brigade

34 ms·

391.
by brigade 5y ago
Overlay support tends to be very limited - it's effectively blitting to the final framebuffer after compositing, which means no transparency and a limited number of surfaces. The desire is for the compositor or any other shaders to be
392.
by brigade 5y ago
The only important bit of pedantry over analog/digital and prime is reminding people that no one deals with linear light unless you already know you're doing so, and that blending in nonlinear space is rarely correct, no matter ho
393.
by brigade 5y ago
https://socialimpact.facebook.com/charitable-giving/live-don... I don't use Facebook so I don't know if that page is misleading, but it certainly looks like you can donate money (in the legal meaning!) throug
394.
by brigade 5y ago
> everyone gets their own connection Technically , everyone on the same cable node shares the same spectrum (typically targeting 200-500 subscribers.) Which is effectively 1GHz at the moment, a decent chunk of which is dedicated to TV.
395.
by brigade 5y ago
I exaggerate, but it seems like they’re focusing on OTT, which provides full turnkey websites and apps you can slap your brand on and then fill with videos. Kinda the polar opposite of YouTube. Personally, I don’t know how much future these
396.
by brigade 5y ago
They don’t want to be anything like YouTube, they want to be a video hosting backend instead.
397.
by brigade 5y ago
> Speculatively, AVX512 processes multiplies serially, one 256-bit lane at a time, losing half its parallelism. Sort of, in Skylake AVX512 fuses the 256-bit p0 and p1 together for one 512-bit µop, and p5 becomes 512-bit wide. So theoreti
398.
by brigade 5y ago
Well for one, the language says a copy is made at the time of the function call , and it's perfectly valid to modify the original before the copy is finished being used. So pretty much any potentially aliasing write or function call i
399.
by brigade 5y ago
That would get exponential real quick. But Intel’s Tremont does have an instruction decoder that’s 6-wide only when decoding a taken branch, since it also starts decoding the branch target in the same cycle.
400.
by brigade 5y ago
Well... when the operation of register renaming itself depends on the vtype register, that isn't a complete solution.
401.
by brigade 5y ago
SVE was designed mindful of how CPUs currently operate, whereas RISC-V vector extensions were designed with fondness for how CPUs operated decades ago. Well that's somewhat of an exaggeration, but XT-910 speculatively executes vector i
402.
by brigade 5y ago
Intel has had PCIe cards targeted at this market, reusing their own HW encoder, e.g. the VCA2 could do up to 14 real-time 4K transcodes at under 240W, and the upcoming Xe cards would support VP9 encode. (XG310 is similar albeit more targete
403.
by brigade 5y ago
The only OSS VP9 encoder is Google’s own libvpx, which is what ffmpeg uses.
404.
by brigade 5y ago
It is two streams but it actually isn’t MST. It’s two independent HBR2 links; MST would have been two streams over one HBR2 link, which doesn’t have the bandwidth for 5k.
405.
by brigade 5y ago
That link must include new registrations for used cars, and exclude trucks. After all, the Focus was discontinued in 2019. All the manufacturers report new US car sales, and except for the RAV4, in 2019 the cars on the list were outsold wi
406.
by brigade 5y ago
Each bit used halves the potential encoding space, e.g. 2^32 -> 2^31 possible instruction encodings. For thumb, 32-bit instructions can be allocated up to 3/32 of the potential 32-bit space, and 16-bit instructions can use 29/3
407.
by brigade 6y ago
On the other hand, did Musk ever get in trouble for his “stock price too high” tweet?
408.
by brigade 6y ago
Each cable node shares ~1GHz of bandwidth among all the separate runs to individual houses. Relative to that, 5G has similar local bandwidth, given the low propagation of mmwave. It's not going to completely replace hardwired internet
409.
by brigade 6y ago
"5nm" is too simplistic; the material shortage continues upstream to substrates and raw wafers that are used across nodes
410.
by brigade 6y ago
The better answer is that the closer an amp's output impedance is to the speakers' impedance, the greater the variation in amplified frequency response. Because impedance varies depending on the frequency of the signal. So an amp
411.
by brigade 6y ago
Strong disagree that it hasn't evolved; modern advances in materials and modeling mean drivers have wider ranges and lower distortion. Modern DSP can extend the low frequency response even further than just porting, and provide a cross
412.
by brigade 6y ago
The blog post links to the benchmark... It's repeatedly populating a 20k entry array with the results. godbolt clang compiles it to: .LBB5_2: // =>This Inner Loop Header: Depth=1
413.
by brigade 6y ago
The dependency chain is state += 0x60bee2bee120fc15ull or (state += UINT64_C(0x9E3779B97F4A7C15)); the rest of the calculations are independent per iteration. Anyway, the more important fact is that 64x64b -> 128b mul might be one instru
414.
by brigade 6y ago
I read somewhere that the crunch to get 14nm fixed burned a lot of goodwill among their top engineers, and when they saw delays looming for 10nm as well they looked for greener pastures rather than go through the 14nm crunch again. But yeah
415.
by brigade 6y ago
You do realize that $1k is about 10% of the median household income tax liability? I can't imagine the majority of people happily accept a 10% (more for half of households) tax hike, when companies already CC the IRS on all the forms t
416.
by brigade 6y ago
Uh no, as defined by Reg T, brokers are free to let cash accounts immediately buy with unsettled funds. It's only a violation if you sell a security before the funds used to purchase it settle. Margin accounts have a different legal
417.
by brigade 6y ago
You're mostly wrong; MST does not increase bandwidth, it splits the bandwidth of a single DisplayPort link. To increase raw bandwidth for 5k/6k they combine multiple HBR2/HBR3 links (over Thunderbolt or with multiple cables),
418.
by brigade 6y ago
Spec-compliant USB-C cables are still required to have the USB2 data lines hooked up, even "power-only" cables, and Android Auto / Carplay don't need beyond USB2
419.
by brigade 6y ago
USB-C cables either have all 4 superspeed lanes wired up, or none of them. Which is why USB-C docks with video worked (2 lanes for USB, 2 lanes for DisplayPort) and why 2x2 isn't supposed to require new cables. Thunderbolt still works
420.
by brigade 6y ago
Intel's TB3 controllers can't do both TB3 daisy-chain and provide two DisplayPort endpoints. That's why most TB3 docks that have two DisplayPort ports don't have a second TB3 port, and why the HP had to use MST to have

More ›