Solaris dbx did have this capability -- patch and replace code with newly compiled code from your updated source files -- circa 2000. You just typed "fix" and it worked. I've been wishing for that capability on x86 Linux ever since.
821 karma · joined June 26, 2019
Solaris dbx did have this capability -- patch and replace code with newly compiled code from your updated source files -- circa 2000. You just typed "fix" and it worked. I've been wishing for that capability on x86 Linux ever since.
From a practical perspective, it interferes with getting your point across to overload a symbol that already has strong associations. From a social perspective, it's considerate not to use symbols that have strong negative associations for others (which is not a view shared by the "fuck your feelings" crowd, although I've noticed that they're not all that consistent in that view, and it tends to depend on whose feelings are getting fucked.)
I was interested in how the percentage accuracy metric is computed, but the link from the tutorial was dead. I was wondering if it could be useful to supply a bit more information. The example I was thinking of is linked from the math.js complex sqrt example, where the fastest version is rated more accurate than the original, because it avoids some catastrophic cancellation, but does not even use all the inputs to the problem. (Maybe that's not as bad as it sounds? I didn't get deep into the FP semantics but on the surface it sounded like something you'd be unhappy about if you blindly copy-and-pasted it into your product or library.) It seemed like % accuracy was hiding an important property that you might want to know about, which once you got into the details, you'd have an "aha, but no thanks" moment about. I'm not sure what metrics would make that clearer at an earlier point, but it's sort of an interesting problem.
Another question -- I see that you can restrict the ranges of the inputs to get improved answers. Another very common fact you might know about an expression is that one input is expected to be much larger than the other. Could Herbie take that as another assumption, and could it help? Maybe as a suggestion as to what ratios between inputs are most important when evaluating accuracy, or as a runtime test to break it into cases? Which it looks like you are already doing for ranges of specific variables (I guess the search space increases pretty fast when you start considering relationships between the inputs.)
Very cool work -- both esoteric and practical all at once!
Well, in the US you can have a benefit corporation, or else certification as a "B Corp", which I just learned are different things while googling it to put a link here. Previously my impression was that "B Corp" was a legal status, but that's wrong, it's a certification by a nonprofit. In the US, a benefit corporation has a separate legal status as "a type of for-profit corporate entity whose goals include making a positive impact on society."
Both are kind of a niche thing still. I've seen a few "B Corp" logos among sustainable food companies, like King Arthur flour.
https://en.m.wikipedia.org/wiki/B_Corporation_(certification...
The HN audience has folks from all aspects of hardware design, from architecture, to RTL coding, to physical design, or hobbyists programming FPGAs, etc., Saying "hardware was an entirely new world for me. I found it to be an unforgiving, harsh, complex terrain," just sounds wrong to them, like you're drawing conclusions about a huge field based on a narrow hobby project.
You seem to have set out to write an opinionated piece with a particular point of view, and said in your intro that you were going to be controversial, so you must have expected not everyone would agree with everything you said.
Maybe someday the LLM's will just read the spec and come up with new designs for which they have no examples!
https://www.nolo.com/legal-encyclopedia/is-federal-parole-sy...
https://www.justice.gov/usao-sdny/pr/united-states-attorney-...
"CAROLINE ELLISON, 28, is charged with and has pled guilty to two counts of conspiracy to commit wire fraud, each of which carry a maximum sentence of 20 years in prison; two counts of wire fraud, each of which carry a maximum sentence of 20 years in prison; one count of conspiracy to commit commodities fraud, which carries a maximum sentence of five years in prison; one count of conspiracy to commit securities fraud, which carries a maximum sentence of five years in prison; and one count of conspiracy to commit money laundering, which carries a maximum sentence of 20 years in prison. ...
"The statutory maximum sentences are prescribed by Congress and are provided here for informational purposes only, as any sentencing of the defendants will be determined by a judge. "
And while John Ray did say they were bad at record keeping, he also said that “This is just old fashioned embezzlement, taking money from others and using it for your own purposes,” he said. “This is not sophisticated at all.”
The subjects were asked about symptoms over a nine month period. The incidence of unexpected bleeding in the 4 weeks post-vaccination was 2X-3X higher than at other times in post-menopausal women, and 3X-5X higher for pre- or peri-menopausal women at times when they were not having a menstrual period. Spikevax was associated with 32% increased incidence compared to Comirnaty.
I've never heard of that happening, let alone many times to most institutions.
I haven't seen any news on since January, and no concrete shipping date. I signed up on the announcement email list and haven't seen anything there either (or at least I think I signed up; I don't even see an acknowledgement email.)
That's not really a criticism or doomsaying about Intel or SiFive from me -- I'm really interested to see what they have -- but they could do with a better communication strategy.
https://www.zdnet.com/article/experimental-intel-chip-could-...
Programmability is always the biggest issue, and that's not really a chicken-and-egg problem because decades of research have gone into writing compilers and languages for massively parallel machines -- it's just hard, some would say intractable (and local memories tend to make programmability issues worse.) There are niche or embarrasingly-parallel problems that will run great. But it's hard to sell hardware that will solve only some of your problems well. And GPUs have taken over for many of those very regular problems as well.