Branches are particularly bad on GPUs. In that SIMD model of computation you usually have multiple hardware blocks (called threads but unrelated to CPU threads) implementing the same instruction on different data in parallel. Since some of the threads will evaluate the branch condition to true and others false, and they are all operating in lockstep, that means the hardware needs to execute both branches every time, with some of the threads just sitting doing nothing while the other threads execute the "if", and visa-versa while the others execute the "else".

At least that was the case when I was doing CUDA 5+ years ago. Don't know if things have changed since then.

Vorbis was significantly better quality than MP3 and roughly on par with WMA and AAC. But yeah everything but MP3 had an uphill battle, and without a big company to push it, vorbis only ever had limited support on portable players.

FLAC was completely uncompetitive for portable players at the time because it uses like 5-10x the space than high quality AAC or Vorbis encoding that could not be distinguished in side-by-side listening tests with high quality audio equipment. FLAC was a good format for CD rips and online purchases because it was lossless, but if you wanted to take your music with you had to reencode with something else.

It was at least a decade after the first iPod until storage on portables became large enough that 5-10x wasn't worth worrying about and lossless encoding become obsoleted by convenience.

No, adding friends is opt-in, but they changed what information is shared with friends, and that change was opt-out and done without notice.

This mission is part of a program that is all about low-cost projects that are willing to take on more risk. The total cost of the mission was $80 million including $20 million for launch. A Falcon 9 launch costs around $70 million, which would have been more expensive than all the other hardware and operations costs combined. So if it works NASA get a great bargain, and whether it works or not they are helping develop competition in the launch market.

That is one of four factors *considered* in determining fair use. Something can be radically transformative and still be considered copyright infringement if it fairs poorly on the other factors. The effect upon the work's value is a big one when talking about ML, particularly where the value the model provides is competing directly with the works used to train it (for example, stock art).

Sure, the RX 7600 is basically is on par with the RX 6650 XT, and doesn't really push the midrange significantly forward. But it is launching at an MSRP that is $130 less than the RX 6650 XT launch price. Attracting customers that skipped the last two generations due to price is more important to the video card manufacturers than providing an increase in capability for those who purchased a card recently, especially in the mid-range.

The $4.5 million of the criminal judgement may be classified as restitution (I haven't read the decision), but I don't think you can say the same for the $10 million civil settlement. In that case Nintendo was asking for actual, statutory, and punitive damages, and the actual damages were significantly smaller that the other two. Since this settlement is an agreement to cover all of the claims of the lawsuit, it is fair to say that most of the amount is punishment not restitution.

Thanks, I wouldn't have thought to look there. I disabled the voice assistant first thing when setting up a pixel 6, and have been confused about why the power button did absolutely nothing on this phone. Even the power + volume buttons didn't bring up a menu - maybe because the assistant is supposed to capture and relay the button press? The only power down option was to hold it for about 15 seconds until it forcible rebooted.

Hence why they tried to make it about a violation of the ToS rather than copyright, since they know they don't have standing when it comes to copyright.

So far the largest attempt at writing a Linux driver in Rust has been the Asahi Apple M1 GPU driver. That has a few hundred lines of unsafe code* out of about ten thousand lines total. The author said this has both decreased the number of memory corruption bugs that they would expect in a new driver this size and made them much easier to track down when they did occur.

* Not counting common rust wrapper code not written for this driver. There are a few thousand lines of unsafe code there, not all of which are used by the driver, and which in the long run will be amortized across many drivers.

There wasn't a single sentence in that article mentioning Musk, let alone giving him credit for this. It was all attributed to SpaceX in general, as it should be.

Building a Raptor a day isn't an individual effort. Back when SpaceX was a small startup building Falcon 1 it was not hard to find articles crediting individual engineers for the work they did, like Tom Mueller, who designed the Merlin engine used on Falcon 1, and lead all the improvements made to for the Falcon 9. But when people try to give him credit for the Raptor he pushes back. To paraphrase (don't have the book in front of me): "I may have made the very initial designs, but it has been reworked so much there is little left for me to take credit for. I assembled a team that has done incredible work on the Raptor - that is all I can take credit for." There are literally thousands of people working on Raptor at this point. Crediting it to any individuals would be a misattribution.

They are saying that it was overkill for the specific mission requirements that were put out for bid. Other proposals were less risky from a technical point of view while still accomplishing the job. That is a non-controversial assessment of the proposals, and something that needs to be taken into consideration when awarding a contract. But it isn't the only thing that needs to be considered, and NASA decided that SpaceX's superior track record of meeting technical challenges balanced that risk.

I very much prefer smaller mass market paperback books over the larger trade paperback format that everything is printed on these days. For older books, I go out of my way to find used copies printed in the 90â(TM)s or earlier rather than recent printings since they are so much more compact. Would love to be able buy new books in that size more often.

Background Information:
https://news.slashdot.org/stor...
https://yro.slashdot.org/story...
https://news.slashdot.org/stor...

In addition, and contrary to the summary, at the start of the pandemic they decided to start "lending out" unlimited copies, beyond the number of books they had actually purchased. That is what triggered the lawsuit, and I really don't see how they are going to justify that as being legal. I only hope that stupid stunt doesn't destroy all the good things that archive.org has built up over the years.