe 2y ago
The vast vast vast majority of instruction pipelines are sparse. You can pack in non-dependent instructions essentially for free.

More ›