danvet

9 ms·

31.
by danvet 15y ago
Iirc the way to reboot an x86 machine is to triple-fault, i.e. execute an interrupt in an interrupt in an interrupt. Could be done by e.g. marking the pagefault handler's stack space as "not available" in the pagetables. Now if the virtual
32.
by danvet 15y ago
Exactly. All the time while I was studying I've lamented about contributing to open source, especially to the kernel. I've read tons of things, code, os books. I think I've teached myself a pretty good understanding of how it all works (nev
33.
by danvet 15y ago
On a hunch: Run Pagerank (the original thing, i.e. just the stable distribution of the random walk) on the graph of who-upvotes-whom. Then tally upvotes weighted with the upvoters pagerank. It's probably safe to assume that the pagerank of
34.
by danvet 15y ago
You still have that rough idea of social proof in the comment ordering. Of course there's the occasional very recent comment sprinkled in, but I think the error this introduces isn't worse that popular, but slight wrong comments upvoted wit
35.
by danvet 15y ago
Looks like taken with rather dim lighting, i.e. high-iso. In which case all these dots in the dark area could be explained as hot pixels (pixels that have a tiny bit of leakage that then shows up when massively amplified). Of course, that l
36.
by danvet 16y ago
The problem is that the amount of raw cpu power you can burn for a storage miss is _much_ larger than for a cache miss. So the tuning is invariable completely different, leading sometimes to completely different algorithms.
37.
by danvet 16y ago
One of the best examples for an exam like this was the final exam in history at my high school. Half of it was a few class-wide topics of 20th century history. The other half was anything of the students choosing, I've chosen the policy of
38.
by danvet 16y ago
I've seen this site before and frankly I think it's rather irrelevant. Recent innovations (past years) in search are mostly spelling corrections/related topics suggestions, visual presentation of aggregated results (how many images, results
39.
by danvet 16y ago
Which is why my current employement contract has an addadendum that exempts my leisure-time open-source contributions from any claims by my employer. Despite the fact that their default contract has some harsher restrictions/non-compete cla
40.
by danvet 16y ago
At least for me, commuting by bike is just routine. Now that my current gig is within walking distance, these about 5 hours of biking per week just disappeared. I think that's because I find biking, running, swimming, actually anything stre
41.
by danvet 16y ago
Not at all. Most series I watch don't show here (or with crappy translations), and even if they do sometimes only a few years later. So I just torrent it. If I actually like something I tend to buy it as a physical disc from amazon.com and
42.
by danvet 16y ago
No wasteful redundant resource management. It might look like that, but for every data structure in the kernel there's a use-case that absolutely needs the corresponding lookup/access to go fast. A bunch of examples from memory management:
43.
by danvet 16y ago
But you need tons of additional water if you deliver it with a helicopter.
44.
by danvet 16y ago
.. which I've lately found out the hard way: Code of mine randomly crashed and valgrind told me it's due to use of unintiliazed memory. An awful lot of time later (i.e. one valgrind --track-origins=yes run) I know that it's from function fo
45.
by danvet 16y ago
With just 100 hundred git branches, I think you're not really using git often and/or to its potential. On a busy coding day, I easily create a few branches. If you also count temporary branches on non-development testing machines (I do driv
46.
by danvet 16y ago
That would be a rather awesome feature for evil people: Just copy&paste your competitions content onto SO (or any other specially protected site) and their google ranking will drop like a stone.
47.
by danvet 16y ago
The problem is more one of cost and efficiency. You can still run your care for 100k of miles (the 1M miles is fiction, imho). But when e.g. the engine is broken, you don't fix it anymore but replace it wholesale. Less hassle, guaranty on
48.
by danvet 16y ago
As some other commenter already noted, this breaks browser text-search. In case you want to quickly checked whether what you want to say has been said already (and upvoted that comment instead or add something as a reply).
49.
by danvet 16y ago
I'd very much welcome that. My gut feeling is that on quite a few posts, discussion seems to stick to the first comment. I'm assuming that starts to happen as soon as this top-level comment has enough upvotes to quickly displace any new pos
50.
by danvet 16y ago
Frankly, 0xdeadbeef is a rather bad value for a pointer poison: It doesn't have enough leading zeros to be caught in unmapped area used to catch null pointer access (it has a high change of being a valid kernel address on linux if you have
51.
by danvet 16y ago
If I'm reading the article correctly, they do fix up all the pointers on the next gc run. For every scanned pointer just check first whether it points to a stale page (and update it), then do the liveness marking. btw, they're using 2mb pag
52.
by danvet 16y ago
Furthermore they're using 2mb pages instead of the usual 4kb (on x86). That cuts the overhead by a factor of 512 alone. The remaining problem I see is cache-trashing because on every minor page-fault you have to run a part of your mark&