WallWextra

First, I would dispute that a cortex A9 is comparable to a Core 2. Second, if ARM is in fact getting that close to Intel, I would attribute it to diminishing returns, and not some mythical property of the ISA that makes it difficult to implement.

I'm not sure what you mean by wide SIMD extensions providing "most of the chip's power." If you mean for vectorized DSP loops or whatever, sure, but I was talking about performance on branchy pointer-chasing spaghetti, which is what most real code is, and where it would be the hardest to catch up to Intel.

A lot less "interesting," in a very good way.

Also, OSX applies unicode normalization to the filenames, which can cause some interoperability problems with unix, e.g. you try to copy files forth and back between the systems and end up with two copies of a file, with identical-looking filenames that differ in their unicode normal form.

It is here: http://www.azulsystems.com/blog/cliff/2008-11-18-brief-conve...

It looks like Azul has basically given up on their custom hardware, I guess? They found a way to make their GC work on commodity x86.

Replacing x86 with a RISC instruction set would do little good. Performance doesn't come from the RISCness of the decoded uops, it comes from superscalar out-of-order execution, big caches, finely-tuned branch predictors, forwarding networks, a good memory controller, etc. That all takes up more transistors than x86 decode. POWER8 is the only RISC that is comparable to high-end Intel parts, and it is not a particularly small chip.

ARM used to have some ISA extensions for directly-executing JVM bytecodes, but it's deprecated. An optimizing JIT compiler gets much better results.

Charge balancing is a common feature, often built in to charge controller ICs. Of course Tesla has to do this on a much larger scale, like with all of its power electronics. But so does a Prius.

I'm not sure how much of a technological lead Tesla has versus major car companies already producing hybrids and electrics, but it can't be that huge, and those guys have way bigger R&D budgets.

Oops! Good catch.

If Apple tried that, Intel would stop giving them such sweet prices.

Even if Apple could match Intel's performance in the laptop power envelope (implausible), that still wouldn't be enough. They'd have to beat it by a big enough margin to make binary translation feasible. When they switched from PPC to x86, they got enough of a performance boost to make Rosetta Stone work ok, despite the overhead. I'd be surprised if it's even possible to beat the current crop of x86 processors in single-threaded performance by such a huge margin, ever. Let alone beat whatever Intel has out by that time.

The reasons I've heard were to kill US oil exploration, and also to teach the high-cost producers (Venezuela, etc) in OPEC a lesson. It's common for these countries to "accidentally" pump over quota to make some extra cash. So Saudi just refused to cut the quota this time around. The Saudis could also be interested in messing with Iran, which has much higher production costs.

On a Galaxy S5, it works glitchily in Chrome and not at all in Firefox.

It's worth pointing out that Sandy Bridge was a massive microarchitectural change from its predecessors, using a totally different out-of-order engine.

I don't know why the performance characteristics are so similar. I guess once you have a big fat memory subsystem on-die, everything else is diminishing returns?

I can't think of a worse idea than prosecuting Bush administration officials. Jailing the previous administration would set a terrible precedent, and would fatally undermine the stability of our government.

There are many people who are convinced, e.g., that abortion is murder and that politicians who enable it ought to be prosecuted. Or, more on topic, that Diane Feinstein is a traitor for making these torture memos public. They believe this with as much moral certitude as anyone here believes torture is a crime. It is only a matter of time before another right-wing president is in the White House, and when there is I don't want there to be a precedent for jailing one's political rivals.

It's great to have a coarse-grained sanity check from a trusted source, but I don't think this is really a plausible replacement for NTP (not that you need NTP when you have a mobile baseband and GPS).

Semaphores are basically unused in the kernel these days, abandoned in favor of mutexes. I really recommend reading kernel/locking/mutex.c. You can learn a lot about the details of low-level synchronization, and it's also just a really impressive piece of ruthlessly-optimised code. Tricks like reading the lock's 'owner' field without any sort of synchronization, to decide whether to spin on the lock or to sleep.

Using two semaphores to implement a bounded buffer has the advantage that writers and readers don't interfere with eachother until the queue is either empty or full. Of course, that might be the only good use of semaphores, and a direct implementation can be better then one using standard semaphores.

Semaphores can be implemented in a way that doesn't require a spinlock in the fast path. Still not as fast as you can make a mutex, but a lot better than the pseudocode given in many classes (which is almost always wrong, too.)

The Pentium Pro was what killed Alpha and MIPS, not Itanium.

It would have been nice if they had a MIPS core actually designed by MIPS.

Uber is a bunch of sketchballs and a bunch of sockpuppets flagging the story is totally within the realm of plausibility.

The intra-frame format of every video codec.

ARM has that inline shifter and predicated instructions, both of which were removed in arm64. Those are the most salient things.

Pugh taught my ODEs course in college and basically turned it into a dynamical systems party. Great class, but I still don't know anything about ODEs. Really cool guy.

Construction of the reals is in an appendix at the back. He starts with ordered field axioms.

Frankly there's something about the presentation in "discrete math" books that I find much more confusing and difficult than Spivak, which is just talking logically about numbers and their properties.

There are also some very good books by Israel Gelfand, written for a high-school level. Titles are Algebra, Trigonometry, The Method of Coordinates, etc. All by Gelfand and one other author. Sort of unusual because they are basic books written by one of the greatest mathematicians of the 20th century.

I honestly couldn't tell what he meant, whether he had stumbled through calc 3 and gotten sick of memorizing arbitrary rules, or if he can't solve a quadratic.

I tried to include some good high-school level math resources.

Sure, right from the axioms.

I got started on "real" math with Spivak's Calculus. Some people start with Topology by Munkres, which is not a difficult book but is very abstract and rigorous so makes a good introduction. If you feel like you have ok calculus chops, maybe Real Mathematical Analysis by Charles Pugh. Other good books are Linear Algebra Done Right by Axler, or the linear algebra book by Friedberg, Insel, and Spence. Maybe even learn linear algebra first. It's so useful.

Do plenty of exercises in every chapter, and read carefully. Count on about an hour per page (no joke). Plenty of math courses have their problem sets published, so you can google a course which uses your chosen book and just do the exercises they were assigned.

If you don't feel comfortable with basic algebra and other high school math, there's Khan Academy, and some books sold to homeschoolers called Saxon Math.

If you haven't had a course in calculus before, maybe you should skim a more intuitive book before or alongside reading Spivak. I don't know of any firsthand, but I heard Calculus for the Practical Man is good. Scans are freely available online (actually, of all these books) and Feynman famously learned calculus from it when he was 12.

I've sort of been assuming that Julia is capable of inlining the vectorized notation into a single traversal over the array, at least for simple types. Is that not true?

EDIT: Not yet, I guess. http://pastebin.com/Tw5PuCcJ

It's longjmp.

JPEG 2000 isn't actually better than plain old jpeg.