Hmm, though the other factor is cache page size - you can fit twice as many 32 bit pointers in the same cache page - so for traversing large data trees you could potentially get a significant performance boost by using smaller pointers. Provided of course that you never wanted to deal with more than 4GB of total data.
Bingo! Cache miss latency is horrible on modern CPUs. x32 reduces your data size down a bit, particularly for things like STL structures that have pointers and sizes all over the place. It's a nice idea in theory, but the trouble is you have to build your own userland, and forget third-party libraries.
Also PC-relative addressing works in x32 mode, and that's a huge gain over i386 for position-independent code (think shared libraries and ASLR). It's supposed to help reduce the size of the working set so you don't thrash the cache as much as you would with 64-bit pointer, size_t etc. Cache miss latency is horrible on modern systems. The trouble is, there are very few libraries built for it, so you pretty much have to build your own userland before you can do anything.
I thought you were just saying that because your wife would kill you if you didn't, but then I found a recent photo, and the last ten years have not been kind to Ms Meng.
The problem is, the users who really care about good native apps -- users who know HIG violations when they see them, who care about performance, who care about Mac apps being right -- were mostly already on the Mac. A lot of newer Mac users either don't know or don't care about what makes for a good Mac app.
No, some time around 10.3 Apple forgot what made for a good Mac app. They went for skeuomorphism where it wasn't useful, per-window brushed metal appearance, non-standard UI widgets all over the place, etc. They deprecated actual useful UI features like drawers, and violated their own guidelines left, right and centre. Then they decided they needed to remove "Save As..." at some point, only to bring it back later with different behaviour (making it save at the old location before also saving at the new location).
Sure I dislike apps that don't feel right. I never liked how Qt applications always felt not-quite-native on OSX, and the same with Firefox where menus don't quite feel right because they aren't native. I dislike the situation Linux inherited from UNIX where every UI toolkit has a different look-and-feel so it varies from app to app. Microsoft is a big mess, too. It's probably best demonstrated with settings, where you have classic control panels that open in their own windows and behave like regular Win32 apps, sort of web-like control panels that open in the "All Control Panels" window and can be navigated with back/forward buttons, MMC snapins inherited from WinNT and things that mimic that approach, and then the flat-look touch-optimised Metro-style settings. Nothing's unified at all.
The OS vendors/distributors themselves aren't providing coherent, unified UI language, so how do you expect third-party app developers do do so? Everyone's bundling frameworks with apps now since Apple made it the trendy thing to do (just include the frameworks inside the app bundle), Microsoft made it easy (use WinSxS so you can have per-app versions of DLLs), and the Linux distros jumped on the bandwagon (Flatpak? Snap). Electron may be a particularly heavy-weight framework, but it's definitely not the only one that gets bundled with applications. But this has lead to a vicious cycle where apps bundle frameworks to avoid incompatibilities, so the frameworks are emboldened to make incompatible changes on minor releases, necessitating the bundling. It isn't an option to use a system-wide installation of a lot of frameworks, because they don't maintain compatibility.
Even if MS manages to make a single system-wide instance of Chromium for running Electron apps, it's not going to solve bloat issues with duplication of the JavaScript frameworks used by the apps themselves. You can't use a single instance of them because of incompatibility between versions, and the fact that JavaScript lets you inject members into any object. You'll still have the bloat of loading, JIT-compiling and caching the JavaScript frameworks for every app.
The trouble is, we've reinvented Java but it's worse in almost every way. We're running JIT-compiled code from untrusted sources on a sprawling runtime with vulnerabilities discovered regularly. But now every app also pulls in a huge pile of dependencies, bloating it further, and greatly increasing the number of third parties you need to trust when you run the code. I don't see how this is any better.
I've owned a whole lot of Macs, and the quality has dropped. The 2001 iBook was very tough and well-built, but the first-generation MacBooks were terrible. Top case broke, optical drive got misaligned with the slot on the case, GMA950 graphics would cause regular kernel panics. The unibody MacBook Pros bend far too easily, and because they're made of aluminium, they stay bent. The older designs were a lot more resilient. I have a 2010 MacBook Pro, and the interrupt controller now only works intermittently, so it will kernel panic or lock up regularly. Funny thing is, I have a "Snakebite" dual G4 that still works fine, and even a couple of working 68k Macs. Apple quality isn't what it used to be.
Well the solution would be to provide some amount of guard band, like a "Please remove your finger and read this" prompt if you have a finger on the sensor before the message appears.
Yeah, the Direct3D drivers for GMA950 were absolutely terrible. Somehow the Windows OpenGL drivers actually support more features (e.g. Windows OpenGL driver supports non-power-of-two textures while Direct3D driver doesn't). They were pretty bad on Mac OS X as well. I had a white MacBook with one, and using an external monitor to extend the desktop would cause regular kernel panics. The way it stole RAM bandwidth from the CPU made performance suck, and that got worse with an external monitor connected, too.
I'll believe Intel can build a discrete GPU worth buying when I see it. Every attempt so far has been flawed (Real3D i740 starved for texture bandwidth), weak (Silicon Image GMA950 with terrible performance and even worse drivers on Windows), or vapourware (Larrabee). There's no indication that it'll be different this time.
Only the CPU core itself is open source. You'll need a memory controller, cache controller, PCI-e bridge, and a lot of other stuff to make a working CPU. The RISC-V people will be happy to sell you this stuff under a proprietary license.
Lie I personally find most annoying: you can't use civilian reactor bred plutonium to make nuclear weapons, it's in there too long, too much of the isotopes that are hot or fission too easily.
Quite a few commercial power reactor designs can be used to breed plutonium. MAGNOX and RBMK were openly designed to allow simultaneous power generation and plutonium production. India clearly found a way to breed plutonium in a CANDU reactor. You could probably find a way to use an AGR to produce useful amounts of plutonium, too.
MAME developer here: we cracked CPS2 encryption with a known plaintext attack. It wasn't until much later that we burned the top off the encrypted CPU and took a photo through a microscope.
It's easy to fix the "windows 10 update ills" by using Windows 10 Enterprise on a domain with a local WSUS server. If your IT department can't set that up, they probably aren't worth what they're paid.
Honestly, it's considered a harder game to solve in terms of computational power vs. chess. Because xiang qi is considered a harder game, you could make an argument that China spending so many resources to win chess championships is kind of a case of them slumming and trying to win an "easier" game.
Are you sure you aren't thinking about Japanese Shogi? Shogi has a much higher branching factor than chess due to optional promotions and the ability to re-use captured pieces ("drops"). As far as I know, Chinese Xiangqi isn't as computationally expensive due to the area restrictions on the general, guards and elephants, and the relative lack of long-ranged pieces. The cannon's requirement to jump when capturing and the ability to block knights and elephants help make it interesting though.
No, the rook is the chariot. King -> king, adviser -&;gt queen, elephant -> bishop, horse -> knight, chariot -> rook, foot-soldier -> pawn. Wikipedia even agrees with me.
The Linux ecosystem isn't in great shape either. Just look at systemd - some good ideas, but horrible implementation. Far too much attack surface on PID 1, no privilege separation, effectively proprietary interfaces. Gnome removes functionality with every release, and KDE just messes up the UI for no reason. When I boot up a Fedora 20 KDE system, the KDE UI looks and feels so much better.
Have you used any software lately? There's no sense of quality or pride in work.
macOS had a bug where you could log in as root without a password.
Microsoft keeps breaking MSVC in horrible ways and their response is to tell you to wait a couple of years for the next major update. Windows gets bugs like preventing you from changing keyboard/layout input method settings, and they won't fix it until the next six-monthly update, by which time they'll have broken more stuff. The bug where files were deleted if you had custom library paths had been reported by "insiders" yet they still rolled out the update without fixing it. Oh and the way they've sacked their QA department and try to sell being an unpaid beta tester as being an "insider" is disgusting. You're just doing unpaid testing at the risk of your own data/productivity, you're not an "insider" in any meaningful sense.
Meanwhile, Red Hat has forgotten what "stability" means. RHEL7 has broken not just binary compatibility but also source compatibility for kernel extensions on a point release (the APIs for IPv6 stuff changed in an incompatible way), and they subsequently completely changed how Infiniband and RDMA work on a point release, breaking compatibility with everything that uses it. Their graphical installer now needs 2GB of RAM to run properly, even though you can actually run an installed system in 512MB. The default installation contains a whole pile of WiFi firmware packages and daemons that are only useful if your network configuration changes dynamically (like a notebook used on public WiFi) which you'd never need on a server. They've completely lost touch with their customers.
Games are all developed with a "ship buggy, patch later" mentality. There have been games released in a state where it's impossible to finish the first level. You know what was good about cartridge-based consoles with no connectivity? They needed some QA because it was expensive to pay for a run of mask ROMs that you had to dump in landfill.
Software is almost uniformly shit these days. Apparently no-one thinks it's worth making good software any more.
Arcade games are upfront and predictable about what you get for a credit (a certain number of lives, health, time extend in a driving game, whatever). Lootboxes contain random, often duplicated items. They're gambling.
Slashdot Mirror
Bingo! Cache miss latency is horrible on modern CPUs. x32 reduces your data size down a bit, particularly for things like STL structures that have pointers and sizes all over the place. It's a nice idea in theory, but the trouble is you have to build your own userland, and forget third-party libraries.
Also PC-relative addressing works in x32 mode, and that's a huge gain over i386 for position-independent code (think shared libraries and ASLR). It's supposed to help reduce the size of the working set so you don't thrash the cache as much as you would with 64-bit pointer, size_t etc. Cache miss latency is horrible on modern systems. The trouble is, there are very few libraries built for it, so you pretty much have to build your own userland before you can do anything.
I thought you were just saying that because your wife would kill you if you didn't, but then I found a recent photo, and the last ten years have not been kind to Ms Meng.
Nope, I always notice because I wonder why the app is such a resource hog, then I find out it's one of these Chromium things.
That's my point - the OS/DE people provide guidelines, then violate them all over the place themselves. They're leading by setting a very bad example.
No, some time around 10.3 Apple forgot what made for a good Mac app. They went for skeuomorphism where it wasn't useful, per-window brushed metal appearance, non-standard UI widgets all over the place, etc. They deprecated actual useful UI features like drawers, and violated their own guidelines left, right and centre. Then they decided they needed to remove "Save As..." at some point, only to bring it back later with different behaviour (making it save at the old location before also saving at the new location).
Sure I dislike apps that don't feel right. I never liked how Qt applications always felt not-quite-native on OSX, and the same with Firefox where menus don't quite feel right because they aren't native. I dislike the situation Linux inherited from UNIX where every UI toolkit has a different look-and-feel so it varies from app to app. Microsoft is a big mess, too. It's probably best demonstrated with settings, where you have classic control panels that open in their own windows and behave like regular Win32 apps, sort of web-like control panels that open in the "All Control Panels" window and can be navigated with back/forward buttons, MMC snapins inherited from WinNT and things that mimic that approach, and then the flat-look touch-optimised Metro-style settings. Nothing's unified at all.
The OS vendors/distributors themselves aren't providing coherent, unified UI language, so how do you expect third-party app developers do do so? Everyone's bundling frameworks with apps now since Apple made it the trendy thing to do (just include the frameworks inside the app bundle), Microsoft made it easy (use WinSxS so you can have per-app versions of DLLs), and the Linux distros jumped on the bandwagon (Flatpak? Snap). Electron may be a particularly heavy-weight framework, but it's definitely not the only one that gets bundled with applications. But this has lead to a vicious cycle where apps bundle frameworks to avoid incompatibilities, so the frameworks are emboldened to make incompatible changes on minor releases, necessitating the bundling. It isn't an option to use a system-wide installation of a lot of frameworks, because they don't maintain compatibility.
Even if MS manages to make a single system-wide instance of Chromium for running Electron apps, it's not going to solve bloat issues with duplication of the JavaScript frameworks used by the apps themselves. You can't use a single instance of them because of incompatibility between versions, and the fact that JavaScript lets you inject members into any object. You'll still have the bloat of loading, JIT-compiling and caching the JavaScript frameworks for every app.
The trouble is, we've reinvented Java but it's worse in almost every way. We're running JIT-compiled code from untrusted sources on a sprawling runtime with vulnerabilities discovered regularly. But now every app also pulls in a huge pile of dependencies, bloating it further, and greatly increasing the number of third parties you need to trust when you run the code. I don't see how this is any better.
I've owned a whole lot of Macs, and the quality has dropped. The 2001 iBook was very tough and well-built, but the first-generation MacBooks were terrible. Top case broke, optical drive got misaligned with the slot on the case, GMA950 graphics would cause regular kernel panics. The unibody MacBook Pros bend far too easily, and because they're made of aluminium, they stay bent. The older designs were a lot more resilient. I have a 2010 MacBook Pro, and the interrupt controller now only works intermittently, so it will kernel panic or lock up regularly. Funny thing is, I have a "Snakebite" dual G4 that still works fine, and even a couple of working 68k Macs. Apple quality isn't what it used to be.
You must have never worked with safety-critical equipment. This kind of interface design is common in robotics and heavy machinery.
That wasn't about developers employed by Microsoft - it was about trying to ensure that Windows is the preferred desktop OS for developers in general.
Well the solution would be to provide some amount of guard band, like a "Please remove your finger and read this" prompt if you have a finger on the sensor before the message appears.
Yeah, the Direct3D drivers for GMA950 were absolutely terrible. Somehow the Windows OpenGL drivers actually support more features (e.g. Windows OpenGL driver supports non-power-of-two textures while Direct3D driver doesn't). They were pretty bad on Mac OS X as well. I had a white MacBook with one, and using an external monitor to extend the desktop would cause regular kernel panics. The way it stole RAM bandwidth from the CPU made performance suck, and that got worse with an external monitor connected, too.
I'll believe Intel can build a discrete GPU worth buying when I see it. Every attempt so far has been flawed (Real3D i740 starved for texture bandwidth), weak (Silicon Image GMA950 with terrible performance and even worse drivers on Windows), or vapourware (Larrabee). There's no indication that it'll be different this time.
Only the CPU core itself is open source. You'll need a memory controller, cache controller, PCI-e bridge, and a lot of other stuff to make a working CPU. The RISC-V people will be happy to sell you this stuff under a proprietary license.
Quite a few commercial power reactor designs can be used to breed plutonium. MAGNOX and RBMK were openly designed to allow simultaneous power generation and plutonium production. India clearly found a way to breed plutonium in a CANDU reactor. You could probably find a way to use an AGR to produce useful amounts of plutonium, too.
MAME developer here: we cracked CPS2 encryption with a known plaintext attack. It wasn't until much later that we burned the top off the encrypted CPU and took a photo through a microscope.
I always think of the Mitsubishi Colt, the 1983-1987 one specifically.
Have you scrolled up a bit and read the comments suggesting bombing the UK over this?
It's easy to fix the "windows 10 update ills" by using Windows 10 Enterprise on a domain with a local WSUS server. If your IT department can't set that up, they probably aren't worth what they're paid.
Are you sure you aren't thinking about Japanese Shogi? Shogi has a much higher branching factor than chess due to optional promotions and the ability to re-use captured pieces ("drops"). As far as I know, Chinese Xiangqi isn't as computationally expensive due to the area restrictions on the general, guards and elephants, and the relative lack of long-ranged pieces. The cannon's requirement to jump when capturing and the ability to block knights and elephants help make it interesting though.
No, the rook is the chariot. King -> king, adviser -&;gt queen, elephant -> bishop, horse -> knight, chariot -> rook, foot-soldier -> pawn. Wikipedia even agrees with me.
Boats? Aren't they equivalent to the elephant in the earlier Indian games chess is descended from?
The Linux ecosystem isn't in great shape either. Just look at systemd - some good ideas, but horrible implementation. Far too much attack surface on PID 1, no privilege separation, effectively proprietary interfaces. Gnome removes functionality with every release, and KDE just messes up the UI for no reason. When I boot up a Fedora 20 KDE system, the KDE UI looks and feels so much better.
In the US maybe. It's not legal in most of Europe.
Have you used any software lately? There's no sense of quality or pride in work.
macOS had a bug where you could log in as root without a password.
Microsoft keeps breaking MSVC in horrible ways and their response is to tell you to wait a couple of years for the next major update. Windows gets bugs like preventing you from changing keyboard/layout input method settings, and they won't fix it until the next six-monthly update, by which time they'll have broken more stuff. The bug where files were deleted if you had custom library paths had been reported by "insiders" yet they still rolled out the update without fixing it. Oh and the way they've sacked their QA department and try to sell being an unpaid beta tester as being an "insider" is disgusting. You're just doing unpaid testing at the risk of your own data/productivity, you're not an "insider" in any meaningful sense.
Meanwhile, Red Hat has forgotten what "stability" means. RHEL7 has broken not just binary compatibility but also source compatibility for kernel extensions on a point release (the APIs for IPv6 stuff changed in an incompatible way), and they subsequently completely changed how Infiniband and RDMA work on a point release, breaking compatibility with everything that uses it. Their graphical installer now needs 2GB of RAM to run properly, even though you can actually run an installed system in 512MB. The default installation contains a whole pile of WiFi firmware packages and daemons that are only useful if your network configuration changes dynamically (like a notebook used on public WiFi) which you'd never need on a server. They've completely lost touch with their customers.
Games are all developed with a "ship buggy, patch later" mentality. There have been games released in a state where it's impossible to finish the first level. You know what was good about cartridge-based consoles with no connectivity? They needed some QA because it was expensive to pay for a run of mask ROMs that you had to dump in landfill.
Software is almost uniformly shit these days. Apparently no-one thinks it's worth making good software any more.
Arcade games are upfront and predictable about what you get for a credit (a certain number of lives, health, time extend in a driving game, whatever). Lootboxes contain random, often duplicated items. They're gambling.