frogblast

8 ms·

31.
by frogblast 4y ago
Not the least of which is likely power and clock state management (both on the CPU and GPU)
32.
by frogblast 4y ago
It is available via public APIs, but the hardware instructions themselves are not documented. This lets the instructions change in future CPUs, vs having the ISA be baked in stone forever. Example: AMX predated the standard ARM matrix mult
33.
by frogblast 4y ago
Regarding having drivers respect preferences... Shippers end up eating the cost of these packages, and they are making an economic tradeoff: Slower deliveries by following preferences vs. eating the cost of loss. They are choosing the lat
34.
by frogblast 4y ago
Well… Yes. It is a scenario where you you have a blind spot on the road ahead of you, so I’d hope a human driver would do the same.
35.
by frogblast 4y ago
The 2xFP32 solution is also dramatically faster than FP64 on nearly all GPUs. While most GPUs support FP64, unless you pay for the really high-end scientific computing models, you're typically getting 1/32nd rate compared to FP32
36.
by frogblast 4y ago
I recall a news story from a few years ago (can't find it now)... There was once a bank that looked the other way when lots of shady cash came in, allowed transfers of those amounts to to foreign banks, basically ignored KYC rules, etc
37.
by frogblast 4y ago
Malloc doesn't guarantee zero filling memory, but ObjC and Swift currently need to zero memory immediately after malloc returns it, and they both share C's malloc pool. This makes a lot more performance sense for those languages.
38.
by frogblast 4y ago
Because Amazon can arbitrarily raise the prices of goods sold at Amazon, and have no concern about being undercut. For example, you are selling a Widget for $20. Amazon says "we'll take $10 of that ourselves". You can sell
39.
by frogblast 5y ago
What you wish for is exactly what HealthKit is. All app access is gated by ACLs that require per-app and per-record-type user approval, and always encrypted when transmitted off of your device. This is a quote from the Apple Health Privac
40.
by frogblast 5y ago
> So given no effective sanction options and no military options, the US and the EU should've realized the only way to defuse this situation was to negotiate. Yet Biden categorically ruled out the possibility of formally refusing to
41.
by frogblast 5y ago
In California that is defined as the “Wildland/Urban Interface”. That zone has been subject to much stricter building codes since 2008. Timber frame homes are allowed, as other factors (venting, roof material, rain gutters, roof ma
42.
by frogblast 5y ago
You can always run Linux in a VM too
43.
by frogblast 5y ago
In swift: // a is a UInt64 let a = Int.random(in: 0..<Int.max) // causes a fatal runtime error if out of range, halts execution let b = UInt32(a) // returns a UInt32?, whic
44.
by frogblast 5y ago
I have a 2 year old Niro EV with 20K miles. It is a fantastic little car, no issues or defects with it all so far. The dealership experience sucked, but that’s not the car’s fault.
45.
by frogblast 5y ago
Are there any good resources out there describing in practice how existing training workloads are distributed among GPUs? (using tensorflow, pytorch, or whatever else?). I'm curious how the problem effectively gets sliced.
46.
by frogblast 6y ago
This is fearmongering. The bill was introduced a year ago, went nowhere, and is now dead. This WSJ article was published a month after the bill was officially killed… The broader legislature has no interest in such a tax. This is the bi
47.
by frogblast 6y ago
Covid’s timing was devastating for Boeing and the 737 Max. When the planes were grounded, Boeing kept producing them assuming the issue would be resolved quickly. As of January 2020, slightly before Covid, they had 400 planes built, but no
48.
by frogblast 6y ago
Apple’s core is very wide. But wide usually means very high power consumption. M1 apparently has much lower power consumption. So something else is going on other than “Apple is willing to spend more to make cores wider”.
49.
by frogblast 6y ago
Right, lots of confusion in this thread. A12Z/DTK: HW does not support virtualization at all. Apple M1 / New Apple Products: HW does support virtualization for ARM64 guests (both windows and linux demonstrated). What about x86 s
50.
by frogblast 6y ago
The primary feature of Swift is that it a more safe language (via many of the same mechanisms in languages you referred to above), while being extremely compatible with the extremely large API surface area of iOS/macOS (which is both i
51.
by frogblast 6y ago
I'm not involved in firefighting, I just spend a lot of time with maps planning out hikes: I find that paper maps are dramatically better for me than anything on a screen, simply because they can be so much bigger and higher resoluti
52.
by frogblast 6y ago
I live in an area strongly impacted by Google making ‘individually optimal’ decisions for each driver, and actually leaving those drivers in a dramatically worse situation. I live in a rural area, between a major population center and a maj
53.
by frogblast 6y ago
I’m not following the privacy issue: The UI shown in the screenshot requires explicit opt-in to which app is launched, so the user has consented to that app receiving the image that is taken? I’m also not immediately seeing a non-privacy
54.
by frogblast 6y ago
This is what I don't get. If the solution is "install a AOA Disagree light, and disable MCAS in that scenario", why has it taken two years to develop this solution?
55.
by frogblast 6y ago
A contactless card just proves physical possession of the card. An Apple Pay payment means that a password or biometric authentication was performed, AND possession of the device, which raises the barrier to fraud. So there is a good reas
56.
by frogblast 6y ago
You're correct. For 3rd-party provided dylibs, it also isn't necessary to get them into the system cache. dlopen() will look for the dylib first (as today), and if the file doesn't exist, it'll look in the shared cache.
57.
by frogblast 6y ago
The 1080 and the rumored 3080 are not the same architecture: they didn't just get die shrunk. For the same number of SMs, you're also paying substantial area for the tensor cores, ray tracing acceleration, independent thread sched
58.
by frogblast 6y ago
The tensor cores are basically more efficient because they don't implement a 'read from register file; execute one instruction; write to register file'-style architecture: They are basically a big array of ALUs hard-wired in