HNHacker News
TopNewBestAskShowJobs

rbultje

401 karma · joined February 22, 2014

submissionscomments
rbultje··on Show HN: WarpBuild – x86-64 and arm GitHub Action runners for 30% faster builds
I've been using github actions cache to store build artifacts (object files) between builds. It takes a bit of fiddling but it's possible.
rbultje··on AV1 video codec gains broader hardware support
AV1 is more complex. Prediction can be more complex with things like "combined inter/intra" or "warped motion" or "overlapping block MC" compared to HEVC/VP9. Then there's additional postfilters like loop restoration, cdef and film grain that didn't exist in VP9 (just deblock - which also exists in AV1) or HEVC (deblock + sao). Entropy coding is more expensive than VP9 with per-symbol entropy updates (which HEVC also has). Bigger blocks are probably an overall win, but bigger transforms can be painful with many non-zero coefficients. And intermediates in 2D MC and bi-directional ("compound") prediction are 12-bit instead of 8-bit in VP9 for SDR/8bit video. This is more similar to HEVC. So overall, AV1 > HEVC > VP9 in terms of runtime complexity, this is expected, nothing you can do about it.
rbultje··on The Tyranny of the Marginal User
> When a profitable startup sells out, that's the founders also saying "profitable isn't enough for me".

Or just "I'm done with this shit". It's remarkably difficult to run a small (software) business in the US. Section 174, multi-state income tax filings. It's a bloody pain if your goal is not just to operate the business and be overhead, but to actually do stuff. Regulatory/tax complexity thresholds depending on business size would be a welcome improvement.

rbultje··on Ask HN: What is your policy regarding smartphones for your children?
I have 2 boys, 10 & 12. The 12YO has an Apple Watch which is useful to keep track of where he is and being able to reach each him (or him reach us). No social media or games, which is what most people are worried about at that age. So it seems like a decent trade-off. You can set it up to not allow phone calls or messages from unknown numbers, which means no scam calls/messages.
rbultje··on FCC imposes record penalty against transnational illegal robocalling operation
> What makes this a problem in the U.S., but not in Europe?

Setting up a business in Europe (for example in the Netherlands) requires significant paperwork. In the US, it's done in a few seconds. In the US, it requires years of court proceedings or (in this case) government agency investigation to shut down a business. (No idea how long that takes in Europe, but I hope your governments are more powerful than this.)

Regulation can be a good thing.

rbultje··on Software firms across US facing tax bills that threaten survival
> Isn't this all referring to the "R&D tax credit"?

No. The terminology is confusing. Section 174, specifically point 3, classifies software development as R&E as of 2022 and requires amortization over 5-15 years. This prevents software engineering salaries as a standard business tax deduction. This has nothing to do with the R&D tax credit (section 41).

rbultje··on Software firms across US facing tax bills that threaten survival
Amortizing makes sense for things that have a huge up-front cost and value over time, like - indeed - a robot, or a laptop, or a house. But salaries are not that, they are a continuous cost. I can take a loan on a house (mortgage), and laptops can be bought on 0% payment plans over 3 years. But how would that work for salaries? I'm forced to take out loans to pay the duplicate tax over salaries (and other minor expenses) over the amortization period. That's the insane bit here.
rbultje··on Software firms across US facing tax bills that threaten survival
Actually. For foreign expenses, the amortization period is 15 years instead of 5, so the cash flow problem is even worse. :-(.
rbultje··on The FFmpeg/Libav situation (2012)
It's like saying it would be good if there were two competing projects writing something akin to the Linux kernel. It's just not true. The mutually beneficial thing is to have one project where all developers are happy to work together. This is obviously idealized, but FFmpeg vs. Libav was 10x worse than what we have today. Imagine how Intel or Nvidia would contribute (either to the broken-up Linux or broken-up FFmpeg) new drivers or new hardware codec integration if there was competition. It would be a horrible end user experience.
rbultje··on Tell HN: Uber has blocked my account for years, won't tell me why
> I had multiple drivers in a row cancel on me before pickup

I had that exact same experience. Some of them even messaged me encouraging me to please cancel for them so that they would not incur a penalty. (I refused.)

I've been using yellow cabs in cities again. Simple & straightforward.

rbultje··on FTX held less than $1B in liquid assets against $9B in liabilities
Liquid doesn't just mean "can trade out of it at any time", but also "would typically be willing to trade out of it at any time". If a market is in a downswing, you're typically told to hold on to your stocks: buy low, sell high. The same rule would apply here. But that does mean they're not guaranteed to be liquid ("willing to trade") at all times, given that there's a risk of loss if you are forced to trade out of it at times of a market-low.
rbultje··on BCI lets completely “locked-in” man communicate
"On average, 121 min were spent spelling and the average length of these communications was 131 characters per day. The patient’s intelligible messages comprised 5747 characters produced over 5338 min, corresponding to an average rate of 1.08 characters per minute."
rbultje··on Dav1d, fast AV1 decoder, version 1.0
Icelake and tigerlake. On AWS, this is m6i/c6i. [edit] earlier AWS machines (c5/m5) have AVX512 also, but not the subsets required by dav1d's assembly.
rbultje··on America's 1% Has Taken $50T From the Bottom 90%
The official inflation numbers do not appear to reflect people's significant worries that rent prices or health care cost (etc.) keep going up.
rbultje··on Attention Must Be Paid: The Electoral Lessons of the Working Class
> How working class voters end up aligning themselves with a party that is less likely to care about them, while cutting taxes for earners well above the working class income bracket.

That's words vs. actions. His words were that he was going to take care of them, and his voters believed him. The fact that he didn't do it is just like any politician lying. They might not even have believed him either, but at least his voters felt like he listened to their pains.

rbultje··on Trump administration announces overhaul of H1B visa program requiring higher pay
Because educating underprivileged kids takes up to 20 years(school, college, potential post-grad studies), whereas the H1B worker can start immediately. H1B workers also typically have several years experience in the specific position where they're being considered. You need both.
rbultje··on Losing the war against surveillance capitalism letting Big Tech frame the debate
I fully agree with you. I did finish the book, but it took forever. Some of my friends did not finish the book, but instead read shorter versions by the author, for example: https://www.nytimes.com/2020/01/24/opinion/sunday/surveillan...
rbultje··on Ask HN: Dear open source devs how do you sustain yourself
This response reminds me of: https://twitter.com/ID_AA_Carmack/status/1258531455220609025

Opensource/free software powers most of the stuff we take for granted. The software and their maintainers are equally taken-for-granted.

rbultje··on Health insurers to investors: We’re good. Health insurers to lawmakers: Help
The affordable care act specifies that 80% of premiums must be spent on medical expenses. If not, the consumer gets a rebate at the end of the year.

See e.g. https://www.investopedia.com/terms/m/medical-cost-ratio.asp

rbultje··on Tell HN: C Experts Panel – Ask us anything about C
> for (i = 0; i <= N; ++i) { ... }

The worst thing is that people take it as acceptable that this loop is going to operate differently upon overflow (e.g. assume N is TYPE_MAX) depending on whether i or N are signed vs. unsigned.

rbultje··on Tell HN: C Experts Panel – Ask us anything about C
I/we understand the optimization, and I'm sure you understand the problem it brings to common procedures such as DSP routines that multiply signed coefficients from e.g. video or audio bitstreams:

for (int i = 0; i < 64; i++) result[i] = inputA[i] * inputB[i];

If inputA[i] * inputB[i] overflowed, why are my credit card details at risk? The question is: can we come up with an alternate behaviour that incorporates both advantages of the i<=N optimization, as well as leave my credit card details safe if the multiplication in the inner loop overflowed? Is there a middle road?

rbultje··on Tell HN: C Experts Panel – Ask us anything about C
I'd love your opinion on the abundance of "undefined behaviour" (as opposed to implementation-defined, or some new incantation such as "unknown result in variable but system is safe") for relatively trivial things such as signed (but not unsigned) integer overflows. I've heard that this is to allow for non-twos-complement implementations. However, in practice, you notice that most people use ugly workarounds which lead to ugly code that (because of e.g. casting to unsigned and allowing the same overflow to happen anyway) only work correctly on twos-complement anyway. Is this intended to be addressed in the future in some way?
rbultje··on Dav1d: performance and completion of the first release
Yes, it supports Windows natively. The tests were run on Windows.
rbultje··on Dav1d: performance and completion of the first release
10/12-bit can usually be done together, but are completely different from 8-bit. However, it's possible we'll do 10-bit first and then later on make the tiny adjustments that allow us to use them for both 10-bit as well as 12-bit.
rbultje··on Dav1d: performance and completion of the first release
dav1d is a decoder, not an encoder.
rbultje··on Dav1d: performance and completion of the first release
Right now, 10-bit decoding is horribly slow because the assembly optimizations only cover 8-bit, so it's probably 10-20x slower. We'll work on 10-bit next, and in the end, I'd expect it to be 30-50% slower than 8-bit.
rbultje··on Dav1d: performance and completion of the first release
No, it is written in C and assembly. See the gitlab graphs:

https://code.videolan.org/videolan/dav1d/graphs/master/chart...

rbultje··on Dav1d: performance and completion of the first release
Hi! You have 2 options: 1) write pthread emulation for your target system. We wrote one for windows native threads, but others should be straightforward. 2) if you want thread-less, that's possible (single-threaded performance shows 1080p is easy, and on high-end systems even 4K single-threaded might be doable), which basically just involves putting the two functions in thread_task.c under #if HAVE_THREADS, along with any coded calling pthread_() functions or using pthread_ types from <pthreads.h>, and then enforcing that Dav1dSettings.n_{tile,frame}_threads is always 1 (that means it won't ever enter these codepaths). Then, you always get single-threaded and (p)thread-less decoding.

Feel free to come on IRC, happy to help you dive into this, it's not very difficult.

rbultje··on A million-dollar drug
As sad as this story is, the silver lining is that patents expire, and 20 years from the patent filing point (probably 5-10 years from now?), patients will be able to be affordably cured. It sucks but is better than nothing.
rbultje··on Introducing Dav1d: a new AV1 decoder
Bitrate reduction for the same quality. So a file of identical quality - in some metric - will be 20% smaller (e.g. 8 MB instead of 10 MB for a video).
← PreviousPage 2 of 3Next →