Similarly, if not performance-focused, I can wholeheartedly recommend Building Git[0], which walks you through building your own git clone in Ruby (although the language is immaterial).
It's definitely possible to do correctly, but looking through the code for both crates it doesn't look like they take the necessary precautions (issuing a fence or using RDTSCP). Which is a little weird because at least quanta explicitly checks for RDTSCP support, but then doesn't seem to use it.
(I'm not a Rust expert and I'm on my phone though, so I might be missing something.)
Thanks for sharing this! I hadn't seen it before, I'll definitely add a mention in the post.
This approach is a smart way to get the same precision as the kernel without tying it to the vDSO implementation as in the post, so presumably someone at Google was worried about very similar issues. I also considered an implicit automatic refresh, but ultimately for my purposes controlling when to take the hit was more important than the slight increase in API complexity introduced by an explicit call.
This is explicitly called out in the post as well as the Intel instruction manual. Every codebase I've ever seen that reads the TSC either issues an LFENCE or uses RDTSCP.
In my benchmarks RDTSCP has a slight advantage, despite the slower latency on paper, because it doesn't fully serialise the instruction stream (later instructions can start executing, unlike with LFENCE). Whether the ECX clobber will outweigh that will depend on the situation.
You might've misunderstood the requirements. The time scale was 1-10 micros per component; 100 ns was the overhead per span we were aiming for.
In this case distributed tracing absolutely was the right choice. These were not simple computational tasks. The components were highly stateful and interconnected both on- and cross-host. Between this and the timescale, as well as the volume of events and the dollar-value impact of each potential failure (of which there were many), we needed real-time analysis capabilities, not a profiler.
Yep, I should probably have mentioned this option for completeness. In practice, however, it only saves you a few cycles/nanos and adds more complexity and failure points downstream.
It depends on your use case, as always. Correctness is not always black and white (hence my favourite compilation flag, -funsafe-math-optimizations) and time complexity can be misleading (O(log N) with a large base is O(1) in practice), but a correct, theoretically optimal algorithm might still be leaving a lot of performance on the table. If you're a hyperscaler, a high-frequency trader, or perhaps a game programmer pushing the limits of the platform, small gains can accumulate meaningfully, as can the thousand cuts caused by small inefficiencies spread out over the codebase.
>It all seems very brittle, though. And that something has gone very wrong with our ecosystem of tools, languages, and processes when it becomes advisable to massage source until specific passes in a specific version LLVM don't mess things up for other passes.
I would say the main takeaway is actually to not do that, precisely because it's brittle, difficult to understand, and can backfire by making things harder for the compiler. As I point out at the end, the vast majority of developers will be better served by sticking to common idioms and making their intent as clear as possible using language facilities, rather than trying to outsmart the compiler or, conversely, relying excessively on the optimiser as a magic black box.
That being said, I do find it helpful to understand the broad strokes of the optimisation pipeline (things like "callees are optimised before callers", "maintaining invariants enables more simplifications", or "early simplifications are better") to make the most of it. Like with any other tool, mastery means letting it do its job most of the time but knowing when and where to step in if necessary.
Full disclosure: I work for Veniam on connected vehicle technology, including LTE, Wi-Fi, and DSRC stacks.
Unfortunately there's also still no agreement on what 5G will actually be, what technologies it will entail, or how it will be deployed. Of course, since none of this is defined yet, there are also no guarantees on what performance will be achievable in practice.
The industry will need to solve this issue long before 5G is widespread. The solution will likely involve DSRC, C-V2X, or LTE-V/LTE Direct, which are tailored for the vehicular communication use case (involving both vehicle-to-vehicle [V2V] and vehicle-to-infrastructure [V2I]). It's unlikely a single technology will be able to handle all the challenges of a V2X scenario (highly mobile nodes, congestion, low SNR, requirements for sub-millisecond connection times), and so a second connection management layer will have to be built on top of multiple existing PHYs to enable simultaneous low-latency, high-throughput, and secure communications.
(Note that the linked article is actually about in-vehicle communication, however.)
We are looking to fill several openings across Engineering, Sales, and Management: https://veniam.com/jobs/. We have opportunities for cloud, analytics, front-end, hardware design, and embedded software work.
Veniam is building the Internet of Moving Things by turning vehicles into WiFi hotspots and building networks of connected vehicles that expand wireless coverage and collect terabytes of urban data for smart city applications.
As a member of our team at Veniam, you will have the opportunity to work in fast moving teams of 3-5 people, work closely with our Customer Success teams and often directly with key customers worldwide. We are organized in changing mission-driven teams that deliver continuously and you will have the opportunity to work on something new every few weeks.
This is the service supervision approach, which is indeed the best, avoiding race conditions and nasty polling. If you really must monitor processes which you cannot parent, at least under Linux you can do this using ptrace to sort-of-reparent the processes (as long as you have CAP_SYS_PTRACE or your ptrace scope is set to 0 in Yama), with waitpid to monitor all of them.
Someone with entrepreneurial spirit who is different, drives change and shares our passion for developing cutting-edge products that people and businesses love and use everyday.
Critical thinking, deliberate problem solving, disruptive mindset, and excellent communication are essential qualities, as you will be working in an exceptionally talented team with know-how across different disciplines, constant learning, boundless creativity, and outstanding execution – always in tandem with our headquarters in Silicon Valley.
Unless I misunderstood, that's what the TIMER_ABSTIME option is for. This is actually what I use for soft real-time loops, when I'm concerned about both drift and jitter.
[0]: https://shop.jcoglan.com/building-git/