this has been bugging me a lot. like, yeah, there’s definitely AI scams out there. and yeah, a lot of people are using it from the wrong end, but it’s also clearly a substantial technology. time to realize that https://mas.to/@carnage4life/112484753548884371
@kellogh Definitely doesn't help AI's image that the chief hype man of AI spent the previous few years building out a large, privacy-invading crypto scam 😅
I am (unfortunately) on a Macintosh. A program is segfaulting, and I need to run in a debugger to find out why. I run lldb executable-name, and then "run". It prints:
error: the platform is not currently connected.
I do not know what this means. Googling for this error message turns up various things involving iOS and XCode, but I am not using either of those things, I am using lldb at the command line.
@glyph@mcc yeah, I recommend always setting up new Macs from scratch. Even without silicon changes, I’ve had so many problems like this with cruft leftover from older OS versions…
Reinstalling rustup from scratch appears to have fixed things, but I can't say I'm thrilled that my operating system decided to spontaneously excommunicate my compiler toolchain
Say I am developing an Android app. I have the source, I built it myself.
It has a problem where occasionally Android pops up a box saying "[App] isn't responding. * Close app * Wait".
I would like to know what the app is doing when this occurs— stack traces, 1 second of profiling, something.
I expect if I tap "Close app", it should drop something useful in the logcat. I've nabbed a logcat within 120 seconds however and see nothing involving "crash" or "ANR".
@mcc The most common approach would be to hookup the app to Firebase Crashlytics, and then your ANR traces will end up in the cloud.
I think your odds of getting ADB connected and logcat running before the ANR falls out of the buffer are slim-to-none. Although connecting ADB over wifi and just having your PC continually spool logcat to a file the whole day might work?
@mcc Have you explored the /data/anr/ directory? In theory a copy should be stored there for each ANR (unclear what the timeline for eviction is, however)
This is cool. A lot of people misunderstand what “real time” means — generally, “late is an error”. For example, he wants to use it in games, where at 60 fps, you have 16.6ms to do all computations and render the screen. Lateness causes lag. Or worse, in a fly-by-wire steering system, late means difficulty steering a multi-ton vehicle.
in a RT system, you need to be able to statically calculate how long all compute will take. Most memory allocators don’t work like that
@fasterthanlime I tend to think they are also much less willing to hang their brand on non-determinism than many other tech firms. LLMs are cute when they work, but the failure modes are also pretty severe. Siri may be dumb, but he's also quite consistent...
@onepict Could be, but these days the powerr grid is as relaible as anywhere else, and I live on the edge of a dairy farming region, so I don't get it at all 😅
IMHO this makes a lot of sense, the native WebGPU implementation libraries turned out to contain quite a bit of fat.
TBH looking at Dawn I was rolling around that idea in my head too (WebGPU without the shader compiler isn't all that much code, and I would have moved that offline like in sokol-gfx).
@floooh Part of the problem here is that WebGPU took some decisions that make runtime shader compilation mandatory (i.e. there are pipeline states that have to be implemented by changing the shader compilation on various backends). If those states could be eliminated, everything could become a lot simpler
There’s a voice in my head suggesting a $4,000 MacBook upgrade to… be able to build machine learning docker images without hating my life. Feels like we took a wrong turn somewhere along the way
(side-effective of growing up poor, I guess. Took me a full day to introspect why I feel this way about a work-related purchase. Can‘t ever turn the “save a penny” voice off, even in cases where it doesn’t make sense)
@jonikorpi I'm with you on the downside of laptops. Unfortunately, I'm fairly locked into the Mac ecosystem for web+ML, and desktop macs are perenially worse than their laptop counterparts...