6 ms·

Meh. I see their point. Subpixel AA was about using software to simulate better hardware. But, at the end of the day you really do want the better hardware so t

Meh. I see their point. Subpixel AA was about using software to simulate better hardware. But, at the end of the day you really do want the better hardware so that you can use your software budget to simulate an even better generation of hardware. Repeat ad-infinitum, modulo scaling limits. Pretty sure this is what Alan Kay means when he says that people who are serious about software will make their own hardware.

adrianN 8y ago
Maybe we could wait with the step where we don't simulate better hardware until (almost) everybody has upgraded to hardware where that is unnecessary.
- coldtea 8y ago
  Well, now is close to that time. Everybody who cares for graphics already has a Retina Mac or a hi-dpi display.
  - jordache 8y ago
    > Everybody who cares for graphics already has a Retina Mac or a hi-dpi display. Elementary level of product owner analysis I must say. Proactively constraint your product into a niche category.. tsk, tsk.
    - coldtea 8y ago
      >Proactively constraint your product into a niche category.. tsk, tsk. Yeah, it hasn't been working for Apple this past 2 decades. They're only near the $1 trillion mark, could have been way better...
- jchw 8y ago
  One problem: people have no good reason to. The software folks have a good reason to remove subpixel AA; it's a hack to get better horizontal spatial resolution out of cheap monitors. The users have no reason, however, to buy better, now reasonably-priced, displays that would make subpixel AA unnecessary. By removing subpixel AA at a time when HiDPI monitors are cheap, Apple is probably doing us a service. In a few years, hopefully everyone will be on a DPI that can render nice text. You may think it's unnecessary, but consider complex glyphs like Hanzi and Kanji, which really don't get enough mileage out of subpixel AA really. I know it's pretty irritating to suggest that forcing users to buy new hardware is doing them "a service." But really, I see no conflict of interest here. Apple themselves probably hasn't shipped a 96 DPI display in half a decade, and they don't even sell their own external displays, so they don't stand to gain that much here other than software performance.
  - admax88q 8y ago
    > In a few years, hopefully everyone will be on a DPI that can render nice text. My 1x monitor can render text just fine using sub-pixel AA. It's a proven technique that has been around for ages and the text looks great on it. > consider complex glyphs like Hanzi and Kanji, which really don't get enough mileage out of subpixel AA really. If I start to become a heavy user of Hanzi and Kanji then I will consider purchasing a monitor that displays it better, but currently not an issue. > Apple is probably doing us a service. Requiring me to spend money on new hardware in order to avoid a regression in my experience is somehow doing me a favour? Really? > so they don't stand to gain that much here other than software performance. I don't understand the performance argument people keep bringing up. Subpixel AA has been around for ages and worked fine on much older hardware. Why is it suddenly a bottleneck that needs to be removed when my decade old hardware can still do it?
    - jchw 8y ago
      >Requiring me to spend money on new hardware in order to avoid a regression in my experience is somehow doing me a favour? Really? No, of course not. If you phrase it like that, that is, ignore the point of removing it entirely and all of the benefits that come with it, of course it's not beneficial. However, 96 DPI was never ideal. HiDPI won't just avoid a regression, it'll improve your experience, period. And with this new software change, you'll get a performance improvement, from better caching and easier GPU acceleration, and software engineers can delete thousands of lines of potentially buggy code made to work around the interaction of subpixel rendering with other graphics code. Like, subpixel rendering is easy to understand in theory and that might delude one into thinking that it's just a minor annoyance, but it's an annoyance that permeates the stack. You have to special case the entire stack wherever text is concerned. Effectively, the text is being rendered at 3x the horizontal resolution of everything else. This brings in complication across the board. An example is that you pretty much have to do the alpha blending in software, and when you do it requires knowledge of the final composition. Doing alpha blending in software sounds easy... But it's only easy if you do it wrong. Dealing with gamma for example. Which makes life worse because you already have to do this for the other part of your graphics library, probably with a case for both GPU acceleration and software. Basically, you'd have to render the whole screen at 3x horizontal resolution to remove the need for all of these compositing hacks. Nobody does this because subpixel AA was a hack from the get go. In an era where battery life and TDP are actually important, finite resources, we want to remove these hacks. P.S.: if you've never tried rendering large amounts of text on screen before, I highly recommend you try writing a piece of software that needs to do this. A few years ago I struggled to get 60fps rendering text for an IRC client, and of course caching the glyphs was horrendously complex thanks to subpixel AA. It would've been easy if I could've just used GPU compositing like all other graphics, but instead I resorted to hacks everywhere. Maybe today's computers don't need the hacks, but it's worth noting that yes, rendering fonts is resource intensive - you're just living on top of decades of optimizations and hacks to make it look like it's easy.
- vbezhenar 8y ago
  You can use Windows for that. Apple have different mindset and it shows in many aspects, e.g. removing big USB ports, audio jacks.
- hrktb 8y ago
  Mac sales massively happen on machines with retina display though. A 1x monitor will be a cheap second monitor or something attached to a Mac Mini (please keep it alive Apple) or Mac Pro (one day it will come). From Apple's point of view I imagine all of these are just super small niches.

adrianN 8y ago

Maybe we could wait with the step where we don't simulate better hardware until (almost) everybody has upgraded to hardware where that is unnecessary.

coldtea 8y ago
Well, now is close to that time. Everybody who cares for graphics already has a Retina Mac or a hi-dpi display.
- jordache 8y ago
  > Everybody who cares for graphics already has a Retina Mac or a hi-dpi display. Elementary level of product owner analysis I must say. Proactively constraint your product into a niche category.. tsk, tsk.
  - coldtea 8y ago
    >Proactively constraint your product into a niche category.. tsk, tsk. Yeah, it hasn't been working for Apple this past 2 decades. They're only near the $1 trillion mark, could have been way better...
jchw 8y ago
One problem: people have no good reason to. The software folks have a good reason to remove subpixel AA; it's a hack to get better horizontal spatial resolution out of cheap monitors. The users have no reason, however, to buy better, now reasonably-priced, displays that would make subpixel AA unnecessary. By removing subpixel AA at a time when HiDPI monitors are cheap, Apple is probably doing us a service. In a few years, hopefully everyone will be on a DPI that can render nice text. You may think it's unnecessary, but consider complex glyphs like Hanzi and Kanji, which really don't get enough mileage out of subpixel AA really. I know it's pretty irritating to suggest that forcing users to buy new hardware is doing them "a service." But really, I see no conflict of interest here. Apple themselves probably hasn't shipped a 96 DPI display in half a decade, and they don't even sell their own external displays, so they don't stand to gain that much here other than software performance.
- admax88q 8y ago
  > In a few years, hopefully everyone will be on a DPI that can render nice text. My 1x monitor can render text just fine using sub-pixel AA. It's a proven technique that has been around for ages and the text looks great on it. > consider complex glyphs like Hanzi and Kanji, which really don't get enough mileage out of subpixel AA really. If I start to become a heavy user of Hanzi and Kanji then I will consider purchasing a monitor that displays it better, but currently not an issue. > Apple is probably doing us a service. Requiring me to spend money on new hardware in order to avoid a regression in my experience is somehow doing me a favour? Really? > so they don't stand to gain that much here other than software performance. I don't understand the performance argument people keep bringing up. Subpixel AA has been around for ages and worked fine on much older hardware. Why is it suddenly a bottleneck that needs to be removed when my decade old hardware can still do it?
  - jchw 8y ago
    >Requiring me to spend money on new hardware in order to avoid a regression in my experience is somehow doing me a favour? Really? No, of course not. If you phrase it like that, that is, ignore the point of removing it entirely and all of the benefits that come with it, of course it's not beneficial. However, 96 DPI was never ideal. HiDPI won't just avoid a regression, it'll improve your experience, period. And with this new software change, you'll get a performance improvement, from better caching and easier GPU acceleration, and software engineers can delete thousands of lines of potentially buggy code made to work around the interaction of subpixel rendering with other graphics code. Like, subpixel rendering is easy to understand in theory and that might delude one into thinking that it's just a minor annoyance, but it's an annoyance that permeates the stack. You have to special case the entire stack wherever text is concerned. Effectively, the text is being rendered at 3x the horizontal resolution of everything else. This brings in complication across the board. An example is that you pretty much have to do the alpha blending in software, and when you do it requires knowledge of the final composition. Doing alpha blending in software sounds easy... But it's only easy if you do it wrong. Dealing with gamma for example. Which makes life worse because you already have to do this for the other part of your graphics library, probably with a case for both GPU acceleration and software. Basically, you'd have to render the whole screen at 3x horizontal resolution to remove the need for all of these compositing hacks. Nobody does this because subpixel AA was a hack from the get go. In an era where battery life and TDP are actually important, finite resources, we want to remove these hacks. P.S.: if you've never tried rendering large amounts of text on screen before, I highly recommend you try writing a piece of software that needs to do this. A few years ago I struggled to get 60fps rendering text for an IRC client, and of course caching the glyphs was horrendously complex thanks to subpixel AA. It would've been easy if I could've just used GPU compositing like all other graphics, but instead I resorted to hacks everywhere. Maybe today's computers don't need the hacks, but it's worth noting that yes, rendering fonts is resource intensive - you're just living on top of decades of optimizations and hacks to make it look like it's easy.
vbezhenar 8y ago
You can use Windows for that. Apple have different mindset and it shows in many aspects, e.g. removing big USB ports, audio jacks.
hrktb 8y ago
Mac sales massively happen on machines with retina display though. A 1x monitor will be a cheap second monitor or something attached to a Mac Mini (please keep it alive Apple) or Mac Pro (one day it will come). From Apple's point of view I imagine all of these are just super small niches.