No need. I gave ChatGPT this prompt: "Write a data mover in Xilinx HLS with Vitis flow that takes in a stream of bytes, swaps pairs of bytes, then streams the bytes out"
And it did a good job. The code it made probably works fine and will run on most Xilinx FPGAs.
I don't even work in hardware, and yet even I have still heard of the Pentium FDIV bug, which happened despite people looking a lot more closely than "probably works fine".
The next operation might need the data in column major order to read it fast. So you might have to transpose first. And these maybe be concurrent stages of a processing pipeline.
Now I'm curious, how many times do you have to fully read the matrix in GPU for the total impact of reading columns to be higher than one-off actual transpose and then sequential row reads? I know it depends on lots of things, I'm after a rough estimate.
It's quite rare. Usually problems are tiled anyway and you can amortize the cost of having data in the "wrong" layout by loading coalesced in whatever is the best layout for your data and then transposing inside your tile, which gives you access to much faster memory.
The one pure transpose case that does come up occasionally is an in-place non-square transpose, where there is a rich literature of very fussy algorithms. If someone managed to make any headway with compiler optimization there, I'd be interested.
Each AIE tile can stream 64 Gbps in and out and perform 1024 bit SIMD operations. Each shares memory with its neighbors and the streams can be interconnected in various ways.
There is more to it than this. My children have had severe food allergies since birth: we know this b/c they had reactions to breast milk based on what mom would eat.
The also have had eczema since birth and one developed life threatening asthma at one year old. His trigger is the normal cold virus which he is constantly exposed to.
With that said, however, my kids are filthy all the time and play outside constantly in the dirt. One of the seven allergies has resolved completely and one appears to be resolved (haven't done a full challenge yet). No luck with the other five foods yet though.
We don't know about the asthma. Seems better but its managed with meds. Its also way scarier than anaphylaxis (and will be triggered by anaphylaxis and gets worse over the course of a week until he is in the hospital unless we aggressively treat it the moment he has breathing symptoms). The eczema comes and goes. Sometimes it responds to various treatments but usually it doesn't.
Its complicated. But kids should be filthy most of the time.
The hygiene theory also doesn't make 100% sense to me, I had no food allergies, no pollen allergies or asthma as a kid. Then in my late 20s, I got an egg allergy bad enough to require an epipen, as well as seasonal allergies, both out of nowhere.
I used to eat eggs several times a week, and now I have the type of egg allergy now where I can't even eat baked eggs, so no muffins, brioche bread, cookies, etc.
It's probably a bit of one, a bit of two. Despite spending everyday in the garden growing up, I've always struggled with allergies and they've changed throughout my life. I was deathly allergic to tomato's and eggs as a child but grew out of it around 11-12, became very allergic to pollen and pet dander at 5, which is now life long, and recently in my late 20s become allergic to hazelnut and some other foods I've yet to pin down (my mouth comes up in hives with no rhyme or reason while eating occasionally). I've also developed oesophagitus in my late 20s, to the point where I choke on foods daily unless chewed to a pulp before swallowing.
To me, it seems the parasite theory makes sense in tandem with hygiene. Some people by upringing become predisposed to allergies, while others have genetically as a result of humanities constant fight against intestinal worms. As our genetic profile changers as we age, so does how our bodies express said genes which would be designed to fight something we no longer have.
The immune system is a complex, only partly understood system, and there isn't a single unifying solution to all of its edges. Broad understandings don't necessarily translate to individual cases.
At some point your immune system faced an adversary and conquered it, but one of the signatures it learned from the enemy encounter unfortunately also matches some molecular component of eggs. There is evidence that some people come out of norovirus infection with an egg allergy, for instance. Similar to how a bite from a lone-star tick can give you a meat allergy.
All sorts of auto-immune diseases can be kicked off by relatively benign things, and often we might never discover the cause. Our immune system learned the heuristics of something, but it's too broad so ends up looking too much like our thyroid glands, nerve myelin, pancreas, etc.
Maybe one day we'll be able to enum all of the signatures an immune system has learned and delete some of them.
The hygiene hypothesis is mostly for childhood allergies and even then it wouldn't explain everything. Here is a counter anecdote: my sister has some skin allergies but I don't. I was allowed to play outside and used to just run around in my backyard and my neighbours. Those places can be pretty muddy. My sister was not allowed to and she has allergies.
Sound very stressful. I hope you and your family are doing well. I had problems with asthma in kindergarten and it is very scary for the first few times, after that I at least knew the drill and that I wouldn't die.
I interpret this as replacing a strange but more optimal call to the runtime with a straight forward but less fast stdlib implementation. rsc's comment seems more like a todo note than and ideological stance (to me at least). Something like: "I removed this weird fast way to do this that was breaking stuff b/c of linker things and replaced it with some straightforward code that's slower. If we want this to be more optimal in the future we can fix it at the compiler level.". But I'm just reading into it. Someone should ask him and tell him to explain himself :) It's been 8 years. What have you been doing?
And it did a good job. The code it made probably works fine and will run on most Xilinx FPGAs.