> This intel will have 56 cores and each core presumably has 4 four-byte simd channels. It will likely hyperthread (maybe not) and have pipelined instructions and predictive branching and larger caches.
Intels have supported AVX512 for a couple generation now. So ech core will probably be able to do 2 512-bit FMA per clock cycle.
The main difference in speed between CPUs and GPUs has been in the memory subsystem more than flops. (And also on programmability CUDA is much easier to write than AVX code...)
I am well involved in organizing scientific conferences. We organize in the US most of the time but went to Canada last year. And we had WAY more attendees that could not get a visa than any other year on record. In particular north african and middle eastern attendees were the most impacted. So it may not be true that getting through visa and emigration is going to be easier in Canada than it is in the US. Though I have only the data point of a single conference.
Maybe we need to have that little talk - you know: about where electricity comes from.
If the electricity comes from a diesel, or worse yet, coal plant, then I doubt there is any environmental benefit.
While that's a good question I disagree with you. It is a lot simpler to have particle filtering at the plant rather than in many buses. Also it makes your energy consumption depend on the grid which gives you a choice of burning diesel, or natural gas, or use a renewable when it is there.
Of course there is some loss in transport and battery storage, but you have also a gain of efficiency because it is simpler to have one efficient plant than 1000 efficient buses engines.
The release note does not seem to mention OpenMP or OpenCL. The release note does not contain the string "OpenMP" or the string "OpenCL". Am I missing something obvious? What has actually changed in that regard?
A gave a shot to the steam link at home. And essentially, you can forget it over wifi. So good luck getting it to work on the go; where network connectivity is typically terrible.
In the future, I could imagine that steam would be bundled on smart TVs that you find in hotel rooms and if they have decent internet and are wired to the network, that it could work.
Same here. I went ahead and read a bunch about LASIK. First, the process is terrifying. Then the potential side effect are not so uncommon and pretty bad. I can stick with glasses.
I had tried the hololens 1 that a colleague use in her VR/AR research. And I liked Hololens. Yes it was a bit heavy, yes the field of view was pretty narrow. But for a version 1, that was pretty good AR. Much better than the VR headset I had tried at the time.
So I hope I'll get to try one out and see how better this is going to be. Probably one of my colleagues will get one, and I'll play with it.
I teach computer science at a state school. And some students laptop can not quite run VMs, for a bunch a reasons (that range from shitty hardware, to not enough memory, and through the virtualization bit got disabled in the BIOS because the constructor wanted to sell a more expensive laptop for developpers).
I had used WSL for the class. And that was a complete catastrophe. The filesystem interaction were just not working right. I am guessing it is because students went editting the file through the file explorer. But we were getting the strangest bugs.
So I have told students to absolutely use the VM and we are working around the few cases of laptops that are not quite good enough. And the clas goes fine so far.
If MS can deliver a working WSL, I may consider using it again.
It really depends who are the target users of textbook. As a first time introduction to a topic, textbooks tend to be a little rough. But as collections of carefully curated piece of knowledge, they are valuable.
There is a need for a carefully written description of a concept and the various versions of it, with 200 exercises some of them with solutions. That is what a textbook is. Whether it is print or not is an orthogonal issue. OpenDSA ( https://opendsa-server.cs.vt.e... ) is a pretty good online textbook for data structures and algorithms for example.
Now, there is value to a printed copies of some books and there is value to electronic ones depending on you usage of them.
We can't replace all our textbook with 3 minute video discussion of the topic. Such videos do not get in enough details and can not be precise enough and can not provide with dozens of alternate versions. But they can be a pretty good introduction to a topic.
We should replicate studies and confirm analysis independently before we get too excited about any results. It is hard to run this kind of studies, there are lots of variable to consider, lots of potential misreported event. We need to be careful. I don't mean that the authors are always lying on purpose, but they could have missed something important, they could have made an error. Peer review does not quite catch these things.
Any kind of study should be taken with a grain of salt until it is replicated in multiple place.
And I say that as a computer scientist. There are things that appear to make a lot of sense when you describe them. It does not always mean they will work. And sometimes it works in one paper, and not in other ones. Machines are never quite the same, the instances could have slightly different characteristics. Lots of things can happen. We need to be careful.
Of course, the media just loves a good headline. So they'll print pretty much anything to sale some papers.
And this my friends is exactly why I am not a big fan of the argument that we should keep taxes low and rely on charities to pay for stuff. Once the money is given by companies, or billionaire, their donations become concerns for any negotiations. Tax them and where the money is spent is no longer their decision but the public's decision.
The style looks about right! The guest book was a nice touch. Though I don't remember guestbook needing "sign in", they usually just let you post whatever, sometime required manual moderation.
But, that's a lot of javascript for the 90's. And that "one page" format is very modern. All these things would have been on different pages.
Well, it does not really matter which one you use in this kind of analysis. U3 and U6 are very correlated. So the analysis you do using U3 will most likely be still correct using U6.
The media and Department of labor quotes U3 because it is the metric that resembles the most what other countries are measuring when reporting unemployment. The media do not report all metrics because there is such a high correlation between them that the story is the same. It is also the metric that is the closer to a natural understanding of what it means to be unemployed: currently not working and looking for a job.
There is a good case for tracking U6 to do various kind of macro economic analysis. But if you are the kind of person that does detailed analysis you know that they are quoting U3 and not U6; so it is not quite confusing.
I am not too sure if counting U5 and U6 as the real unemployment makes a lot of sense. I knew a lot of people that would count in U6 in the sense that they said they wanted a full time job. And then they found one and quit it within two month because it is too much work.
I know a lot of people that count in U5 that claim they want a job but in practice are very happy that their spouse brings home the bacon: "Honey, I'd love to work, look I edited my resume 3 weeks ago; but you know with the crisis it is hard!"
Why not, the measure is in "do you know this language?" Java was the language most chosen in CS1. And bootcamp companies trying to place people in software contracting firm were mostly teaching Java for 15 years. It is not surprising that Java was up there in the "do you know this language?" ranking.
Personally, I have always taken this kind of online rating with a 20% confidence interval. I use to read IMDB scores like that: >9 people got super excited about it. Probably the movie does not deserve that score but that also means it is probably going to become some kind of a cultural reference. So you probably want to watch it. >8 There is a consensus that the movie is decent. >7 People who are not particularly about that style of movie liked it fine. It is a good bet you'll like it too. >6 About only the people enthusiastic about that type of movie found it enjoyable. Approach with care. https://www.imdb.com/title/tt0... Titanic https://www.imdb.com/title/tt0...
Popular musicians are pirated and they sell more concert tickets.
Popular movies may be pirated, but the sequel will sell more entries. (Or you'll listen to the on Youtube and get some revenue from there.)
The show may be pirated, but you'll sell the t-shirts.
Almost all entertainment products have multiple revenue streams. Maybe the primary product you don't sell, but you'll sell derivative products. Or you'll sell to the same people 5 years down the road once they'll have the income to buy it.
Slashdot Mirror
I know it is completely out of topic. But yeah it is terrible.
I grew up in Paris, I have been bummed out all afternoon.
> With AVX instructions, I believe that each core can perform 32 fused add/multiply operation per clock cycle.
With AVX-512, architectures should have 32 FMA per cycle single precision. So that should be 7TFlops single precision.
Intel was playing with half precision, I wonder they are going to go that route and give us 14TFlops half precision.
> This intel will have 56 cores and each core presumably has 4 four-byte simd channels. It will likely hyperthread (maybe not) and have pipelined instructions and predictive branching and larger caches.
Intels have supported AVX512 for a couple generation now. So ech core will probably be able to do 2 512-bit FMA per clock cycle.
The main difference in speed between CPUs and GPUs has been in the memory subsystem more than flops. (And also on programmability CUDA is much easier to write than AVX code...)
I am well involved in organizing scientific conferences. We organize in the US most of the time but went to Canada last year. And we had WAY more attendees that could not get a visa than any other year on record. In particular north african and middle eastern attendees were the most impacted.
So it may not be true that getting through visa and emigration is going to be easier in Canada than it is in the US. Though I have only the data point of a single conference.
Maybe we need to have that little talk - you know: about where electricity comes from.
If the electricity comes from a diesel, or worse yet, coal plant, then I doubt there is any environmental benefit.
While that's a good question I disagree with you. It is a lot simpler to have particle filtering at the plant rather than in many buses.
Also it makes your energy consumption depend on the grid which gives you a choice of burning diesel, or natural gas, or use a renewable when it is there.
Of course there is some loss in transport and battery storage, but you have also a gain of efficiency because it is simpler to have one efficient plant than 1000 efficient buses engines.
The release note does not seem to mention OpenMP or OpenCL. The release note does not contain the string "OpenMP" or the string "OpenCL". Am I missing something obvious? What has actually changed in that regard?
A gave a shot to the steam link at home. And essentially, you can forget it over wifi. So good luck getting it to work on the go; where network connectivity is typically terrible.
In the future, I could imagine that steam would be bundled on smart TVs that you find in hotel rooms and if they have decent internet and are wired to the network, that it could work.
Same here. I went ahead and read a bunch about LASIK. First, the process is terrifying. Then the potential side effect are not so uncommon and pretty bad. I can stick with glasses.
Or maybe she is a maniac instagrammer so she has pictures of every hour of her life?
Main stream VR and AR is still 10 years away due to cost/quality combination.
VR is already a (small) market in video games.
AR is still not quite there. But that's why they target at particular industries at this point.
I had tried the hololens 1 that a colleague use in her VR/AR research. And I liked Hololens. Yes it was a bit heavy, yes the field of view was pretty narrow. But for a version 1, that was pretty good AR. Much better than the VR headset I had tried at the time.
So I hope I'll get to try one out and see how better this is going to be. Probably one of my colleagues will get one, and I'll play with it.
Bruce Schneier, thank you for the fish!
I teach computer science at a state school. And some students laptop can not quite run VMs, for a bunch a reasons (that range from shitty hardware, to not enough memory, and through the virtualization bit got disabled in the BIOS because the constructor wanted to sell a more expensive laptop for developpers).
I had used WSL for the class. And that was a complete catastrophe. The filesystem interaction were just not working right. I am guessing it is because students went editting the file through the file explorer. But we were getting the strangest bugs.
So I have told students to absolutely use the VM and we are working around the few cases of laptops that are not quite good enough. And the clas goes fine so far.
If MS can deliver a working WSL, I may consider using it again.
There was a good discussion on the vlogbrothers youtube channel (video and comments): https://www.youtube.com/watch?...
It really depends who are the target users of textbook. As a first time introduction to a topic, textbooks tend to be a little rough. But as collections of carefully curated piece of knowledge, they are valuable.
There is a need for a carefully written description of a concept and the various versions of it, with 200 exercises some of them with solutions. That is what a textbook is. Whether it is print or not is an orthogonal issue. OpenDSA ( https://opendsa-server.cs.vt.e... ) is a pretty good online textbook for data structures and algorithms for example.
Now, there is value to a printed copies of some books and there is value to electronic ones depending on you usage of them.
We can't replace all our textbook with 3 minute video discussion of the topic. Such videos do not get in enough details and can not be precise enough and can not provide with dozens of alternate versions. But they can be a pretty good introduction to a topic.
NVidia's GPUs already do that (I assume AMDs too).
It is called unified memory in CUDA. It's been out there for years now.
We should replicate studies and confirm analysis independently before we get too excited about any results. It is hard to run this kind of studies, there are lots of variable to consider, lots of potential misreported event. We need to be careful. I don't mean that the authors are always lying on purpose, but they could have missed something important, they could have made an error. Peer review does not quite catch these things.
Any kind of study should be taken with a grain of salt until it is replicated in multiple place.
And I say that as a computer scientist. There are things that appear to make a lot of sense when you describe them. It does not always mean they will work. And sometimes it works in one paper, and not in other ones. Machines are never quite the same, the instances could have slightly different characteristics. Lots of things can happen. We need to be careful.
Of course, the media just loves a good headline. So they'll print pretty much anything to sale some papers.
Try and live in NYC on $50K.
Indeed, $50,000 seems not enough to survive in New York for a month.
They are not scientists at all. They must be paid by you-know-who.
Voldemort? Maybe we should call Harry Potter!
And this my friends is exactly why I am not a big fan of the argument that we should keep taxes low and rely on charities to pay for stuff.
Once the money is given by companies, or billionaire, their donations become concerns for any negotiations. Tax them and where the money is spent is no longer their decision but the public's decision.
The style looks about right! The guest book was a nice touch. Though I don't remember guestbook needing "sign in", they usually just let you post whatever, sometime required manual moderation.
But, that's a lot of javascript for the 90's. And that "one page" format is very modern. All these things would have been on different pages.
Where is the "webring" banner?
Well, it does not really matter which one you use in this kind of analysis. U3 and U6 are very correlated. So the analysis you do using U3 will most likely be still correct using U6.
The media and Department of labor quotes U3 because it is the metric that resembles the most what other countries are measuring when reporting unemployment. The media do not report all metrics because there is such a high correlation between them that the story is the same. It is also the metric that is the closer to a natural understanding of what it means to be unemployed: currently not working and looking for a job.
There is a good case for tracking U6 to do various kind of macro economic analysis. But if you are the kind of person that does detailed analysis you know that they are quoting U3 and not U6; so it is not quite confusing.
I am not too sure if counting U5 and U6 as the real unemployment makes a lot of sense. I knew a lot of people that would count in U6 in the sense that they said they wanted a full time job. And then they found one and quit it within two month because it is too much work.
I know a lot of people that count in U5 that claim they want a job but in practice are very happy that their spouse brings home the bacon: "Honey, I'd love to work, look I edited my resume 3 weeks ago; but you know with the crisis it is hard!"
Why not, the measure is in "do you know this language?"
Java was the language most chosen in CS1. And bootcamp companies trying to place people in software contracting firm were mostly teaching Java for 15 years. It is not surprising that Java was up there in the "do you know this language?" ranking.
Personally, I have always taken this kind of online rating with a 20% confidence interval. I use to read IMDB scores like that:
>9 people got super excited about it. Probably the movie does not deserve that score but that also means it is probably going to become some kind of a cultural reference. So you probably want to watch it.
>8 There is a consensus that the movie is decent.
>7 People who are not particularly about that style of movie liked it fine. It is a good bet you'll like it too.
>6 About only the people enthusiastic about that type of movie found it enjoyable. Approach with care.
https://www.imdb.com/title/tt0...
Titanic https://www.imdb.com/title/tt0...
Now when you see a movie with more than half a point of shift you know there is something going on:
The ghostbusters all female reboot https://www.imdb.com/title/tt1...
The girl next door https://www.imdb.com/title/tt0...
You'll see movie with a clear age scoring difference. That probably means something too.
please ignore that idiot!
We love you. And we love the diversity you bring to the web market and your commitment to internet freedom.
A faithful user.
Popular musicians are pirated and they sell more concert tickets.
Popular movies may be pirated, but the sequel will sell more entries. (Or you'll listen to the on Youtube and get some revenue from there.)
The show may be pirated, but you'll sell the t-shirts.
Almost all entertainment products have multiple revenue streams. Maybe the primary product you don't sell, but you'll sell derivative products. Or you'll sell to the same people 5 years down the road once they'll have the income to buy it.