Sir Nullptr

WTF Codex! How many times I have to tell you that you don't need to clamp Z buffer reads between [0,1]. The hardware can't write outside that range!

morphology and morphospace for the CS bunny egirls who asked. what the heck is it? morphology is the study of form (the shape of the thing) rigorously. what does it look like? how can we make that a number? and morphospace is the space of all possible forms. you pick your measurable traits like limb ratios, curvature, angles, whatever is meaningful to your subject of study. then, you make axes out of them so every hypothetical form is now a point. now, you have a space! ... unfortunately MOST OF IT IS EMPTY why? well...because life occupies a tiny region. the rest is theoretical organisms that dont exist and (maybe) can’t. morphospace asks why THESE points dont exist? is it constraint? selection? developmental canalization? historical accident? vibes?? dear CS kitty, morphospace is similar to manifold learning style analysis. (but not identical keep that in mind ok..) when you run UMAP PCA, t-SNE on your embeddings you are doing something similar to morphospace analysis on a feature space. the questions are the same. where do things cluster? what regions are empty? what does distance mean? you have been doing this geometric morphometrics is one formal version of this. procrustes alignment + PCA, align your shapes (remove translation, rotation, scale, pure shape remains), find principal components of variation, plot you can model organisms that dont exist and ask why they dont? hmm? you can watch lineages move through the space. you can find convergence! (two clades arriving at the same region independently) which is one of the most interesting things in biology. sometimes people mix up waddington’s epigenetic landscape. it is related but it is mostly a metaphor for developmental trajectories. there’s also the phylotypic stage/developmental hourglass idea where vertebrate embryos have a similar form despite having very different early stages and very different adult forms. waooow how does a cow human rabbit pig chick all pass through a similarish body plan bottleneck? can early shape predict late outcome? how much of the space is accessible given our developmental architecture? that’s it. no language wars with the morphospace octopuses and CS bunnies please

Don’t blame Michael for the result. He has specified the terms very clearly in advance. Testing inside NVIDIA bunker with cherry-picked tests (very light-threaded) and pre-tuned system. NVIDIA being NVIDIA. But no one would like to miss out on such a great opportunity on new CPU

"However, one thing to keep in mind is that the marginal benefit of this scaling method gradually diminishes. The theoretical gain from folding a single layer into two layers is 100%. But moving from two layers to three layers only adds 50%. If Huawei moves from three to four stacked layers by 2035, the incremental gain would be only 33%. Moreover, as the number of stacked layers increases, the three challenges mentioned earlier — cooling, design complexity, and cost — all become more severe. (…) Leakage also remains unresolved. Cooling becomes increasingly difficult, and the 3D EDA toolchain represents an entirely new challenge. Therefore, Tau scaling is not an exponential growth path that can continue for the next ten years." I’m really grateful to be able to read this kind of expert opinion for free!

Doesn't look like a huge breakthrough. They're using a technique that's expected to be done by everyone - because they can't scale litho, they've accelerated other parts of the roadmap. Logic splitting is best shown in AMD's 2021 roadmap. It requires tight pitch TSVs and best in class HBI. So even if it comes out in 2026/2027, Huawei still has a way to go to hit 14A equivalents by 2031 using SMIC multipatterning. Also what is equivalent here? Chip perf? Logic density? Power? All of them? DTCO has been part of node and chip design for many years, and is often heralded as a 'node or better' uplift when done right. 2031 having 14A equivalent might be SMIC 7+++++ in reality. As in, each node has lots of improvements to make if you get stuck on it for a while.

A lot of people seem to think that Huawei has side-stepped EUV and developed a method to scale to 1.4nm without it But as u can see from their graph, a huge density & clock speed jump happens between 2030 & 2031 After the 2019 sanctions, Huawei started project Mount Everest in 2020 The goal was to develop and mass-produce EUV machines within 10 years The tentative date of China's EUV is 2030 rn

how do i get a badge on 𝕏?