mgaunard

47 ms·

811.
by mgaunard 3y ago
The concern isn't that humans are magic, it's that we haven't found the trick yet. When we do, most likely there will be a big revolution. But we haven't and throwing more money at it isn't going to make us reach it
812.
by mgaunard 3y ago
Today I discovered _mm512_mask_compressstoreu_epi64. Neat instruction.
813.
by mgaunard 3y ago
A normal network runs at 10Gbps or 1.25GiB/s. 200MiB/s is too slow by a factor of 7. Though obviously some people have optimised it.
814.
by mgaunard 3y ago
We'd had computers before we started making them out of silicon semiconductors. Computer Science still puts the same restrictions on what we can compute.
815.
by mgaunard 3y ago
That's the whole problem of inference: it's just wrong outside of the trivial cases. Humans are not limited to inference-based reasoning.
816.
by mgaunard 3y ago
isn't the pronunciation wrong for sublicense and noninfringement?
817.
by mgaunard 3y ago
You could encode the data in a format that would be resistant to this transformation. We'd need to know the specifics of this transformation to build something adequate.
818.
by mgaunard 3y ago
It's not difficult to consume third party C++ libraries -- what's difficult is to ensure third party code is maintained up to your standards and compatible with all the other choices you've made about your environment.
819.
by mgaunard 3y ago
Dependency hell means having many dependencies that are related to each other in non-trivial ways, making it difficult to ensure a compatible set of versions is maintained.
820.
by mgaunard 3y ago
Yours looks much better to be honest.
821.
by mgaunard 3y ago
I'm not too familiar with Rust, but took a look at the Cargo.lock file of this project. It depends on half the universe for some reason, sometimes multiple versions of the same library. And that doesn't even include the main depen
822.
by mgaunard 3y ago
The parent of the thread is https://news.ycombinator.com/item?id=39950725 It argues that you shouldn't debug optimized builds. Obviously, as I pointed out, a real programmer needs to be able to debug optimized builds.
823.
by mgaunard 3y ago
The bugs you find while working on something are trivial ones; the hard bugs come from integration with other components. Additionally most software runs as a service this day, so you have fast deployment iterations to collect errors. Even
824.
by mgaunard 3y ago
The crash when your program is distributed is the normal workflow. What matters is that your program runs flawlessly in its distributed environment, not that your test suite passes locally. Testing is but a proxy to achieve the true goal, a
825.
by mgaunard 3y ago
Of course you use the debugger to debug the core. Doesn't change the fact it's completely unrealistic to expect that you can do that with a full debug build. The industry has moved towards software as a service, and sometimes your
826.
by mgaunard 3y ago
Debug/sanitized builds are for unit tests. Crashes in production happen with optimized builds. You need to be able to inspect the core and figure out what happened from there, as usually you can't reproduce the scenario.
827.
by mgaunard 3y ago
Debugging optimized binaries is harder than debugging non-optimized ones, and requires more advanced debugging skills and gdb knowledge. There is a reason why crash dumps on optimized builds used to be called "guru meditation". On
828.
by mgaunard 3y ago
I feel like btrfs has been in development since forever and getting no adoption at all. When is the year of the btrfs file system coming?
829.
by mgaunard 3y ago
Why would you be downvoted for speaking facts? Also why should you restrain from pointing out weaknesses in other people's comments due to the fear of negative karma? Karma is meant to be burned.
830.
by mgaunard 3y ago
You need to be able debug optimized builds. That's a requirement of the real world if you're a software engineer. Just learn how to use gdb, it's quite capable.
831.
by mgaunard 3y ago
well clearly that doesn't work, you can't pass it a buffer to write into. This necessary allocates, and I guess it formats on the hot path too.
832.
by mgaunard 3y ago
Web/cloud tech is usually at the complete opposite of the software development spectrum from where C++ shines. I thought they invented Go for these things?
833.
by mgaunard 3y ago
That's why you need an extensible serialization system: to handle data that isn't copyable or whose copies would require memory allocation.
834.
by mgaunard 3y ago
Users should make their own opinions and evaluate what is the solution that best satisfies their requirements -- which in most cases should be to build it themselves (third party code is mostly a liability in software development). It seems
835.
by mgaunard 3y ago
Most logging libraries are bad because they're not designed to account for the requirements of a real-time system. What you need: - an asynchronous design where both the formatting and the writing to disk can happen on another thread.
836.
by mgaunard 3y ago
Not nearly as good as the explanation of the Fourier transform we had the other day.
837.
by mgaunard 3y ago
-Wstringop-overflow is the first warning I disable because of all the false positives. I doubt the analyze variant would fare any better.
838.
by mgaunard 3y ago
> How did you get the idea that I mistakenly believed WG21 is a "single person or entity"? > > classic WG21 choice > This is how WG21 ended up here. Programmers who believe that it just "doesn't really make m
839.
by mgaunard 3y ago
Ignoring the fact that WG21 is not a single person or entity, but 200 people with disjoint interests, your whole argument doesn't make sense. std::set preserves sorted as an invariant, so there is no such precondition on its lower_boun
840.
by mgaunard 3y ago
I've been involved with the C++ standards committee for 12+ years and my experience is that it is largely political; championing a proposal is more of a demagogic endeavor than anything else. Influential players have been priming contr

More ›