jrk

15 ms·

91.
Google publishes algorithm behind Pixel's “highest rated smartphone camera”(hdrplusdata.org)
3 points by jrk 10y ago|0 comments
92.
by jrk 10y ago
Breaking on every kernel upgrade is a configuration issue (and a result of Linux's driver model); the "quality" of the NVIDIA Linux driver being discussed is an issue of the actual performance, compliance, stability, and feat
93.
by jrk 10y ago
I'm curious: why is the pricing linear in RAM for the S* and M* types, but double that for the L* types? L3 = 4xM3, but the pricing is 8xM3. The pricing is very appealing on the lower end, but much less appealing for the still-quite-sm
94.
by jrk 10y ago
I quibble with your label of "bad GPU," though the relative magnitude of resources and performance is right. Modern GEN is actually a good-to-great GPU architecture (an assessment widely shared in the GPU architecture community to
95.
by jrk 10y ago
The paper isn't up yet, but Disney has a paper at SIGGRAPH this year on a DSL and compiler for controlling knitting machines: http://s2016.siggraph.org/technical-papers/sessions/cloth Unlike most additive man
96.
by jrk 10y ago
"Colistin is widely used in Chinese livestock" oh for fuck's sake…
97.
by jrk 10y ago
In addition to open source drivers, there has been work to reverse engineer the binary formats and write open source assemblers for recent versions of SASS (e.g., https://github.com/NervanaSystems/maxas ).
98.
by jrk 11y ago
The close parallel is AsmJit: https://github.com/kobalicek/asmjit . I'm curious how it compares.
99.
by jrk 11y ago
I clearly should have known, but: oh my god the comments section…
100.
by jrk 12y ago
I think the idea is not that they wouldn't have goggles—these lab prototypes have goggles—but that they might be much more constrained than a free, walk-around head mount. Think: Imagineering-like controlled experiences down to arcades
101.
MarkovRMS, automatically generated from Richard Stallman's email correspondence(twitter.com)
1 points by jrk 12y ago|0 comments
102.
by jrk 12y ago
The "hybrid" IO architecture is historically known as AMPED (asynchronous multi-process event driven): https://www.usenix.org/legacy/event/usenix99/full_papers/pai...
103.
by jrk 12y ago
This fails to launch with the 1.8.0_20 JRE installed on Yosemite (the only supported version, now that Apple Java 1.6 is gone). Fails with "No matching VM found." Anyone from JetBrains: how does this work with newer Java releases
104.
by jrk 12y ago
Dear god, can we please stop saying "Golang" before it's too late?
105.
by jrk 12y ago
This does not read like the investigative work of someone who has actually read any technical history on the subject. (The Making of the Atomic Bomb is probably the best scientific history ever written—I cannot recommend it highly enough.)
106.
James Mickens drops wisdom on distributed systems(vimeo.com)
3 points by jrk 12y ago|0 comments
107.
by jrk 13y ago
(My post got long, so I've put a standalone version, with better markup rendering, here: http://gist.io/9161898. ) [TikZ] is a perfect example of TeX's missed potential: it is an outstandingly well designed DSL for
108.
by jrk 13y ago
The rated CPU speed is nearly meaningless for machine responsiveness or even bursts of heavy usage (Photoshop edits; compiling changed files) given the dynamic voltage and frequency scaling ("Turbo"). Both the "1.7 GHz"
109.
by jrk 13y ago
Yes, ARMv8 has twice as many architectural GPRs and NEON/FP registers, but: 1. This has nothing to do with 64-bitness, it could just as easily have been applied to a 32-bit iteration of the ARM ISA. (You're not saying it does, b
110.
by jrk 13y ago
They did say something fairly concrete: they said it's "twice as fast" as the A6 in "CPU performance." This has nothing to do with virtual address size and pointer width (which only have a negative impact on relat
111.
by jrk 14y ago
…and even with simple point-wise filters, Instagram had to manually code all its filters for OpenGL ES to run on the GPU to get reasonable performance. Last I heard (months before the acquisition), only one or two people at the company coul
112.
by jrk 14y ago
It's also wrong on the opposite front: ZFS cannot change geometries on the fly. I am a fairly big fan of ZFS, but the ability to mix disk sizes and change numbers of disks in a RAIDZ set is not why--it effectively cannot do either.
113.
by jrk 15y ago
Moore's law only describes silicon processes: http://en.wikipedia.org/wiki/Moores_law It has nothing to do with computing architectures.
114.
by jrk 15y ago
This is precisely the analysis given in the per-transistor pricing crossover chart. That shows that the old rate of wafer price increase still allowed for significant per-transistor cost advantage fairly quickly. The whole point is that the
115.
AbstractSingletonProxyFactoryBean is abstract(static.springsource.org)
22 points by jrk 15y ago|7 comments
116.
by jrk 16y ago
$2.5B/yr seems like a lot of added revenue for the industry, but this is a very, very high-revenue, low-margin business. Has anyone actually modeled the cost of especially first bag fees? Some obvious sources include: - Longer boarding ti
117.
The Unofficial Guide to Migrating Off of Google App Engine(www-cs-students.stanford.edu)
163 points by jrk 16y ago|57 comments
118.
by jrk 16y ago
ML: There is not enough radical innovation in Silicon Valley. Focus is too narrow. Too many companies that are featurettes. Companies…like Slide?
119.
by jrk 16y ago
They aren't as essential for desktop browsers (now), since the relative resource constraints on a given tab are so much smaller, but just as relevant as "desktop" runtimes scale down in power/up in efficiency demands, and as the apps run wi
120.
by jrk 16y ago
Yes, MobileSafari plays music (with <audio> and direct links to MP3/M4A, etc.) quite well. The biggest issue with this as a path to replace native audio playback apps is how the browser manages resources for background tasks (tabs):

More ›