"Windows 95 was 30 MB" is such an ignorant, obnoxious, trite take. a triple buffered framebuffer (which you want for smooth scrolling) for my 4K display is 70 MB in *pixels alone*. Obviously a complete webpage with precomposed textures would take more.https://twitter.com/julienPauli/status/1042113172143067138 …
-
Show this thread
-
Replying to @whitequark
"Precomposed textures" doesn't sound very reasonable for a webpage. I can't think of any good reason for a webpage whose purpose isn't delivering high-res media to be more than a few hundred kB.
1 reply 0 retweets 12 likes -
Replying to @RichFelker
The current generation of browsers is optimizing for input latency at cost of memory consumption, which is a tradeoff that isn't right for everyone (probably not for you!) but is a decision that should be acknowledged as valid
2 replies 1 retweet 27 likes -
Replying to @whitequark
Maybe I misunderstand; I thought you were talking about loading that much data over the network, not rendering pipeline on the client. Still not sure how it helps input latency though.
1 reply 0 retweets 0 likes -
Replying to @RichFelker @whitequark
Ideal performance and memory usage should be achieved by compressing all image data, decompressing on the fly when displaying it. (Decompressing jpeg is a lot faster than memcpy if done right.)
1 reply 0 retweets 0 likes -
Replying to @RichFelker
ideal performance and power usage (which is more important than memory usage for the vast majority of browser users today) is achieved by doing everything on GPU, which precludes this kind of tricks
2 replies 0 retweets 0 likes -
Replying to @whitequark @RichFelker
when just scrolling, you do not want to touch the DOM or pixels on the CPU at all unless you need to reflow the webpage. you render precomposed textures into the framebuffer, including any desired animation or such, all done on the GPU. Servo does this well, WebKit not so much
1 reply 0 retweets 0 likes -
Replying to @whitequark @RichFelker
in theory if you had non-interleaved RAM dice and the OS could compact physical memory and manage NVM cache cleverly, using less RAM would actually reduce power consumption by powering off unused RAM, but I think it's not significant enough to spend time getting it to work
1 reply 0 retweets 2 likes -
Replying to @whitequark
Yeah, an ideal laptop/mobile memory architecture would work like that. With a smart OS, you'd get sleep lifetimes comparable to hibernate, but with instant resume.
2 replies 0 retweets 0 likes -
Replying to @RichFelker
by the way, do you know where GPU compression provides huge gains? on modern laptops GPU compresses the framebuffer before sending it to the *display panel*, which lets it power down the LVDS link, which has power savings on the order of 1 W (!)
2 replies 1 retweet 6 likes
I mean power down the LVDS link faster, of course.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.