> It sometimes takes a lot of time but I read everything. It still faster than nothing.
Opposite experience for me. It reliably fails at more involved tasks so that I don't even try anymore. Smaller tasks that are around a hundred lines maybe take me longer to review that I can just do it myself, even though it's mundane and boring.
The only time I found it useful is if I'm unfamiliar with a language or framework, where I'd have to spend a lot of time looking up how to do stuff, understand class structures etc. Then I just ask the AI and have to slowly step through everything anyways, but at least there's all the classes and methods that are relevant to my goal and I get to learn along the way.
What I don't get is why they don't at least assign a dev or two to make the poster child of this work: llama.cpp
It's the first thing anyone tries when trying to dabble in AI or compute on the gpu, yet it's a clusterfuck to get to work. A few blessed cards work, with proper drivers and kernel; others just crash, perform horribly slow, or output GGGGGGGGGGGGGG to every input (I'm not making this up!) Then you LOL, dump it and go buy nvidia et voila, stuff works first try.
How so? It's rock solid for me. I use ollama but it's based on llama.cpp
It's quite fast also, probably because that card has fast HBM2 memory (it has the same memory bandwidth as a 4090). And it was really cheap as it was on deep sale as an outgoing model.
Ah I see. Yes, but you pick the card for the purpose of course. I also don't like the way they have such limited support on ROCm. But when it works it works well.
I have Nvidia cards too by the way, a 4090 and a 3060 (the latter I use for AI also, but more for Whisper because faster-whisper doesn't do ROCm right now).
Aside from the fact that gfx906 is one of the blessed architecture mentioned (so why would it not work). Like how do you look at your specific instance and then turn around and say "All of you are lying, it works perfectly." How do you square that circle in your head
No I was just a bit thrown by the "sometimes". I thought they were referring to a reliability issue. I am aware of the limited card support with ROCm and I complained about this elsewhere in the thread too.
Also I didn't accuse anyone of lying. No need to be so confrontational. And my remark to the original poster at the top was from before they clarified their post.
I just don't really see what AMD can do to make ollama work better other than porting ROCm to all their cards which is definitely something they should do.
And no I'm not an AMD fanboi. I have no loyalty to anyone, any company or any country.
libgd, the graphics backend of php. It's an ancient pile of shit yet the "maintainers" refuse An help, contributions and improvement as good as they can. The most popular example is emoji support on their text rendering. Questions on SO, bug reports to libgd and php go back at least 10 years, yet no interest by the devs to either fix it or accept any help in doing so.
A couple years ago I started investigating and it starts with the fact that their utf8 decoder accepts no more than 3 bytes per character, thanks to an ifdef. There's code to handle 4, so you change the ifdef and then it doesn't compile anymore because code rot. Fixing that you see the correct code points going into freetype but then something else I forgot about breaks. Then I saw there were already a couple patches on the mailing list and checking their github just now I see at least one PR with a bunch of back and forth and another couple issues with mostly no reaction or at best a few lame excuses.
If you could donate negative money to a project this would be a prime candidate. Sure, move fast and break things is bad, but this is the opposite of that, and not good either.
And if the maintainers are dragging their feet because they hate emoji(my guess, it's common enough), what else do they expect than to be constantly nagged by users.
Interesting, at the airport I always see general "no (devices with) batteries in checked luggage" signs.
One time I had an old phone with battery removed in checked luggage and when I arrived at my destination, I saw they fiddled with the tsa lock and the phone was taken out of the envelope I had it in and just lying on top of my clothes. I mean maybe they saw it in an xray and wanted to steal it and then saw it was some old junk phone, no idea how good the xrays come out to tell beforehand whether you're dealing with an iPhone or 7 year old android midrange...
This was the norm in Germany for a while. Then we got a law ("Routerfreiheit") that forces ISPs to allow you to use and modem/router you want. Before that you couldn't do it with DOCSIS, but it kinda worked with DSL, if you could somehow get the PPPoE credentials.
Nowadays they have to provide PPPoE and voip credentials, and cable providers need to provide a service where you give them the Mac of your own modem and then it gets tied to your contract.
That's not their point? Linux has all the drivers lying around on disk and on every boot just does hw detection and loads the appropriate ones. It takes a few seconds. You can take a Linux install from a modern AMD system, stuff it into a 10 year old Intel system and it will boot up instantly. No driver install no "getting your devices ready" screen that shows up for a minute or two.
This has nothing to do with being open source or being customizable. It's simply pointing out how fast hw detection is not only possible but the norm on other systems.
PCI wasn't a given. Plug and play wasn't a given. You couldn't even reliably enumerate all hardware on a system.
Hardware detection back then involved a lot of poking at random IO ports and seeing what happens, using heuristics to select an appropriate driver. This is as dodgy as it sounds and would crash or hang your system if you weren't lucky.
One thing that helps this is that most of that really buggy hardware has fallen into the wastebin of history and everything attached to a "modern" W98 machine should be plug and play compatible. PCI solved most of these issues. ISA cards gave Windows 98 and especially 95 a bad rap. Well, that and early USB controllers and devices. There was a whole lot of brand new driver code being tested in production back then.
> with a custom data packing method that is optimized for streaming directly from CD to the hard disk without any seeking
This is nice. I've always wondered why they don't do this with the core parts of the os and then only extract additional components and drivers. But maybe back then the core was only a few MB and it wouldn't have helped so much...
I remember the setup taking ages. With 9x I don't think any install ever lasted longer than a year, so I did this a lot. :)
This is definitely an interesting part of it. Copying a few hundred megs from CD to HDD doesn't take long even for old hardware, but when I was setting up a win98 rig last autumn to test some 3dfx cards it took the common ~45 minutes to get through the main install, and that's before you get into the post-install cycle of installing any software/drivers/updates that need reboots. The "unofficial service pack3" is one I'd love to integrate as it includes support for USB mass storage which made the machine a lot easier to work with
On a side-note, walking the line between annoying and entertaining was the noise of the HD during install, which sounded like techno music and I should have recorded it. Weirdly it was only during the win98 install that it made that type of sound.
This was definitely a thing in the optical disc era of games where seek times were horrendous. In record mode, this is done by just overloading the file read functions, recording a list of file, seek position, and read size instructions, and then using that to build a .dat file. In play mode, the function is overloaded to ignore file opens and seeks, and to just read from the contiguous file. This requires the load to be perfectly deterministic, and preferably without redundancy.
I don't think it's clickbait. Even though the title doesn't mention C# or .net explicitly it seems clear from that Span<> stuff that this is talking about some higher level language...
How about spice? I don't know anything about the technical details, but it seems to offer most of the advanced features like usb redirect and I think currently gets video support. It's currently really only used in qemu but I don't see a reason why you can't use it as a general open rdp alternative.
I think most people will agree that ui is the hardest part for normal desktop software. Not getting any kind of ui going, but good ui. Consistency, how to group things, what goes where.
If your app needs two inputs and a button then this won't be so much of an issue, but I've wrestled with qt in the past for more complex layouts that I hate it with a passion.
Opposite experience for me. It reliably fails at more involved tasks so that I don't even try anymore. Smaller tasks that are around a hundred lines maybe take me longer to review that I can just do it myself, even though it's mundane and boring.
The only time I found it useful is if I'm unfamiliar with a language or framework, where I'd have to spend a lot of time looking up how to do stuff, understand class structures etc. Then I just ask the AI and have to slowly step through everything anyways, but at least there's all the classes and methods that are relevant to my goal and I get to learn along the way.
reply