Hacker News Reader

ztenon VMware Fusion Pro: Now available free for personal use

Yeah, if you care about 3D acceleration on a Windows guest and aren't doing pcie passthrough, then KVM sure isn't going to do it. There is a driver in the works, but it's not there yet.

edit: I made a mistake and got confused in my head with qemu and the lack of paravirtualized support. (It does have a PV 3D linux driver, though)

ztenon How to Use the Foreign Function API in Java 22 to Call C Libraries

Hmm, interesting. They do have DLLs in the JAR...

ztenon How to Use the Foreign Function API in Java 22 to Call C Libraries

Yep, you're right, they do exactly that. Apologies for the confusion.

Decompiled class file:

    try {
        var4 = File.createTempFile("libzstd-jni-1.5.0-4", "." + libExtension(), var0);
        var4.deleteOnExit();

ztenon How to Use the Foreign Function API in Java 22 to Call C Libraries

Yes. Check out a library like zstd-jni. You'll find native libraries inside it. It'll load from the classpath first, and then ask the OS linker to find it.

ztenon Exercise, metabolism, and weight: New research from The Biggest Loser (2022)

Your heart literally gets bigger! Your stroke volume goes up, and you probably have more blood plasma, too. This aids in fueling muscles.

Though, if you Google heart hypertrophy, you'll find a bunch of scary things, because your heart can also get bigger in a bad way rather than a good way.

ztenon Ubuntu 24.04 Noble Numbat

Having no snaps able to use DNS after upgrading thanks to AppArmor sure was a fun surprise

ztenon Sleep apnea: Mouthguards less invasive, just as effective as CPAP

> a very fit and very strong person can wrongly be classified (and in the opposite direction too)

Unfortunately, it usually goes in the bad direction, by quite a bit. BMI under-predicts obesity. You only have to hit 25% body fat as a male to be obese, or 32% as a female.

https://academic.oup.com/jes/article/7/Supplement_1/bvad114....

ztenon Hacked Nvidia 4090 GPU driver to enable P2P

This represents pretty well how gamers (residential buyers) are going to feel when the next generation of consumer cards are scooped up for AI.

ztenon Thanks FedEx, this is why we keep getting phished

No, it goes further than that, all the way back to 1934. USPS is the only authorized service to use a mailbox. Here is an altogether far too detailed study of the law: https://www.gao.gov/assets/ggd-97-85.pdf

You need separate bins/boxes/whatever for other services to use.

ztenon SSDs have become fast, except in the cloud

> On the contrary, young people often show up having learned on their super fast Apple SSD or a top of the line gaming machine with NVMe SSD.

Yes, this is often a big surprise. You can test out some disk-heavy app locally on your laptop and observe decent performance, and then have your day completely ruined when you provision a slice of an NVMe SSD instance type (like, i4i.2xlarge) and discover you're only paying for SATA SSD performance.

ztenon Meta AI releases Code Llama 70B

Multi-GPU in desktop chassis gets crazy pretty quickly. If you don't care about aesthetics and can figure out both the power delivery and PCI-E lane situation, https://timdettmers.com/2023/01/30/which-gpu-for-deep-learni... has an example that will make a Bitcoin mining rig look clean.

Water cooling can get you down to 2x slot height, with all of the trouble involved in water cooling. NVIDIA really segmented the market quite well. Gamers hate blower cards, but they are the right physical dimensions to make multi-GPU work well, and they are exclusively on the workstation cards.

ztenon Meta AI releases Code Llama 70B

Well, the workstation-class equivalent of a 4090 -- RTX 6000 Ada -- has enough RAM to work with a quantized model, but it'll blow away anyone's budget at anywhere between $7,000 and $10,000.

ztenon S-LoRA: Serving Concurrent LoRA Adapters

While not related, thank you for mentioning PyTorch eager mode. I had no idea this existed. There is just SO MUCH documentation to read, and it is bewildering how bad the defaults are in the whole deep learning ecosystem.

ztenon S-LoRA: Serving Concurrent LoRA Adapters

> The Stable Diffusion AI Horde already does this, just with no batching and tremendous inefficiency.

Yes, the inefficiency here is the sheer number of models that are served. Instead of every worker having SDXL and loading LoRAs, the worker is very rapidly switching between one of ~230 models... _and_ loading LoRAs, if the worker allows it! From discussions I've seen, there is a lot of encouragement to use LoRAs (and textual inversions) instead of a unique model, but there's still a lot of demand among clients for unique base models. Additionally, being a volunteer system, a lot of the workers are, well, strange. You have people contributing high-end GPUs, and then other people are trying their best to (abuse?) use Colab and Kaggle within whatever their TOS allows, so the capacity of the workers varies quite a bit.

ztenon ChatGPT-4 significantly increased performance of business consultants

I've seen code bases survive three different ticket management systems. Meanwhile, the tickets never made it between the different systems, so if the 'why' isn't in the commit message, then it got lost to time.

I will admit that a lot of the really old decisions don't have much relevance to the current business, but the historical insight is sometimes nice.

ztenon Fine-tune your own Llama 2 to replace GPT-3.5/4

Thank you for posting this. I had to go look for your HuggingFace data sets to find the labeled variety you produced with GPT-4, but other than that, everything was easy to follow.

ztenon Windows 11 will happily execute a binary compiled 30 years ago

Ah, yes, that's true, it's usually not done maliciously like the ATI example. It came to mind because it was a well-documented case of the driver being aware of the application and changing behavior.

ztenon Windows 11 will happily execute a binary compiled 30 years ago

See the infamous story about Quack 3 on ATI Radeon cards.

ztenon Why host your own LLM?

    ~ > sudo nvidia-smi -pl 300
    Power limit for GPU 00000000:09:00.0 was set to 300.00 W from 450.00 W.

ztenon MS Teams channels cannot contain MS-DOS device names

For the absolute worst modems, I believe you could just send that in a PING on IRC and their client would write the string back and hang up

ztenon Commuters prefer origin to destination transfers

> Offhand, I can't think of any 'park and ride' style Car <> Transit (train or not) interface in the USA that _does_ charge anything to park there. Finding a spot might be an issue, but there's no cost as that's sort of built into getting someone to take the transit system at all.

Almost every MBTA garage/lot in Massachusetts charges something. BART parking in the Bay Area also does. It's a pittance, though.

ztenon One million cancel broadband as living costs rise

I've got Verizon 5G for $25/month extra on top of my cell phone bill, and the tower is right across the street. It's pretty competitive with Xfinity's 1.2Gbps service since they didn't bother throttling it. I also get 350Mbps up. Seeing 15ms pings to servers in Sacramento from SF.

ztenon How to design software architecture for startups

The way to do this wrong (by forcing the use of the "blessed" client or common platform dependencies, instead of leaving it open to the consumer as you suggest) is discussed in this talk: https://www.microservices.com/talks/dont-build-a-distributed...

ztenon It’s Not Just You: 5G Is a Big Letdown

I think I've seen the same thing. I use Verizon in SF for both an iPhone and for their 5G home internet service, so I've got some basis for comparison.

During the setup for the home internet, they were very particular about where to place the device for line of sight to the tower. It consistently gets 1Gbps down and 300Mbps when in 5G UWB mode. There are rare situations in which it falls back an LTE connection that is somehow unimaginably worse than phone tethering.

On the other hand, when I'm using my phone, it'll sometimes _think_ it has the ability to do ultra wideband, but it doesn't work, and the icon disappears and suddenly I'm back on ordinary 5G again with a bit of a hiccup that you note. If I ensure that my phone has clear line of sight to the base station, and confirm that it is using UWB, then I can get similar speeds and latency to the 5G home internet device.

The ultra wideband mode is very sensitive to where you are in relation to the base station, so on a phone moving around your house it may be inconvenient or frustrating to have your phone attempt to use it.

ztenon Associations between alcohol consumption and gray and white matter volumes

This is in a different measurement from a casual understanding of 6 drinks. It’s 6 daily alcohol units. UK’s NHS alcohol calculator says 2 pints of 5.2% beer is 6 units.

ztenon Car Seats as Contraception (2020)

Malthusian thinking is rampant in public policy discourse. Many people will become intent on stopping planning for growth when they buy a home.

ztenon Amazon to Acquire One Medical

I agree and I thought that might be a problem. I would be in the same position.

ztenon Amazon to Acquire One Medical

I know this is the wrong answer, because you should be expecting reasonable care from your physician, but if you're on an HDHP plan just buy the service from LabCorp and reimburse yourself. For some reason, direct-to-consumer blood work is becoming a thing.

ztenon Ask HN: Does Apple secretly want to prevent gaming on macOS?

> Try booting up any Mac game from the 2010s on Apple Silicon; there's a 90% chance it depends on a 32-bit library Apple won't give you anymore.

The funny thing is... you know what does work on Apple Silicon? Wine, using 64-bit Windows binaries, DXVK, and a Vulkan to Metal wrapper. Don't expect stunning results on a just-released title like Elden Ring, but there's a project called XIV on Mac for an old title, FF14, that has at least proved to me that it can be done.

ztenon Uber and Lyft’s new road: Fewer drivers, thrifty riders and jittery investors

I remember there being signs on Spear St at Folsom with specific destinations for carpools. I can see them on Google Street View in 2021, but they appear to be removed in 2022. (or maybe they're designated on the paper you can't read on street view right now?)