This readme, this header do not seem to discuss in any way the tradeoff, which is that you're paying by the same factor with median latency to buy lower tail latency. Nobody thinks of a load as taking 800 cycles but that is the baseline load latency here.
Also, having sacrificed my own mental health to watch the disgustingly self-promoting hour-long video that announces this small git commit, I can confidently say that "Graviton doesn't have any performance counters" is one of the wrongest things I've heard in a long time.
Overall, I give it an F.
Anyway if you want to hide memory refresh latency, IBM zEnterprise is your platform. It completely hides refresh latency by steering loads to the non-refreshing bank, and it only costs half the space, not up to 92% of your space like this technique.
Nope, there isn’t a tradeoff; median latency isn’t affected. I don’t think you understand the code. The p50 is identical between a single read and the hedged strategy.
The clflush is there because the technique targets data that will miss the cache anyway. If your working set fits in L1, you don’t need this.
Also, AWS Graviton instances absolutely do not expose per-channel memory controller counter PMUs. That’s why you have to use timing-based channel discovery.
The IBM z-system is neat! But my technique will work on commodity hardware in userspace, and you can easily only sacrifice half the space if you accept 2-way instead of 8+ way hedging. It’s entirely up to you how many channel copies you want to use.
Your reply was quite rude, but I hope this is informative.
I was just trying to reconcile his reply with the charts. Have you tested how this scales down for smaller systems, as one might find in on the management side of a network switch?
I won't be tone-policed by a person who is clearly trying to mislead and confuse people. I leave it to the other HNers to read your benchmark code and see for themselves that it is an exercise in absurdity, a work-around for its own library that doesn't measure anything other than with N threads, because of the laws of probability, this technique of reading timestamps as fast as possible and cramming them into a vector yields lower measurements with higher N.
You were rude for absolutely no reason. You could point out where you think the article comes short and make suggestions on how to improve it. With this approach, you achieved nothing.
Being competent requires being knowledgeable AND getting things done. You might be knowledgeable, but you need to learn how to work with other people.
Even current Dell servers less than a year old ship with G200 graphics. If it works, why change it? A 1998 ASIC can be put in the corner of a modern chipset for pennies or less.
Sure. I picked IMPACT for having hardware textures. An interesting list would contain the first examples of processors that had things we still have today, like geometry processors. It would also contain evolutionary dead ends that tried to do things differently, like the Rendition Vérité.
It's easy to think about. Google reported a global average power consumption of 3.7GW in 2024, so you can think of this deal as representing an expansion of something like 10-15% of that 2024 baseline, if you assume 50% capacity utilization.
For homicide this is very correct. You could live dead center in the statistically most violent block of Chicago and still cut your personal risk of being a homicide victim by 1) not being a criminal, and 2) not posting diss raps to your 11 followers on Soundcloud. There are not really dangerous neighborhoods but there are dangerous social networks.
Living in Indianapolis (higher homicide rate than Chicago, but not drastically so), I feel the same way.
And for non-homicide violent crime, you're probably more likely to get jumped on a side street near a bar district than you are in a "bad neighborhood" unless you do something yourself to incite violence.
And not having any vehicle problems, because you usually only are rolling through bad areas to get to better areas. Most people in violent cities have no occasion to stop in violent areas. On one occasion I was forced to work overnight for critical hospital operations in a bad part of town, on my way back my tire went flat and when I was distracted fixing it the locals noticed I was weak and they put a gun to my head.
There's a gigantic global industry of telling people that a far away place they've never been is bad and dangerous. This industry exists to make people overlook the objectively bad and dangerous place where they actually live. This plays out on global and national scale. For example, my family who live in Oklahoma City are constantly yapping about some fearmongering they saw about San Francisco, when their city has like 4x the homicide rate, traffic violence, poverty, substance abuse, even manmade earthquakes. But they are plugging into a 24x7 stream of disinformation.
If you visit some place outside your local city, you might experience a broader swath of people, culture, and beliefs. That makes you a dangerous person. You might start questioning why you've only been told one thing. Sadly, just traveling within the US can get you some exposure, but it's mainly just variations of the same thing with perhaps some external influences that have been absorbed. It's nothing like going to abroad and experiencing totally different cultures.
Scary social media videos about those foreign lands is just another tool to dissuade people from wanting to expand their experiences. If you research into some of this global industry, you'll see a lot of the influencers are spokes from central hubs of propaganda sources that shockingly are financed by the same people.
Absolutely. Every country's local media will try to get people to stay home and spend money locally, and not complain because elsewhere is much MUCH worse!
It also makes it easier to dehumanize people in-case of a conflict.
This isn't limited to far-away places. There are plenty of influencers and social media shills that specifically parrot how bad cities are - to people who live in those cities. Hyper-local news outlets that no one outside of the city would read are filled with crime reports and people giving all kinds of warnings. Even decades ago, local news would excitedly report on crime because it draws viewers.
Almost every single large city subreddit needs to have some kinds of rules restricting crime posts, because people love to post every single act of crime that happens and it eventually takes over the sub. Then "alternative" subreddits are created and the cycle continues.
Scaring people has always been profitable, social media just lowers the bar. It's not just old people anymore.
--
I almost fell into one of these traps. It was a guy walking around the dead San Francisco mall and talking about how it was DEVASTATED and SF was COLLAPSING. I had actually been there recently and yes, it was pretty empty. The Youtube algorithm seized on this and started showing me more videos.
Except one of the videos was on a street next to me. The video was just the one guy walking around two blocks that were empty, but looped over and over for 20+ minutes. So it was clearly nonsense.
I checked his channel again. It has paid subscriptions and a PayPal link. He gets minimum 50K views on every single video, with some hitting 800K. He makes two a month. That's an easy income.
Everyone hates people who are not them. I live in a pretty plain middle class neighborhood just Southeast of downtown Dallas (oakcliff). People here talk about the absolute cesspools that are the... safest neighborhoods with good jobs and consistently top rated public schools in North Dallas only because they're also the wealthiest. They just can't stand people who are not like them for some reason.
To be fair the wealthy neighborhoods of North Dallas are notorious hotbeds of crime. It's just not the types of crimes that get reported on right-wing TV.
you'll have to give me some examples. My boys are 16 and 14 and have fortunately managed to get in to the top magnet HSs in DISD (TAG and SEM) but if i had to do it over again, and had the means, I'd choose to raise my kids in Preston Hollow/Highland Park any day.
there's a lot more tax evasion going on in oakcliff and other working class neighborhoods than in any of the wealthier areas. Every other person in oakcliff pays with 100s at the grocery store, almost all business is cash under the table.
I don't know anything about Oklahoma City but I'm pretty confident that SF is a shit hole having lived there up to 2024 and seeing it first hand every day. Any stats that claim otherwise are lying.
Also, having sacrificed my own mental health to watch the disgustingly self-promoting hour-long video that announces this small git commit, I can confidently say that "Graviton doesn't have any performance counters" is one of the wrongest things I've heard in a long time.
Overall, I give it an F.
Anyway if you want to hide memory refresh latency, IBM zEnterprise is your platform. It completely hides refresh latency by steering loads to the non-refreshing bank, and it only costs half the space, not up to 92% of your space like this technique.
reply