A non-red @Ferrari? Its like a Mac without the 🍎
La Gazzetta Ferrari@GazzettaFerrari
Lewis and Charles revealed fully electric Ferrari Luce.
English
Federico Perini
782 posts
Federico Perini
@FedericoPerini
Combustion CFD consultant and researcher. Developing the FRESCO CFD code. Sempre fortissimo, e con strepito. 🇺🇸
Mantova / Madison Katılım Haziran 2011
420 Takip Edilen148 Takipçiler
@levelsio It's possible that the US AM / Europe PM working hours overlap corresponds to their compute load peak... our requests may just sit in a queue for longer
English
Much to be said - This is the Fortran equivalent of the C++ version
SudoX7@sudox7
your "clean" C++ is slower than this "ugly" C
English
OMG GPT 5.5 wrote C++ code with a `goto` in it.
What would Dijkstra say?
English
@taisukeOo Nice! what 3d framework does this use? three.js I guess? or vtk.js?
English
物理シミュレーター × トポロジー最適化 × 編集機能
（GPT5.5プロンプト数回で実現）
物理シミュ系のゲーム好きだけど自分で作った方が楽しい。
日本語
@chenna1985 No but for a Combustion CFD solver in Fortran…. Ask me anything!
English
Do any of the genius AI prompt engineers have the prompts for Claude or ChatGPT for writing an FEM solver in C++ for multiphase flows using the phase field approach?
English
Despite what I'm reading on here, my initial feedback is positive about 4.7. At times it feels a bit dumber than 4.6 (4.5 > 4.6 > 4.7 tbh) but it's quite a lot faster, spending far less time staring at the screen, this is going to be a lot more productive overall 🚀
Claude@claudeai
Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.
English
@litocoen Tell me you're cherry picking without telling me you're cherry picking
English
@IvanPribec Would be interesting to see how fast flang runs these tests - it’s a good benchmark
English
One more week and #Fortran is almost on par with C/C++, at least on my M1 Max.
Federico Perini@FedericoPerini
@lemire's work on fast_float parsing caught my attention, and seeing @kolemannix's great C port I decided to attempt a modern #Fortran version. All pure Fortran, -O3, nothing architecture-specific, it's on par if faster than C!
English
@SebAaltonen @BrockRiddickIFB Besides style - taste is personal - there's certainly a technical reason: (H)EVs don't need that much cooling, and fake fins advance flow separation, you don't want that for aerodynamic efficiency!
English
Federico Perini retweetledi
Following up on a tweet by @micheleboldrin
Nature index by university output.
Nr of top 50 in Europe by country (pop in parenthesis)
🇩🇪 18 (84m)
🇬🇧 9 (69m)
🇨🇭 6 (9m)
🇳🇱 6 (18m)
🇫🇷 4 (66m)
🇧🇪 2 (12m)
🇸🇪 2 (10m)
🇩🇰 2 (6m)
🇦🇹 1 (9m)
🇮🇹 0 (59m)
link in next tweet
English
@ChShersh Ah, interesting - so it's essentially an enum of types. In my use case focus is on propagating error handlers across many calls all over the CFD step.
I want a struct of compile-time known size, that gets fast default initialization, fast return-by-value and so no polymorphism
English
What exactly do you want to be resolved at compile time?
std::variant is a runtime structure; you don't know in advance which case it calls. So it has to call the corresponding function.
However, the types of all cases are known in advance, so the implementation immediately knows which to call each time; it doesn't traverse the list of functions at runtime.
English
And here's how you use it.
C++ is so elegant rn
Dmitrii Kovanikov@ChShersh
These 3 magic lines give you pattern-matching in C++
English
@SebAaltonen Are all SDFs for gaming from primitives or also from triangle meshes and more complex geometry structures? Like in a CFD code you have STL meshes and their sdfs are expensive even on GPU
English
Signed distance fields (SDF) are super nice for physics sim. Claybook had SDF based GPGPU physics engine. With SDF you always know where is the closest surface, even inside the objects. Pushing particles out is simple. Triangles = thin surface. Tunneling is a real issue.
Adiyar Aidarbekov - FXology@FX_ology
🤯DistanceFields is just an insane technology O_O ⛓️Sampling the DistanceField CollisionQuery to avoid particle clipping: 🔷if Distance to field is less than ParticleRadius - then it means that Particle is clipping through geometry => I correct it's position back to surface Since I'm using SimulationStage - it has same update rate as Constraint, so it doesn't break. #distancefield #niagara
English
"The practitioner shall implement their own..." "This is too much of a burden for compiler implementors".... I guess I should come to the conclusion that Fortran is faster-no-more. 🎪
English
And of course, a LOOP is always faster than an ARRAY🤦♂️... we'll at least soon get UNSIGNED, it's experimental on GCC.
English