Talking about open-source businesses is missing the point entirely. Most businesses that are successful as a result of open source (or Free Software, for the RMS-style folks) or that contribute significantly to open source are not 'open-source businesses' any more than companies that use Windows and Office are 'closed-source businesses. The difference is that one category of businesses realises that writing software is expensive and copying software is trivial, so spends its investment on the software parts of its infrastructure paying people to write software (typically customising and improving existing projects), whereas the other pays someone for copies of software and hopes that that will give them an incentive to produce software that's more like they want.
You're comparing apples and oranges as far as the technical details. I'm saying Win 3.x let me continue when it saw problems, and NT could also do that.
Not really. The kind of situations where Windows 3.x let you try to continue, Windows NT just handles transparently. In Windows 3.x, with cooperative multitasking, a single application can refuse to relinquish the CPU. If this happens, you have three choices (outlined by the dialog box):
Just wait and see if it eventually recovers.
Kill that application and hope that it isn't holding any handles that other processes need to be able to do useful work.
Restart the entire computer.
In a system with protected memory and preemptive multitasking, an application that refuses to relinquish the CPU will just have its priority downgraded and the only thing that you'll notice is the CPU getting warm. Eventually, you may choose to kill the program, but it never affects system stability.
I'd like to have the *option* to continue to save my work even if there was a chance of data corruption. For example, take the common NT blue screen IRQL_NOT_LESS_OR_EQUAL. That fact that my buggy network driver tried to access paged memory in the wrong sequence is miles away from catastrophic. And it certainly doesn't take priority over something I've been working on for hours. IRQ 0 is me, motherfuckers!
It means that there's a high probability that something has damaged some kernel data structures. If you continue, there's a good chance that this corruption will spread to the buffer cache and you'll end up writing invalid data to disk. If you kill the system, the corruption is limited to the RAM.
Agreed on Chen's blog, but the summary is horrible. This message hasn't been part of Windows since Windows 95 (which introduced preemptive multitasking to the Windows world, so a single application could no longer freeze the system trivially), so the odds are that if you used Windows in the last two decades you've never seen this notice...
The idea is to have a timer that would automatically disable the equipment unless it received an enable signal, either from a satellite or removable medium. It's possible to make such a system that is, at the very least, very difficult to tamper with. Many of the systems on tanks and so on are computer controlled and if the computers stop working then it's a lot less valuable. The goal of such systems is similar to that of crypto: it's not to prevent the enemy from ever using the tanks that they've stolen, it's to prevent them using them quickly. If you have a few weeks to bomb the stolen equipment before it can be used, and the enemy has to invest a lot of high-tech resources into cracking the systems, then that's probably good enough.
MIRVs are a better second-strike weapon. In a first-strike context, your missile bases are all working fine and you can just launch everything that you've got at the enemy. MIRV in a first strike requires fewer missiles to get through, but that just means you need to build fewer missiles, which doesn't reduce the cost by a huge amount compared to the cost of maintaining a first-strike capability at all. In a second-strike situation, however, you're much more likely to have limited launch resources (a few submarines if you're late, the missile silos that you can contact in time if you're not). Being able to have a devastating second-strike capability with just one submarine surviving is a strong deterrent.
Also I'm not sure about your phone but with mine I can disable data over cellular in the swipe down settings. It's literally a swipe and a single click, so if you really want to restrict everything and then only use it as required that would be the easy way to do it system wide and is about as complicated as turning on screen rotation.
You didn't read my post, did you? You can turn it off globally, but it's really hard to then turn it on for a subset of applications. You have to individually disable each one's access (and they all default to 'on', so you have to remember to do it again after installing each new app). If I turn on mobile data somewhere expensive to look something up urgently (or, for example, to get an updated boarding pass for a flight) then suddenly a dozen applications will say 'whee, Internet! Let's download a load of updates!'
Honestly, even if I'm doing it a on a million items, O(N^2) vs O(N) isn't a big deal
If you're doing a million items, then the difference between O(N) and O(N^2) is a factor of a million. If each operation takes one cycle to complete, then on a 2GHz CPU you're going to take 500 microseconds. More plausibly, if each iteration takes 10 cycles, then it will take 5ms. The O(n^2) version, assuming the same ten-cycles per operation (which is pretty small - even a bubblesort on integers in an array will be hard pressed to be that quick), will take one hour and 23 minutes.
If you really think that the difference between 5ms and almost an hour and a half is irrelevant, then I really hope that I never use any code that you write.
De Morgen's laws are widely known and trivial to automate. Any vaguely modern compiler knows about them and will do the transform for you. Not only will it do it for you, it will do it for you everywhere and base on knowledge of the target architecture.
That said, it won't rearrange things that might have side effects. In C, for example, you have short-circuit evaluation, so if you have something cheap on the left of an or and something expensive on the right, then it will be faster unless the compiler can determine that the expensive thing has no side effects. If you have a negation of a long string of ands, the arguments to each of which is expensive to compute, then turning it into a string of negated ors is faster, because the compiler can stop computing once it's found one that's failed.
Exactly what I came here to post. We had the demonstration of what happens when you immerse something in liquid nitrogen vs what happens when you pour it over the top. Even if you didn't get to play with liquid nitrogen in school, there are lots of videos of this.
Some Fortran 66 code that was certified as accurately modelling explosions before the ban, which no one is allowed to modify, run on ever-more-expensive supercomputers.
Or, you put it another way, the most expensive wishful thinking on the planet.
When people say 'first strike' in this context, they mean 'nuclear first strike', as in 'launch enough nuclear missiles to take out your enemy's ability to fire back'. The reason for nuclear submarines is to largely eliminate this possibility: even if you completely destroy every military installation in the USA, there is going to be at least one submarine hidden somewhere that will be able to flatten a few of your cities.
MAD only works as a deterrent if there is a very small chance of getting away with a first strike. If you can shoot down incoming ballistic missiles, then you might be tempted to launch first and just shoot down the retaliatory strike. If you can fire missiles that are fast enough that the enemy doesn't have time to order a retaliatory strike, then you might be willing to live with a couple of submarines shooting at you, especially if you have a good idea of where they are and can have aircraft ready to launch interceptors flying above them.
Aircraft carriers completely suck for launching nuclear missiles (and the USA has not confirmed that they even bother putting them on board), because they're a big obvious target. If, for example, the USA put a nuclear-armed aircraft carrier near Russia, then you can bet that the Russian second-strike capability would be on alert and ready to launch unless they got regular confirmation that the US carrier hadn't launched. Aircraft carriers also suck for second-strike, because they're too big to easily dodge a nuclear strike and so will be on the list for first-strike targets.
They're good for a conventional (non-nuclear) first strike, because they give you a staging post that you can put right next to someone else's country, but that's less of a problem for global stability. A non-nuclear first strike commits you to an expensive war. Aircraft carriers aren't also specifically first-strike weapons. They have support roles and are even occasionally used for disaster relief (a big portable nuclear reactor turns out to be quite handy sometimes).
Five hundred hostiles is well within the targeting capacity of a single Apache, let alone a warship. The reason that the US Navy has been investing heavily in laser weapons recently is that there's a gap between being able to target them all and being able to shoot them all.
There is, but it's to coarse grained. For example, I'd be happy for an IM application to use mobile data, because it's likely to be very low traffic. I'd like my web browser to use mobile data sometimes: when I'm away from WiFi and need to look something up urgently. I want everything else to be restricted to WiFi. I want these settings to vary slightly between when I'm at home and when I'm roaming, as data charges can vary by a couple of orders of magnitude.
Android actually does provide an interface for doing this, but everything defaults to being able to be used all of the time. Within some applications (e.g. web browser) I want something like the old 'show images' button from the MODEM days, so that images, videos, and anything else large will not be downloaded unless I opt in on a specific page.
For comparison, the USA is #1, Germany, France and the UK are 6-8. Now, if you're the worlds second-largest economy, would you prefer to have the worlds eighth-largest economy as a trading partner, or the world first, third, fourth, fifth and sixth? Now, if rational choices aren't enough, keep in mind that (in the scenario under discussion), the current President of #8 is someone who fondly recalls the days when you were a junior partner in an alliance and has demonstrated a willingness to use nuclear weapons.
The BBC news article about the hack had a quote from one of the celebrities saying that the pictures had already been deleted before they were stolen. That is the problem with these services: they don't securely (or, at all) delete things. Google's deletion mechanism, for example, relies on simply not actively copying the files to newer disks so that when the old disks eventually die the files are gone. I wouldn't be surprised if Apple's works in a similar way. Even if you decide you don't trust Apple/Google/Facebook today, you've got a long wait before all of the files that you've uploaded to them are gone.
Step one is to have the big high-profile stories in the press about the problems. Step two is to have the big high-profile stories in the press about the alternatives. The important thing now is for anyone who is contacted by the press as an expert to ask about the iCloud hack to make it very clear that this isn't an Apple-specific problem, it's a problem inherent in the entire design of centralised services and to list alternatives.
It was on the BBC news this morning, which probably counts as more reliable than 4chan. Most interesting was the claim by one of the women involved that the photos had been deleted. If this is true, then it would be a great example of the fact that just because something is 'deleted' in the cloud doesn't mean that malicious people can't get at it in the future...
If I'm going to report errors in a map, I'd rather do so with a map that releases its data under a license that allows reuse. Since such a map already exists and doesn't have the errors in Google Maps, I don't see much incentive. Google can pull the data from there if they want. This is actually one case where Microsoft has been a bit nicer: they allowed OSM to trace their satellite images to improve maps. Google Maps, in contrast, is very protective over their data.
I don't know about rural Japan, but I found the OSM data far better when I visited Tokyo a few months ago. Google didn't know building names and placed a load of things that we were trying to visit a few blocks away from where they actually were. This was very frustrating as the web site that we were using to find vegetarian restaurants used Google maps - we spent half an hour one lunchtime walking in the wrong direction, because we'd come to a junction and, according to the Google map needed to turn left and would then see our destination on the right. It turned out, when we eventually found it, that we should have turned right.
I use OSMAnd on my phone[1], but my girlfriend recently bought a Windows Phone and I've been very impressed with Nokia's mapping app (I actually like a lot of what Microsoft's done with Windows Phone 8, but it's a strange mix of very polished and well-designed UI parts and completely unfinished parts with missing features). It's good to see more competition with Google maps, which is becoming increasingly entrenched in spite of the fact that the UI is pretty poor in many regards and the mapping data is terrible. For example, here they're missing (or have in the wrong places) most of the cycle paths, which ends up with people regularly getting lost if they rely on Google, in spite of the fact that all of this data is in OpenStreetMap.
[1] For me, it's the killer app for Android. Offline maps, offline routing, and open source backed by high-quality mapping data from OpenStreetMap. I use the version from the F-Droid store, which doesn't have the limitations of the free version from Google Play and it's one of the few open source apps that I've donated money to.
Remember when the SoundBlaster Live! came out and Creative Labs were telling you that it had as much processing power as a Pentium 166MHz MMX, dedicated entirely to sound processing? Well, it turns out that now you can have far more CPU power than that dedicated entirely to sound processing without custom hardware...
Just curious, what's so wrong with branch with delay slot and isn't that more native way to look at branch ?
They're a pain for people on both sides of the ISA.
The compiler has to find an instruction that can run after the branch. This is normally trivial for calls, but for conditional branches within a function it's often difficult to find an instruction that you can put there. It has to be one that is either from before the jump (or in both basic blocks after the jump), but that the branch doesn't depend on (because it's executed after the branch instruction). This means that you quite often end up padding the delay slots with nops, which bloats your instruction cache usage. On a superscalar implementation this is the only cost, but on a simple in-order pipeline it's also a completely wasted cycle.
On the other side, it's a pain to implement. It made sense for a three-stage pipeline in the original MIPS, because you always knew the next instruction to fetch. A modern simple pipeline is 5-7 stages though, so your branch is still in register fetch (if there) by the time the delay slot is needed. It doesn't buy anything and it means that, if you're doing any kind of speculative execution (even simple branch prediction, which you really need to do to get moderately good performance) then you have an extra dependency to track - you can't just use the branch as the marker and flush everything after it, you need to do some reordering. In a superscalar implementation, you need to do even more complex things in register renaming to make it work.
No idea. I don't know if the instructions for computing PC-relative addresses in an ISA without an architectural PC are patentable. They also exist in RISC V (not sure which came first), so if they do then it's going to be a problem for Kriste et al. Nothing else in there is especially novel: like ARMv8, it's a nicely designed compilation target, but it doesn't do anything that's especially exciting.
I didn't look at the floating point stuff in much detail, so there may be something there, although the biggest changes in recent versions of the MIPS specs have been that they're more closely aligned with the IEEE floating point standards, so it's hard to imagine anything there.
The biggest difference between MIPS64r6 and ARMv8 is that the MIPS spec explicitly reserves some of the opcode space for vendor-specific extensions (we use this space, although our core predates the current spec - it's largely codifying existing opcode use). This allows, for example, Cavium to add custom instructions that are useful for network switches but not very useful for other things. ARMv8, in contrast, expects that any non-standard extensions are in the form of accelerator cores with a completely different ISA. This means that any code compiled for one ARMv8 core should run on any ARMv8 implementation, which is a big advantage. With MIPS, anything compiled for the core ISA should run everywhere, but people using custom variants (e.g. Cisco and Juniper, who use the Cavium parts in some of their products) will ship code that won't run on another vendors' chips.
Historically, this has been a problem for the MIPS ecosystem because each MIPS vendor has forked GCC and GNU binutils, hacked it up to support their extensions, but done so in a way that makes it impossible to merge the code upstream (because they've broken every other MIPS chip in the process) and left their customers with an ageing toolchain to deal with. I've been working with the Imagination guys to try to make sure that the code in LLVM is arranged in such a way that it's relatively easy to add vendor-specific extensions without breaking everything else.
Imagination doesn't currently have any 64-bit cores to license, but I expect that they will quite soon...
Slashdot Mirror
Would you prefer that mega-corps not make their videos available for streaming at all?
No, I'd rather that copyright on music and video be contingent on distribution in a form that does not lock customers into a particular platform.
Talking about open-source businesses is missing the point entirely. Most businesses that are successful as a result of open source (or Free Software, for the RMS-style folks) or that contribute significantly to open source are not 'open-source businesses' any more than companies that use Windows and Office are 'closed-source businesses. The difference is that one category of businesses realises that writing software is expensive and copying software is trivial, so spends its investment on the software parts of its infrastructure paying people to write software (typically customising and improving existing projects), whereas the other pays someone for copies of software and hopes that that will give them an incentive to produce software that's more like they want.
You're comparing apples and oranges as far as the technical details. I'm saying Win 3.x let me continue when it saw problems, and NT could also do that.
Not really. The kind of situations where Windows 3.x let you try to continue, Windows NT just handles transparently. In Windows 3.x, with cooperative multitasking, a single application can refuse to relinquish the CPU. If this happens, you have three choices (outlined by the dialog box):
In a system with protected memory and preemptive multitasking, an application that refuses to relinquish the CPU will just have its priority downgraded and the only thing that you'll notice is the CPU getting warm. Eventually, you may choose to kill the program, but it never affects system stability.
I'd like to have the *option* to continue to save my work even if there was a chance of data corruption. For example, take the common NT blue screen IRQL_NOT_LESS_OR_EQUAL. That fact that my buggy network driver tried to access paged memory in the wrong sequence is miles away from catastrophic. And it certainly doesn't take priority over something I've been working on for hours. IRQ 0 is me, motherfuckers!
It means that there's a high probability that something has damaged some kernel data structures. If you continue, there's a good chance that this corruption will spread to the buffer cache and you'll end up writing invalid data to disk. If you kill the system, the corruption is limited to the RAM.
Agreed on Chen's blog, but the summary is horrible. This message hasn't been part of Windows since Windows 95 (which introduced preemptive multitasking to the Windows world, so a single application could no longer freeze the system trivially), so the odds are that if you used Windows in the last two decades you've never seen this notice...
The idea is to have a timer that would automatically disable the equipment unless it received an enable signal, either from a satellite or removable medium. It's possible to make such a system that is, at the very least, very difficult to tamper with. Many of the systems on tanks and so on are computer controlled and if the computers stop working then it's a lot less valuable. The goal of such systems is similar to that of crypto: it's not to prevent the enemy from ever using the tanks that they've stolen, it's to prevent them using them quickly. If you have a few weeks to bomb the stolen equipment before it can be used, and the enemy has to invest a lot of high-tech resources into cracking the systems, then that's probably good enough.
MIRVs are a better second-strike weapon. In a first-strike context, your missile bases are all working fine and you can just launch everything that you've got at the enemy. MIRV in a first strike requires fewer missiles to get through, but that just means you need to build fewer missiles, which doesn't reduce the cost by a huge amount compared to the cost of maintaining a first-strike capability at all. In a second-strike situation, however, you're much more likely to have limited launch resources (a few submarines if you're late, the missile silos that you can contact in time if you're not). Being able to have a devastating second-strike capability with just one submarine surviving is a strong deterrent.
Also I'm not sure about your phone but with mine I can disable data over cellular in the swipe down settings. It's literally a swipe and a single click, so if you really want to restrict everything and then only use it as required that would be the easy way to do it system wide and is about as complicated as turning on screen rotation.
You didn't read my post, did you? You can turn it off globally, but it's really hard to then turn it on for a subset of applications. You have to individually disable each one's access (and they all default to 'on', so you have to remember to do it again after installing each new app). If I turn on mobile data somewhere expensive to look something up urgently (or, for example, to get an updated boarding pass for a flight) then suddenly a dozen applications will say 'whee, Internet! Let's download a load of updates!'
Honestly, even if I'm doing it a on a million items, O(N^2) vs O(N) isn't a big deal
If you're doing a million items, then the difference between O(N) and O(N^2) is a factor of a million. If each operation takes one cycle to complete, then on a 2GHz CPU you're going to take 500 microseconds. More plausibly, if each iteration takes 10 cycles, then it will take 5ms. The O(n^2) version, assuming the same ten-cycles per operation (which is pretty small - even a bubblesort on integers in an array will be hard pressed to be that quick), will take one hour and 23 minutes.
If you really think that the difference between 5ms and almost an hour and a half is irrelevant, then I really hope that I never use any code that you write.
De Morgen's laws are widely known and trivial to automate. Any vaguely modern compiler knows about them and will do the transform for you. Not only will it do it for you, it will do it for you everywhere and base on knowledge of the target architecture.
That said, it won't rearrange things that might have side effects. In C, for example, you have short-circuit evaluation, so if you have something cheap on the left of an or and something expensive on the right, then it will be faster unless the compiler can determine that the expensive thing has no side effects. If you have a negation of a long string of ands, the arguments to each of which is expensive to compute, then turning it into a string of negated ors is faster, because the compiler can stop computing once it's found one that's failed.
Exactly what I came here to post. We had the demonstration of what happens when you immerse something in liquid nitrogen vs what happens when you pour it over the top. Even if you didn't get to play with liquid nitrogen in school, there are lots of videos of this.
Or, you put it another way, the most expensive wishful thinking on the planet.
When people say 'first strike' in this context, they mean 'nuclear first strike', as in 'launch enough nuclear missiles to take out your enemy's ability to fire back'. The reason for nuclear submarines is to largely eliminate this possibility: even if you completely destroy every military installation in the USA, there is going to be at least one submarine hidden somewhere that will be able to flatten a few of your cities.
MAD only works as a deterrent if there is a very small chance of getting away with a first strike. If you can shoot down incoming ballistic missiles, then you might be tempted to launch first and just shoot down the retaliatory strike. If you can fire missiles that are fast enough that the enemy doesn't have time to order a retaliatory strike, then you might be willing to live with a couple of submarines shooting at you, especially if you have a good idea of where they are and can have aircraft ready to launch interceptors flying above them.
Aircraft carriers completely suck for launching nuclear missiles (and the USA has not confirmed that they even bother putting them on board), because they're a big obvious target. If, for example, the USA put a nuclear-armed aircraft carrier near Russia, then you can bet that the Russian second-strike capability would be on alert and ready to launch unless they got regular confirmation that the US carrier hadn't launched. Aircraft carriers also suck for second-strike, because they're too big to easily dodge a nuclear strike and so will be on the list for first-strike targets.
They're good for a conventional (non-nuclear) first strike, because they give you a staging post that you can put right next to someone else's country, but that's less of a problem for global stability. A non-nuclear first strike commits you to an expensive war. Aircraft carriers aren't also specifically first-strike weapons. They have support roles and are even occasionally used for disaster relief (a big portable nuclear reactor turns out to be quite handy sometimes).
Five hundred hostiles is well within the targeting capacity of a single Apache, let alone a warship. The reason that the US Navy has been investing heavily in laser weapons recently is that there's a gap between being able to target them all and being able to shoot them all.
There is, but it's to coarse grained. For example, I'd be happy for an IM application to use mobile data, because it's likely to be very low traffic. I'd like my web browser to use mobile data sometimes: when I'm away from WiFi and need to look something up urgently. I want everything else to be restricted to WiFi. I want these settings to vary slightly between when I'm at home and when I'm roaming, as data charges can vary by a couple of orders of magnitude.
Android actually does provide an interface for doing this, but everything defaults to being able to be used all of the time. Within some applications (e.g. web browser) I want something like the old 'show images' button from the MODEM days, so that images, videos, and anything else large will not be downloaded unless I opt in on a specific page.
For comparison, the USA is #1, Germany, France and the UK are 6-8. Now, if you're the worlds second-largest economy, would you prefer to have the worlds eighth-largest economy as a trading partner, or the world first, third, fourth, fifth and sixth? Now, if rational choices aren't enough, keep in mind that (in the scenario under discussion), the current President of #8 is someone who fondly recalls the days when you were a junior partner in an alliance and has demonstrated a willingness to use nuclear weapons.
The BBC news article about the hack had a quote from one of the celebrities saying that the pictures had already been deleted before they were stolen. That is the problem with these services: they don't securely (or, at all) delete things. Google's deletion mechanism, for example, relies on simply not actively copying the files to newer disks so that when the old disks eventually die the files are gone. I wouldn't be surprised if Apple's works in a similar way. Even if you decide you don't trust Apple/Google/Facebook today, you've got a long wait before all of the files that you've uploaded to them are gone.
Step one is to have the big high-profile stories in the press about the problems. Step two is to have the big high-profile stories in the press about the alternatives. The important thing now is for anyone who is contacted by the press as an expert to ask about the iCloud hack to make it very clear that this isn't an Apple-specific problem, it's a problem inherent in the entire design of centralised services and to list alternatives.
It was on the BBC news this morning, which probably counts as more reliable than 4chan. Most interesting was the claim by one of the women involved that the photos had been deleted. If this is true, then it would be a great example of the fact that just because something is 'deleted' in the cloud doesn't mean that malicious people can't get at it in the future...
If I'm going to report errors in a map, I'd rather do so with a map that releases its data under a license that allows reuse. Since such a map already exists and doesn't have the errors in Google Maps, I don't see much incentive. Google can pull the data from there if they want. This is actually one case where Microsoft has been a bit nicer: they allowed OSM to trace their satellite images to improve maps. Google Maps, in contrast, is very protective over their data.
Not sure what post you think you're replying to, but the quoted line didn't appear in my post...
I don't know about rural Japan, but I found the OSM data far better when I visited Tokyo a few months ago. Google didn't know building names and placed a load of things that we were trying to visit a few blocks away from where they actually were. This was very frustrating as the web site that we were using to find vegetarian restaurants used Google maps - we spent half an hour one lunchtime walking in the wrong direction, because we'd come to a junction and, according to the Google map needed to turn left and would then see our destination on the right. It turned out, when we eventually found it, that we should have turned right.
I use OSMAnd on my phone[1], but my girlfriend recently bought a Windows Phone and I've been very impressed with Nokia's mapping app (I actually like a lot of what Microsoft's done with Windows Phone 8, but it's a strange mix of very polished and well-designed UI parts and completely unfinished parts with missing features). It's good to see more competition with Google maps, which is becoming increasingly entrenched in spite of the fact that the UI is pretty poor in many regards and the mapping data is terrible. For example, here they're missing (or have in the wrong places) most of the cycle paths, which ends up with people regularly getting lost if they rely on Google, in spite of the fact that all of this data is in OpenStreetMap.
[1] For me, it's the killer app for Android. Offline maps, offline routing, and open source backed by high-quality mapping data from OpenStreetMap. I use the version from the F-Droid store, which doesn't have the limitations of the free version from Google Play and it's one of the few open source apps that I've donated money to.
Remember when the SoundBlaster Live! came out and Creative Labs were telling you that it had as much processing power as a Pentium 166MHz MMX, dedicated entirely to sound processing? Well, it turns out that now you can have far more CPU power than that dedicated entirely to sound processing without custom hardware...
Just curious, what's so wrong with branch with delay slot and isn't that more native way to look at branch ?
They're a pain for people on both sides of the ISA.
The compiler has to find an instruction that can run after the branch. This is normally trivial for calls, but for conditional branches within a function it's often difficult to find an instruction that you can put there. It has to be one that is either from before the jump (or in both basic blocks after the jump), but that the branch doesn't depend on (because it's executed after the branch instruction). This means that you quite often end up padding the delay slots with nops, which bloats your instruction cache usage. On a superscalar implementation this is the only cost, but on a simple in-order pipeline it's also a completely wasted cycle.
On the other side, it's a pain to implement. It made sense for a three-stage pipeline in the original MIPS, because you always knew the next instruction to fetch. A modern simple pipeline is 5-7 stages though, so your branch is still in register fetch (if there) by the time the delay slot is needed. It doesn't buy anything and it means that, if you're doing any kind of speculative execution (even simple branch prediction, which you really need to do to get moderately good performance) then you have an extra dependency to track - you can't just use the branch as the marker and flush everything after it, you need to do some reordering. In a superscalar implementation, you need to do even more complex things in register renaming to make it work.
I didn't look at the floating point stuff in much detail, so there may be something there, although the biggest changes in recent versions of the MIPS specs have been that they're more closely aligned with the IEEE floating point standards, so it's hard to imagine anything there.
The biggest difference between MIPS64r6 and ARMv8 is that the MIPS spec explicitly reserves some of the opcode space for vendor-specific extensions (we use this space, although our core predates the current spec - it's largely codifying existing opcode use). This allows, for example, Cavium to add custom instructions that are useful for network switches but not very useful for other things. ARMv8, in contrast, expects that any non-standard extensions are in the form of accelerator cores with a completely different ISA. This means that any code compiled for one ARMv8 core should run on any ARMv8 implementation, which is a big advantage. With MIPS, anything compiled for the core ISA should run everywhere, but people using custom variants (e.g. Cisco and Juniper, who use the Cavium parts in some of their products) will ship code that won't run on another vendors' chips.
Historically, this has been a problem for the MIPS ecosystem because each MIPS vendor has forked GCC and GNU binutils, hacked it up to support their extensions, but done so in a way that makes it impossible to merge the code upstream (because they've broken every other MIPS chip in the process) and left their customers with an ageing toolchain to deal with. I've been working with the Imagination guys to try to make sure that the code in LLVM is arranged in such a way that it's relatively easy to add vendor-specific extensions without breaking everything else.
Imagination doesn't currently have any 64-bit cores to license, but I expect that they will quite soon...