mgaunard

33 ms·

661.
by mgaunard 2y ago
I believe clear and succinct writing is the key to effective communication.
662.
by mgaunard 2y ago
For me any potential technical argument and innovation is completely drowned in the needlessly pervasive anti-capitalist genderfluid digital nomad hippie talk.
663.
by mgaunard 2y ago
"a small collective exploring the failability of modern tech" really just means "a couple living on a boat".
664.
by mgaunard 2y ago
Most of those bugs are about crashes during compilation or compilation being slow, and also this is 30 years worth of bugs. Of note, most of those (few) bugs tend to be related to specific language extensions.
665.
by mgaunard 2y ago
libbacktrace does it all, and there is also that feature in the C++ standard library now.
666.
by mgaunard 2y ago
-O3 only breaks your code if it is invalid to begin with.
667.
by mgaunard 2y ago
The rationale is explained in the article; it's for speed.
668.
by mgaunard 2y ago
That's some pretty clean bash. In my experience most bash code is badly written and inefficient, but this is a good example of code that isn't.
669.
by mgaunard 2y ago
Clearly you didn't understand what emplace does. Go is simpler because it is limited in functionality.
670.
by mgaunard 2y ago
The real-time trading systems are in C++, while the data analysis, which happens offline, is either done in Q, R or Python (pandas, polars, etc.)
671.
by mgaunard 2y ago
The whole point of SMT is to maximize utilization of a superscalar execution engine. I wonder if that trend means people think superscalar is less important than it used to be.
672.
by mgaunard 2y ago
because the core of what they do is working with time series and making analysis and predictions.
673.
by mgaunard 2y ago
what's the difference between K, Kx, kdb+, Q? All the companies I see listed with these actually mostly use Q, which is a superset of K.
674.
by mgaunard 2y ago
Just using the standard image they provide for Kubernetes. I didn't do the work of reverse-engineering all the bloat that's in there.
675.
by mgaunard 2y ago
One thing I found interesting recently while working with cloud instances (I usually work with traditional datacenters), is that those things boot extremely slowly. The whole point of those new complex init systems was to boot faster. It de
676.
by mgaunard 2y ago
say you want the median value since the beginning of the day for a time series that has 1000 entries per second, that's potentially hundreds of gigabytes in RAM, or just a few variables if you use a p-square estimator.
677.
by mgaunard 2y ago
any approximation gives a bound on the error.
678.
by mgaunard 2y ago
You could also use one of the streaming algorithms which allow you to compute approximations for arbitrary quantiles without ever needing to store the whole data in memory.
679.
by mgaunard 2y ago
There are other commercial reverse debuggers, like Undo.
680.
by mgaunard 2y ago
That's because there is no systematic solution to the real problems.
681.
by mgaunard 2y ago
most security problems are not memory safety problems.
682.
by mgaunard 2y ago
Nice biased article, I'm surprised they didn't mention Rust.
683.
by mgaunard 2y ago
It's all a guess. You need to find a job where you're happy and that pays well. You don't really know ahead of time whether you'll be happy (or even really what makes you happy) and, if the pay includes any upside, you d
684.
by mgaunard 2y ago
Argument is that KUTrace is faster, which honestly I'm not sure of.
685.
by mgaunard 2y ago
Why not just use the eBPF system?
686.
by mgaunard 2y ago
There are millions of financial instruments and their price change 100 million times per day. That's a lot to process, so yes, it matters.
687.
by mgaunard 2y ago
Representing prices is difficult because a lot of financial instruments have widely different values and price increments. Decimal floating-point gives you the range to cover everything but is generally very inefficient to process. What you
688.
by mgaunard 2y ago
Your hardware is configured by your software. You pre-arm transactions to be released when a trigger occurs. Releasing the transaction when the trigger occurs is the fast path, and what the hardware does. The slow path is what actually dec
689.
by mgaunard 2y ago
In HFT you typically only measure latency for events that you did react on, not all events (at least, in the systems that I've built). That's arguably a bit problematic as events you don't react to can cause you to fall behin
690.
by mgaunard 2y ago
Why not send a patch?

More ›