Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
v-yadli
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
20 ms
·
121.
▲
by
v-yadli
9y ago
I partially agree with you, in that the rendering of the page is good-looking in circa 2007 standard on a mobile device. But rendering one page nicely does not prove it is full-featured. It has to go through some kind of benchmark, which re
122.
▲
by
v-yadli
9y ago
http://mobilehtml5.org/ I'm interested in how you would define full-featured . Please check the symbian & opera columns. Also, iOS 1.0 ships with Safari3.0, not Safari3.1.1 in this test. In iOS 2.0 they introduced
123.
▲
by
v-yadli
9y ago
> Yes, you have webworkers for multi threaded development. They're basically independent applications which run on different threads and you pass messages (which are simply objects) between them. The browsers themselves are also mov
124.
▲
by
v-yadli
9y ago
Can you run web apps in a multithreaded environment? UI remains the largest overhead in a web app in my opinion.. Or, how much speedup would you estimate, if we convert all GoogleDocs functionalities into Word97? I'd estimate 1000 time
125.
▲
by
v-yadli
9y ago
For 1) and 2), I have an idea that it would probably be better if we can have an "Electron factory" running in the background and serve all these apps. This turns the OS task manager into tabs of a browser, and any optimization do
126.
▲
by
v-yadli
9y ago
I think there's misunderstanding here, because I don't think that electron is the only way to go. Like the author of oni wrote here, we can make this frontend happen on any web-friendly runtime. What I mean is that we should put a
127.
▲
by
v-yadli
9y ago
I'd agree with you on the basic idea, but practically from my experience, the qml+js subsystem seems like an isolated piece of "glueware" upon a legacy platform and they do not interpolate very well. Once I did a knowledge gr
128.
▲
by
v-yadli
9y ago
Are you expecting another Visual Studio or IntelliJ? I think Electron is a very cost efficient dev environment. Can't really think of writing all this up in plain C++, wiring up native calls for various OSes, leveraging GPU, support ra
129.
▲
by
v-yadli
9y ago
Likewise, your text objects aren't my text objects. For a fair comparison I think we should first assume compatibility for both unstructured, and structural data. Otherwise I would challenge that your bash script is not my bash script.
130.
▲
by
v-yadli
9y ago
Totally agree. And it turns out not caused by the layout algorithm but exactly the rendering. We've been drawing on canvas without proper acceleration for too long.
131.
▲
by
v-yadli
10y ago
I really love how the system sounds create a musical sense of progression (ascending clips with varying patterns), that you are getting closer and closer to setting things up successfully. It turns the OS itself into something that feels li
132.
▲
by
v-yadli
10y ago
This is Trinity.
133.
▲
by
v-yadli
10y ago
Check this out: https://www.graphengine.io/downloads/papers/Trinity.pdf
134.
▲
by
v-yadli
10y ago
Sure we'll definitely update the docs! Or better to work out one way to put the docs on GitHub and keep it synchronized with the site (they're markdown anyway)
135.
▲
by
v-yadli
10y ago
Yes, we will continue working on bringing both the core runtime and the toolchain (TSL compiler, data management tools, more computation modules, language bindings etc.) online and ready for the *nix world.
136.
▲
by
v-yadli
10y ago
At the current state you could link Trinity.C with a custom Mono host and then execute your GE assemblies/Trinity.Core. Check out CMakeFile.txt for details. :) We're working to interface Trinity with CoreCLR. .NET standard 2.0 wou
137.
▲
by
v-yadli
10y ago
Dev here. Cross-platform for Linux/BSD is on the way, and it's already in the repo now.
138.
▲
Microsoft Open Sources Graph Engine and Language Integrated Knowledge Query
(github.com)
11 points
by
v-yadli
10y ago
|
0 comments
139.
▲
by
v-yadli
10y ago
This could be very useful as a codegen alone.
140.
▲
by
v-yadli
11y ago
yep, thanks for the clarification!
141.
▲
by
v-yadli
11y ago
make extensions on the original. use the single tape to simulate multiple tapes, one for stack(or registers), one for ram, one for rom, and another for encoding how the hardware behaves against the other tapes.
142.
▲
by
v-yadli
11y ago
Talks about desktop interface, and leverages a touch pad. When falling back to a mouse, this design language would not be effective.
143.
▲
by
v-yadli
11y ago
isn't this the minmax of the peak rate?
144.
▲
by
v-yadli
11y ago
Please note that this test also only captures a single profile of Graph Engine. The cell entry lookup table and the MM module are designed to operate in a more complex environment and not optimized towards the direction of this micro benchm
145.
▲
by
v-yadli
11y ago
I'm pretty sure the values in entry will be loaded into registers.
146.
▲
by
v-yadli
11y ago
I don't have a Mac at hand and the best effort I can do is to finish the 90 seconds. I would suggest an attempt to include more info in the stock view... Even the metro map have station names. :-) People do not recognize commits with j
147.
▲
by
v-yadli
11y ago
No commit logs in the graph? I would prefer a list view instead...
148.
▲
Microsoft Graph Engine 1.0
(research.microsoft.com)
19 points
by
v-yadli
11y ago
|
5 comments
149.
▲
by
v-yadli
12y ago
Dat overdriven bassline!
150.
▲
by
v-yadli
12y ago
I thought you mean calc.exe
More ›