yoklov

13 ms·

91.
by yoklov 8y ago
Realistically he would just claim it was those things even though it was verifiably false.
92.
by yoklov 8y ago
> physicists say it’s a reasonable assumption that there might not be anything new to find until energy scales of 100,000,000 TeV or more This is an extremely pessimistic view. Yes, we're almost certainly going to find something the
93.
by yoklov 8y ago
It seems to be related to some of the methodology used in the study. Apparently it ignores some things?
94.
by yoklov 8y ago
Worth noting that auto_ptr is decidedly not modern, it was deprecated in c++11 because it was error prone and replaced with C++11's unique_ptr (and shared_ptr/weak_ptr too sort of, although they have a different use case -- unique
95.
by yoklov 8y ago
One option: Just go further up the chain. Malloc is going to be implemented (probably) using mmap/VirtualAlloc/whatever, just teach the system about those native calls as allocations. (This probably is not a satisfying resolution
96.
by yoklov 8y ago
Right. Just to add on, one reason Quaternions are still used in graphics (despite rotation matrices not suffering from gimbal lock either) is that they're easy to interpolate between, even if you have many. If you just have two you can
97.
by yoklov 8y ago
It also needs data when you click on a link it created in the past, which there's no real way for it to know in advance. (It could possibly use a custom protocol but it would break backwards compatibility with HTTP which would be a big
98.
by yoklov 8y ago
That's true, these are always costly though.
99.
by yoklov 8y ago
Yes that works, but relies on devirtualization in the production code, and isn't at all what you were suggesting before about having separate implementation files.
100.
by yoklov 8y ago
> I don't like the idea of needing the machine to explain to me how my own code actually works. But you're fallible. This is the same reason you need to use a profiler to see if your optimizations actually sped anything up. A d
101.
by yoklov 8y ago
The problem with AVX2 accelerated code (and much of AVX) is that unless you have a lot of it to run you end up with a substantial speed hit that comes from the cost of switching to a different power bin (which often takes 1 or 2ms!) and the
102.
by yoklov 8y ago
One reason is that other test might need the production version. Building twice for that case is more of a pain and not really worth it.
103.
by yoklov 8y ago
Yes, the actual render is crisp for me.
104.
by yoklov 8y ago
I think it makes unsafe code less error prone to write, but additionally it allows you to interop with C APIs that have aliasing pointers. Also, if they were restrict, it would be UB in rust, which I think is something Rust tries to avoid w
105.
by yoklov 8y ago
It’s pretty common in both C and C++ to pass things by value, at least for more modern code. Maybe less common than in Rust, but not by that much.
106.
by yoklov 8y ago
Yep, as well as the untracked cache. I also was using a split index for a while, but it didn’t play nicely with some tools...
107.
by yoklov 8y ago
Dang, this made `git log` inside the repo I use at work (which is enormous with a stupid number of commits) nearly instantaneous. Great work. `git status` still takes over a second for me though, oh well...
108.
by yoklov 8y ago
Ignoring other issues, this will have rather poor power usage, which is especially relevant given how many people listen to podcasts on mobile devices.
109.
by yoklov 8y ago
Generally it's not like you can just swap someone's machine out without any downtime.
110.
by yoklov 8y ago
Sadly a substantial number of Mac users are on os versions that don't support Metal. And a substantial number of then wouldn't even get support if they updated their OS. So Metal-only isn't really an option for many projects,
111.
by yoklov 8y ago
> Sometimes you may want to use SSE41 even with AVX2 is available. I'm glad this is an option. I feel like a lot of libraries assume you always want the widest SIMD width possible, when this is very likely to cause performance issue
112.
by yoklov 8y ago
Didn't they just win a very similar law suit about this? https://www.techdirt.com/articles/20180606/01224439979/court...
113.
by yoklov 8y ago
Neither Lua nor LuaJIT use a generational design, they're both tri-color incremental mark and sweep collectors without a separate nursery. There was a plan for a generational GC for LuaJIT [0] (not to mention a lot of other cool featur
114.
by yoklov 8y ago
Sticking two 32 bit numbers (e.g. `(uint64_t(a) << 32) | uint64_t(b))` will not introduce bias. IIRC, the PCG C++ distribution has a 64 bit variant (it uses 128 bit integers, which are implemented in software). I don't know if th
115.
by yoklov 8y ago
Not sure I think using AVX512 makes sense in an RNG unless you're already using those instructions for a lot of stuff. Using instructions from outside the current power bin will adjust the clock rate and frequency of the core, and doin
116.
by yoklov 8y ago
> Second, it is an error to call gcc's behavior "defined in documentation" as if that means it's okay to use. It's just a flag to support obsolete code. Long-term, the assumption that gcc will retain flags to sup
117.
by yoklov 8y ago
It’s been a long time (and this was never my strongest area), but in case someone who really knows what they’re talking about doesn’t get back to you, it’s basically a particle who may or may not exist based on quantum uncertainty, but in p
118.
by yoklov 8y ago
No, I don't think building a neural net for image recognition qualifies. NP problems are essentially problems where the following two properties hold: 1. No better algorithm is known than using brute force -- generating every possible
119.
by yoklov 8y ago
This isn’t entirely untrue in theory, but in practice it is. Compilers optimize under the assumption that undefined behavior never occurs, and the standard is written with this assumption. Any conforming compiler is free to break code that
120.
by yoklov 8y ago
This is doable in rust too.

More ›