I'm actually glad to see that restriction in their study, because there are many cases where you just can't / more difficult to use PGO. One of the most common reasons I ran into is that clients refuse to use it (no, I'm not even joking), either they don't know how to come up with a good training data (in which case they'll point at your nose shouting IT DOESN'T WORK!!) or they thought it's a stupid idea (again, not joking). You'll be amazed by some people's stubbornness
NASA used to be nice over SLS was that they really didn't have a choice + congress pressure (hiring people who lost their jobs due to space shuttle cancelation) but now it seems like NASA _does_ have a choice to choose an alternative (and waaay cheaper) vehicle over SLS. Curious whether those senators will keep their pressure.
I have the same feeling watching Starship when the telemetry on the bottom right of the screen showing “0km” but landing burn hasn’t started. Later did I know that it was meant to start the burn only a few hundreds _meters_ from the ground
I think saving weight is definitely one of the main issues, see those proportionally large legs on Falcon-9, I guess it simply doesn't scale on bigger vehicles like Starship / Heavy booster. Also, by catching the booster on site, they can even save the transportation and do the refurbishing on site, so even shorter turnaround time I guess.
I was just wrapping my head around on how do they even run linux on the RISC-V microcontroller they use. Turns out there is a separate ARM for application processor.
Switching from individual schedulers to unified one for integer execution makes sense to me, but I still don’t quite understand why FP execution units do the opposite, could somebody explain why?
LLVM tries to canonicalize all the loops into one of the two canonical forms[1] before doing any useful analysis and transformations. It has been quite effectives. Not all loops can be shaped into a canonical one though.
I'm not an expert in this field but I feel like the key is to make Dragon as a platform cheap, and once that happens more rides can be carried (to ISS) hence making the system battle-tested. Being cheap and reliable are the two major reasons NASA rode with Soyouz in the past few decades and, for better or for worse, still riding with them even after the Ukraine war started.
could somebody ELI5 what is happening to this program? delays after delays, and now they might need to send another spacecraft to bring people back. It will be hilarious if that turns out to be a SpaceX Dragon.
I feel like Microsoft's current stand on this is weird. They're either not as enthusiasting as Qualcomm and just riding along, or their terrible execution makes them move really slow.
> Not sure I understand that. Is it something specific to Swift, or is it exactly what is expected from using shared libraries?
More of a shared library issue I believe.
> Say my Linux distribution distributes some Swift runtime, then the corresponding Swift packages should have been built for this runtime as well. Just like when my Linux distribution distributes a libc, the corresponding packages need to be built for this libc. Right?
With my past experience in Khronos, NVIDIA is indeed a member and they sent decent guys to the meetings -- but only for strategic reasons, rather than "advocating for open standards" as you described. My experiences there actually told me the opposite that they will never drop CUDA. Objectively they also have incentive to do so: fighting with 100-ish companies to ratify something is always slower than rolling out an feature in an ecosystem you have total control of.
At first I thought it would just be a one-off tool I used for one of my projects, not until I discovered later that it has everything I need and became my daily driver ever since.
I know many tropical countries have successful stories fighting mosquitos by simply removing every single place that might contain water: ponds, flower pots etc.
Motorola 68k has this kind of aligned variable length instructions (in their case it’s always aligned on 16 bit boundaries). It’s not super difficult to support this in compilers though
not sure what you trying to get, but hundreds if not thousands of universities are working closely with DARPA, which is an DoD agency that sponsors mostly high-level and non-classified researches. Many of the researches have weak connections to actual military usages (there has to be _some_ connections, but many of them are pretty weak).