Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
brigade
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
62 ms
·
691.
▲
by
brigade
14y ago
Very few TVs use IPS, since IPS tends to have worse black levels and pixel response time than PVA/MVA (alternatively: it's more expensive to get equivalent specs out of IPS), which are the two specs that TV reviewers care about (black level
692.
▲
by
brigade
14y ago
But the buffering is not required because of the HDMI link; a display could update each pixel as it receives it over HDMI, just like CRTs did. LCDs buffer at least one frame for many reasons, but "because HDMI requires it" is not among them
693.
▲
by
brigade
14y ago
€100,000 was insured. If you want to go along that line of thinking, then he currently has access to less than 4% of the uninsured money.
694.
▲
by
brigade
14y ago
Not everyone is in a relationship. The forcing is due to money - if you're poor enough, then you simply can't afford even a studio and are forced into having roommates to split the cost of rent.
695.
▲
by
brigade
14y ago
It's kind of interesting that he thinks anonymous questions should have just died, due to people abusing them. Whereas my experience of the site was that e-bullies would simply sign up new accounts immediately after being banned, since no b
696.
▲
by
brigade
14y ago
No, the compiler couldn't have kept resVal in a register because it doesn't know resVal doesn't alias with any of val1..val7. Therefore, it must store resVal each iteration, and it must load each of val1..val7 each iteration. Otherwise,
697.
▲
by
brigade
14y ago
No, it got lucky because it's emitting exactly what the author wrote; there's nothing complex going on. In fact, I'm sure the author is stunting where it got smart and hoisted val1+val2+val3+val4+val5+val6+val7 outside of the loop, and it c
698.
▲
by
brigade
14y ago
That's... not really subtle. That's like pipelining 101. Which is nearly 30 years old now... The only weird thing there is the compiler lucks out in generating a version that allows OoOE to save the programmer from themself.
699.
▲
by
brigade
14y ago
The only stack-specific special optimization that's done is fusing the decrement/increment of esp/rsp with the store µop. And that's done mainly since push/pop are one byte opcodes, unlike general load/store. Everything else is general memo
700.
▲
by
brigade
14y ago
One problem is finding people responsible / hardworking enough that you can rely on them doing your errands and chores without constant supervision. Someone that competent isn't likely to be working for $8-10/hour. Obviously there are some
701.
▲
by
brigade
14y ago
Unless I'm mistaken, he only compares the fees for stores. And looking at http://pages.ebay.com/help/sell/fees.html#if_fixed , it seems that for individuals, the new fixed fee of 10% is always worse for anything electronic or $100 or more.
702.
▲
by
brigade
14y ago
You can easily spend $3000 on just a photographer for your wedding. Which is actually many photographers' primary source of income.
703.
▲
by
brigade
14y ago
It isn't; there are very few flags that allow the compiler to perform optimizations not allowed by the language standard. Aliasing is not one of them for any compiler I know of. In fact, there are usually flags to go the opposite direction
704.
▲
by
brigade
14y ago
> Ability to leverage existing drivers implementing the Android driver model [1] This is huge for ARM; very few vendors bother developing any sort of an OpenGL driver for X11, and drivers for same GPU often not only SoC specific, but so
705.
▲
by
brigade
14y ago
Samsung ARM Chromebook. I'm pretty sure all the issues I'm having are specific to that model; x86 Chromebooks should work better with Ubuntu.
706.
▲
by
brigade
14y ago
Not in the default (or default-updated) 12.04 ChrUbuntu install, unless something has changed in the last 6 hours. Also using the Arch files only half fixes the trackpad - using the physical button is still very finicky compared to Chrome O
707.
▲
by
brigade
14y ago
There's ChrUbuntu, which is easy to install but gives a barely functional trackpad, no sound, and a 75% chance of hanging on wake. Supposedly 13.04 is significantly better, but upgrading is stupidly hard and I've yet to manage to do so (cur
708.
▲
by
brigade
14y ago
Personally, I'm impressed by how horribly implemented Steam's OS X gamepad support is. That takes skill to do, when other apps Just Work™. Also, Steam.app used to suck heaps of CPU just idling in the background, often enough that to make th
709.
▲
by
brigade
14y ago
> They must have had a working system to test this on?! The speculation is that Stuxnet was tested on P-1 centrifuges that the US acquired when Libya dismantled its nuclear program, set up in Israel's nuclear arms facility in Dimona. [1
710.
▲
by
brigade
14y ago
Well, I guess the sensor could theoretically be made using the entirety of a wafer. But it would be useless since the lens needed for such a large sensor would be more than the size and weight of a large ground telescope.
711.
▲
by
brigade
14y ago
1.8 Gpixel sensors don't exist, due to physics. It's combining 368 sensors. No one stores or transmits uncompressed video outside of RAM. 15Mbps is decent for 1080p24, so 1.8 Gpixel 24fps would be decent at 15Gbps. Assuming say 14 hours of
712.
▲
by
brigade
14y ago
Given that for EME to not be trivially bypassed, the CDM has to be a blob implementing its own decoders, how is this significantly different than Flash or Silverlight binary plugins? Why would whoever's developing these CDMs (each individua
713.
▲
by
brigade
14y ago
It's pretty rare for a thread to completely stall in modern CPUs as you might imagine an in-order CPU doing, but it is common to achieve less than 1 instruction/cycle throughput out of a theoretical ~3-4 instructions/cycle in the frontend a
714.
▲
by
brigade
14y ago
Opposite actually - instruction scheduling is significantly more useful for in-order processors without renaming. Compilers take advantage of OoOE and register renaming by not caring too much about careful scheduling in favor of reducing re
715.
▲
by
brigade
14y ago
Renamed registers for OoOE, not architectural registers. These number over 100 (200-300 counting FP/SSE) for modern x86 CPUs. Hyperthreading splits resources (including registers, branch prediction, TLB, etc.) between the threads. This spli
716.
▲
by
brigade
14y ago
No, it was referring to bandwidth and latency. 30ms latency is basically the best-case scenario you can expect for consumer internet (with a nearby datacenter and all that.) 10ms latency is about the worst-case latency for a rotational disk
717.
▲
by
brigade
14y ago
It seems unlikely that a Chromebook would have a gigabit link within Google's datacenters.
718.
▲
by
brigade
14y ago
Yet you did basically say that the reason teachers hated NCLB was because it would reveal them not doing their job properly, in the same paragraph as you basically saying that police hate recordings because it would reveal them not doing th
719.
▲
by
brigade
14y ago
These teachers also had some of the highest average standardized test scores in the county. How does that fit in with your personal worldview?
720.
▲
by
brigade
14y ago
It's kind of interesting how every teacher I've talked to that's taught for a decade or three agrees that NCLB is awful, yet among the general populace (HN comments), that's considered more evidence that teachers are incompetents. Possibly
More ›