I actually wouldn't mind buying a usb doohickey that has an accelerometer that sends info to a centralized place to detect and report quakes. I dont, however, think I want to spend battery-life on my phone for it. I mean, wouldn't it need to frequently check the GPS?
I think cell-tower positioning would generally be good enough. If you have thousands of users then the positioning errors will tend to cancel each other out. You probably don't need to update the position very often either, since most people spend most of their time not moving very fast or very far.
I believe the main problems with regards to battery use are: 1. Recording a continuous stream of accelerometer samples without keeping the CPU awake. 2. Measuring the power ad other relevant characteristics of the accelerometer signal without using the phone's main CPU, only waking up the CPU if the signal meets your criteria, so that you can pipe interesting signals (and only interesting signals) over the network to your servers for further analysis. 3. Find a good match between false positives and false negatives so that you don't make network requests too often. The cellular radio on your phone is quite power hungry. The cellular radio in transmission mode is a lot hungrier than the GPS.
Basically, you need to be able to run code on a low power CPU/MCU that the accelerometer delivers data to and then have it wake up the main CPU when you detect something that looks like an earthquake signal. If you're lucky the manufacturer already has that code in place for you, so that you can just tell the system to wake itself up if it detect such and such a signal.
It really depends what you're trying to do. Amdahl's law will always apply. A single fast core will always be more versatile than multiple slower cores because the fast single core can perform equally well in sequential and parallel algorithms. Of course, we will always have to compromise and make do with what we can physically make.
If each step in your algorithm depends on every single preceding step then you will probably want the fastest single core you can get your hands on within a reasonable budget in terms of money, power, etc.
If your algorithm contains lots of simple calculations that can be done in a massively parallel fashion, then you will want a huge number of simple computation units that can run in parallel at low voltage and low clock.
If your algorithm is of the former type, but you're running a large number of them at any given time then you will potentially benefit from having lots of cores running at medium voltage and clock speed.
If you don't know (or don't yet know) what your algorithm is you might benefit from programmable logic, despite its lower performance at any given algorithm.
We can hope that the market will be able to sustain multiple technologies with different trade-offs. It would be nice if we could continue for decades to come to have a line of super-fast quad-core CPU:s for workstations and gaming rigs and a line of medium speed many-core CPU chips for servers and a line of massively parallel chips (basically GPGPU:s) for other stuff and a line of high performance FPGA:s and other programmable logic.
But more than anything of all it would be... interesting... to see some frequency scaling again. I wonder what sort of design and what sort of manufacturing method one would need.
I don't know about the others, but George Carlin was a fairly politically correct comedian. He almost always prefaced and qualified anything non-PC that he said in his routines.
Actually, I just gave it a try, and there is a problem with where you place your defer block.
The following program compiles and the value of the string at the end of execution is "newValue", not "newerValue" as the programmer probably intended.
I think the defer block will always run as the last thing that happens before its containing scope finishes. So as long as you're just writing a function that opens a resource and does not hand that resource over to anything else you should get a very predictable behaviour regardless of what you do inside that function.
The placement of the defer block itself probably doesn't matter. I suppose it would give you a compile time error if you place it before the resource is declared.
I guess defer is only every useful in a function that creates a resource R, uses R itself and does not share R with any code that can run at a later time.
That last requirement might be a little tricky to check for at compile time.
Google Glass was fundamentally flawed in that it did not have a way to block outside light, which means that it basically only worked well in low light conditions.
Ideally, an AR device should be able to filter outside light on a per-pixel basis, so that each drawn pixel on the screen can be mixed with anything from 0% to 100% of the light that comes in from the outside.
The problem is that they don't seem to know where they're going.
The new MacBook should have had not one, but zero ports, because the MacBook is a home device. You're not going to use it for anything that legitimately needs a wired connection. Wired charging? That's barbarism. It should also have been water resistant, because I might want to keep it around when I have a beer which I might spill. It should have had the best webcam on the market. Pricing should have been about $900 for the lowest end model.
The 2015 MacBook Pro is the model that should have had two USB-c ports, in addition to its other ports. The MacBook Pro is power-hungry enough that it probably needs wired charging too, but that's fine sine it's often going to be wired to things in the office (or home office) anyway.
The Pencil for the iPad Pro should also have had a better charing solution. And the iPad Pro is not really a professional device, so it should have been named something else. And the lowest end model should have been at least a hundred bucks cheaper and it should have shipped with the Pencil, because optional add-ons for a device always fragment the software market for that device, which is a very bad thing.
This is all very impressive, but the reason why USAF wants better rockets is probably nothing to do with Mars. They probably want cheaper and better satellites and the ability to put heavy weaponry in space, such as tungsten rods with retrorockets and guidance. That old cold war idea.
The Mars trip will happen when there is a compelling reason for people to go there, like if someone other than the US tries to get there first.
A phone app can use an complicated technology to do something, which pc apps can do without any problems all the time.
Yeah.
An Android or iOS app can silently download a patch that will do bad things with the data that you have created using that particular app. It can also spy on you if you have previously given it permission to do so.
A PC app can silently download a patch that will do bad things with all of your data. It can also spy on you.
How do you have "advanced" mathematics ( or perhaps a better term might be "non-trivial" ) without at least a rudimentary writing system?
You can't. You can do a lot of basic arithmetic and basic geometry.
But you could for example come up with the hypothesis that stars are faraway suns, just by noticing that different stars vary in brightness and guessing that the brighter ones are closer to Earth, with the Sun being much closer than all the others. You could argue that spherical objects are more efficient than other objects because they minimise both their surface area and the distance of any surface feature to the center of themselves for a given volume, then pose the hypothesis that the universe likes to be efficient about things and then conclude that the Earth is therefore probably spherical, like the Moon and the Sun.
It does take writing and some instruments to prove beyond all doubt that these things are true, but the ideas themselves could have been dreamt up by the same people who painted lions on cave walls 30,000 years ago.
Parse is/was a service that allowed people who knew next to nothing about server programming to cobble together a backend for a mobile app (or other app that can make http requests).
So in other words it was a tremendously useful or harmful service depending on your level of cynicism. It is of course hard to monetise a solution like this, since any app that becomes highly profitable will attract developers that know how to build a proper backend and then that app will migrate away and stop paying monthly fees.
Civilizations tend to "discover" philosophy, mathematics, literature, drama and great works of music in the centuries after they invent ways of writing those things down.
What's probably going on is that these things have been cropping up intermittently for thousands of years (or tens of thousands of years), but the ideas would usually not survive for very long because it would take unreasonable amounts of human effort to remember and transmit them.
By the way, video finally made it possible to commit dancing to permanent media in the early 1900's, so future historians will probably think of the 1900's and 2000's as the centuries when great dancing was first invented.
I googled "Hubble replacement" and found that the scientists themselves have apparently taken to naming their proposal for a next-gen optical spectrum telescope the Advanced Technology Large-Aperture Space Telescope, ATLAST.
The previous times that they have attempted to land on a drone ship have all failed due to failures that happened earlier in the descent, so they haven't been able to test how the touchdown itself will work at sea, at least not before today.
The live stream froze just as a yellow reflection (rocket exhaust most likely) became visible on the surface of the ship, which probably means that the rocket was no more than 50 meters or so from the ship at that time. So it seems plausible that it hit the ship. I imagine SpaceX has recovered footage from the ship by now unless the antenna got hit by debris from the explosion.
The fuel and the tank is quite fragile and at least one of the engines is extremely hot and located near the fuel tank, so unless the landing is perfect the tank will burst and the fuel will ignite. I expect we'll see some fiery footage within a day or two once they've had time to analyse it internally.
Yeah, early this week, sometime before Thursday, or at least on Thursday. That is unless there's a lot of unexpected work before then in which case there is always Friday.
The thing is, it no longer matters how much money we throw at making CPU:s and GPU:s do more computations per unit of energy. The laws of physics and the manufacturing technique (lithography) dictate that 4 or 5 GHz is pretty much the limit.
There are people who want faster CPU:s and GPU:s for crazy things like virtual reality games, but they're not getting faster chips anytime soon, so those ideas will have to wait, or will have to be adjusted down to what the hardware can do. Maybe it's okay for VR games to be cartoony and pixelated.
Well, most people only want to use computers for the things that currently available computers can do. That has been true for as long as I can remember.
I remember thinking that a 50 MHz 486 was easily fast enough for everything that one could possible want to do with a personal computer, except for a few things that crazy people suggested we should do, like watching TV-quality videos, or running software in a web browser. Soon enough, Intel launched faster CPU:s. Some of those crazy people wrote software that couldn't quite run on the old 486's boxes and that software turned out to be something that people wanted to use and then people switched to the faster CPU:s. This process continued until Intel and AMD reached the point where they could no longer make faster CPU:s.
Slashdot Mirror
I actually wouldn't mind buying a usb doohickey that has an accelerometer that sends info to a centralized place to detect and report quakes. I dont, however, think I want to spend battery-life on my phone for it. I mean, wouldn't it need to frequently check the GPS?
I think cell-tower positioning would generally be good enough. If you have thousands of users then the positioning errors will tend to cancel each other out. You probably don't need to update the position very often either, since most people spend most of their time not moving very fast or very far.
I believe the main problems with regards to battery use are:
1. Recording a continuous stream of accelerometer samples without keeping the CPU awake.
2. Measuring the power ad other relevant characteristics of the accelerometer signal without using the phone's main CPU, only waking up the CPU if the signal meets your criteria, so that you can pipe interesting signals (and only interesting signals) over the network to your servers for further analysis.
3. Find a good match between false positives and false negatives so that you don't make network requests too often. The cellular radio on your phone is quite power hungry. The cellular radio in transmission mode is a lot hungrier than the GPS.
Basically, you need to be able to run code on a low power CPU/MCU that the accelerometer delivers data to and then have it wake up the main CPU when you detect something that looks like an earthquake signal. If you're lucky the manufacturer already has that code in place for you, so that you can just tell the system to wake itself up if it detect such and such a signal.
Yeah, you could brick the device, or you could pop up a warning on each boot:
"Third-party hardware or hardware tampering detected! Touch ID has been disabled."
Gee, I wonder which solution the customer would prefer!
It really depends what you're trying to do. Amdahl's law will always apply. A single fast core will always be more versatile than multiple slower cores because the fast single core can perform equally well in sequential and parallel algorithms. Of course, we will always have to compromise and make do with what we can physically make.
If each step in your algorithm depends on every single preceding step then you will probably want the fastest single core you can get your hands on within a reasonable budget in terms of money, power, etc.
If your algorithm contains lots of simple calculations that can be done in a massively parallel fashion, then you will want a huge number of simple computation units that can run in parallel at low voltage and low clock.
If your algorithm is of the former type, but you're running a large number of them at any given time then you will potentially benefit from having lots of cores running at medium voltage and clock speed.
If you don't know (or don't yet know) what your algorithm is you might benefit from programmable logic, despite its lower performance at any given algorithm.
We can hope that the market will be able to sustain multiple technologies with different trade-offs. It would be nice if we could continue for decades to come to have a line of super-fast quad-core CPU:s for workstations and gaming rigs and a line of medium speed many-core CPU chips for servers and a line of massively parallel chips (basically GPGPU:s) for other stuff and a line of high performance FPGA:s and other programmable logic.
But more than anything of all it would be... interesting... to see some frequency scaling again. I wonder what sort of design and what sort of manufacturing method one would need.
I don't know about the others, but George Carlin was a fairly politically correct comedian. He almost always prefaced and qualified anything non-PC that he said in his routines.
Actually, I just gave it a try, and there is a problem with where you place your defer block.
The following program compiles and the value of the string at the end of execution is "newValue", not "newerValue" as the programmer probably intended.
var someString = "oldValue"
func deferTest() {
someString = "newValue"
return
defer {
someString = "newerValue"
}
}
deferTest()
someString
I think the defer block will always run as the last thing that happens before its containing scope finishes. So as long as you're just writing a function that opens a resource and does not hand that resource over to anything else you should get a very predictable behaviour regardless of what you do inside that function.
The placement of the defer block itself probably doesn't matter. I suppose it would give you a compile time error if you place it before the resource is declared.
I guess defer is only every useful in a function that creates a resource R, uses R itself and does not share R with any code that can run at a later time.
That last requirement might be a little tricky to check for at compile time.
I have an iPad Mini and a MacBook Pro, so I'm not really in the market for a small home-class laptop.
Google Glass was fundamentally flawed in that it did not have a way to block outside light, which means that it basically only worked well in low light conditions.
Ideally, an AR device should be able to filter outside light on a per-pixel basis, so that each drawn pixel on the screen can be mixed with anything from 0% to 100% of the light that comes in from the outside.
Someone cares about it for some reason. I don't think anyone, even the developers themselves, are claiming that Hurd will be useful in and of itself.
If you ask me I suspect it's a long lead-up to a joke about playing DNF on Hurd...
The problem is that they don't seem to know where they're going.
The new MacBook should have had not one, but zero ports, because the MacBook is a home device. You're not going to use it for anything that legitimately needs a wired connection. Wired charging? That's barbarism. It should also have been water resistant, because I might want to keep it around when I have a beer which I might spill. It should have had the best webcam on the market. Pricing should have been about $900 for the lowest end model.
The 2015 MacBook Pro is the model that should have had two USB-c ports, in addition to its other ports. The MacBook Pro is power-hungry enough that it probably needs wired charging too, but that's fine sine it's often going to be wired to things in the office (or home office) anyway.
The Pencil for the iPad Pro should also have had a better charing solution. And the iPad Pro is not really a professional device, so it should have been named something else. And the lowest end model should have been at least a hundred bucks cheaper and it should have shipped with the Pencil, because optional add-ons for a device always fragment the software market for that device, which is a very bad thing.
This is all very impressive, but the reason why USAF wants better rockets is probably nothing to do with Mars. They probably want cheaper and better satellites and the ability to put heavy weaponry in space, such as tungsten rods with retrorockets and guidance. That old cold war idea.
The Mars trip will happen when there is a compelling reason for people to go there, like if someone other than the US tries to get there first.
A phone app can use an complicated technology to do something, which pc apps can do without any problems all the time.
Yeah.
An Android or iOS app can silently download a patch that will do bad things with the data that you have created using that particular app. It can also spy on you if you have previously given it permission to do so.
A PC app can silently download a patch that will do bad things with all of your data. It can also spy on you.
It takes more like 2-4 hours before the update is rolled out by Google. I believe they run some tests on your update before they let it through.
How do you have "advanced" mathematics ( or perhaps a better term might be "non-trivial" ) without at least a rudimentary writing system?
You can't. You can do a lot of basic arithmetic and basic geometry.
But you could for example come up with the hypothesis that stars are faraway suns, just by noticing that different stars vary in brightness and guessing that the brighter ones are closer to Earth, with the Sun being much closer than all the others. You could argue that spherical objects are more efficient than other objects because they minimise both their surface area and the distance of any surface feature to the center of themselves for a given volume, then pose the hypothesis that the universe likes to be efficient about things and then conclude that the Earth is therefore probably spherical, like the Moon and the Sun.
It does take writing and some instruments to prove beyond all doubt that these things are true, but the ideas themselves could have been dreamt up by the same people who painted lions on cave walls 30,000 years ago.
Parse is/was a service that allowed people who knew next to nothing about server programming to cobble together a backend for a mobile app (or other app that can make http requests).
So in other words it was a tremendously useful or harmful service depending on your level of cynicism. It is of course hard to monetise a solution like this, since any app that becomes highly profitable will attract developers that know how to build a proper backend and then that app will migrate away and stop paying monthly fees.
Civilizations tend to "discover" philosophy, mathematics, literature, drama and great works of music in the centuries after they invent ways of writing those things down.
What's probably going on is that these things have been cropping up intermittently for thousands of years (or tens of thousands of years), but the ideas would usually not survive for very long because it would take unreasonable amounts of human effort to remember and transmit them.
By the way, video finally made it possible to commit dancing to permanent media in the early 1900's, so future historians will probably think of the 1900's and 2000's as the centuries when great dancing was first invented.
A bit aged?
I googled "Hubble replacement" and found that the scientists themselves have apparently taken to naming their proposal for a next-gen optical spectrum telescope the Advanced Technology Large-Aperture Space Telescope, ATLAST.
Not likely, since provocation studies failed to show that electro sensitive people are sensitive to electromagnetic fields.
"Tonight on the History Channel."
But where does Hitler come into it?
Did Hitler attend a secret meeting with alien representatives in an underground meeting room under Washington DC in 1929?
The previous times that they have attempted to land on a drone ship have all failed due to failures that happened earlier in the descent, so they haven't been able to test how the touchdown itself will work at sea, at least not before today.
The live stream froze just as a yellow reflection (rocket exhaust most likely) became visible on the surface of the ship, which probably means that the rocket was no more than 50 meters or so from the ship at that time. So it seems plausible that it hit the ship. I imagine SpaceX has recovered footage from the ship by now unless the antenna got hit by debris from the explosion.
The fuel and the tank is quite fragile and at least one of the engines is extremely hot and located near the fuel tank, so unless the landing is perfect the tank will burst and the fuel will ignite. I expect we'll see some fiery footage within a day or two once they've had time to analyse it internally.
I would rtfa, but I'll do it later.
Yeah, early this week, sometime before Thursday, or at least on Thursday. That is unless there's a lot of unexpected work before then in which case there is always Friday.
Definitely getting to it.
The thing is, it no longer matters how much money we throw at making CPU:s and GPU:s do more computations per unit of energy. The laws of physics and the manufacturing technique (lithography) dictate that 4 or 5 GHz is pretty much the limit.
There are people who want faster CPU:s and GPU:s for crazy things like virtual reality games, but they're not getting faster chips anytime soon, so those ideas will have to wait, or will have to be adjusted down to what the hardware can do. Maybe it's okay for VR games to be cartoony and pixelated.
Well, most people only want to use computers for the things that currently available computers can do. That has been true for as long as I can remember.
I remember thinking that a 50 MHz 486 was easily fast enough for everything that one could possible want to do with a personal computer, except for a few things that crazy people suggested we should do, like watching TV-quality videos, or running software in a web browser. Soon enough, Intel launched faster CPU:s. Some of those crazy people wrote software that couldn't quite run on the old 486's boxes and that software turned out to be something that people wanted to use and then people switched to the faster CPU:s. This process continued until Intel and AMD reached the point where they could no longer make faster CPU:s.