The "MIT Climate Model" is not a full climate model, just a dynamical core (with simplified physics) for the atmosphere and ocean. It's never created for operational uses of predicting climate. Instead, it's configured for academic purposes of understanding some specific dynamical processes, in an idealized setting.Fortran dinosaurs crying, losing hoap
Reminder: Julia depends on LLVM, which is written in C++.
Also a substantial part of the Julia codebase is C and C++, just have a look at the main source repo, bottom right.
https://github.com/JuliaLang/julia
In effect all the speed critical part are C/C++.
You obviously haven't understood what a compiler is or how it works.
This just means the original Fortran code was inefficient and the inefficiency was somehow fixed when they rewrote the thing in Julia.Julia outperforms Fortran on some common code patterns.
https://julialang.org/benchmarks/
Fortran is indeed a poor language for tasks related to I/O and strings. Not sure if the climate model does a lot of that but it’s possible.
Also a substantial part of the Julia codebase is C and C++, just have a look at the main source repo, bottom right.
https://github.com/JuliaLang/julia
In effect all the speed critical part are C/C++.
OK bro! You go ahead and write all of the code for your papers in C. I'm sure you'll get all the code written and executed much faster.
Also a substantial part of the Julia codebase is C and C++, just have a look at the main source repo, bottom right.
https://github.com/JuliaLang/julia
In effect all the speed critical part are C/C++.You obviously haven't understood what a compiler is or how it works.
Can you explain it to me?
This just means the original Fortran code was inefficient and the inefficiency was somehow fixed when they rewrote the thing in Julia.If they had rewritten the code in Fortran but used modern syntax for parallel computing, it would likewise have fixed the problem.
Yes however the fact that you can make a program run faster by parallelizing etc doesn’t say much about the merits of Fortran vs Julia. After all you can do parallel computing using Julia as well.
Also a substantial part of the Julia codebase is C and C++, just have a look at the main source repo, bottom right.
https://github.com/JuliaLang/julia
In effect all the speed critical part are C/C++.You obviously haven't understood what a compiler is or how it works.
Can you explain it to me?
You are aware that Julia nor FORTRAN code actually run? Everything gets translated into machine code. The thing that is fast or slow is whatever the translation is, not Julia or FORTRAN
This just means the original Fortran code was inefficient and the inefficiency was somehow fixed when they rewrote the thing in Julia.If they had rewritten the code in Fortran but used modern syntax for parallel computing, it would likewise have fixed the problem.
Yes however the fact that you can make a program run faster by parallelizing etc doesn’t say much about the merits of Fortran vs Julia. After all you can do parallel computing using Julia as well.
The implication is that Julia was written in parallel and FORTRAN wasn’t
This just means the original Fortran code was inefficient and the inefficiency was somehow fixed when they rewrote the thing in Julia.
If they had rewritten the code in Fortran but used modern syntax for parallel computing, it would likewise have fixed the problem.
Yes however the fact that you can make a program run faster by parallelizing etc doesn’t say much about the merits of Fortran vs Julia. After all you can do parallel computing using Julia as well.
The implication is that Julia was written in parallel and FORTRAN wasn’t
If so that is some seriously lame performance comparison…
Also a substantial part of the Julia codebase is C and C++, just have a look at the main source repo, bottom right.
https://github.com/JuliaLang/julia
In effect all the speed critical part are C/C++.OK bro! You go ahead and write all of the code for your papers in C. I'm sure you'll get all the code written and executed much faster.
I don't write all the code in C, only some critical parts in some specific projects. For all the rest Python is more than fast enough given that most libraries are thin wrappers around native code.
There is not much point to Julia when Python gives you portability/readability/productivity for the vast majority of tasks. When it comes to high performance computing however there is no substitute for properly written C or Fortran. That a bunch of climate scientist got a speedup by porting their model to Fortran is irrelevant. It only implies they had poor, unoptimized Fortran code, and that beginner programmers are unlikely to write efficient native code. Stackoverflow is full of beginners stating that they ported a "slow" Python script to C only to find that C is slower.
The "MIT Climate Model" is not a full climate model, just a dynamical core (with simplified physics) for the atmosphere and ocean. It's never created for operational uses of predicting climate. Instead, it's configured for academic purposes of understanding some specific dynamical processes, in an idealized setting.Fortran dinosaurs crying, losing hoap
Reminder: Julia depends on LLVM, which is written in C++.
Julia --> C++ --> assembly language --> hardware
Fortran --> assembly language --> hardware
The "MIT Climate Model" is not a full climate model, just a dynamical core (with simplified physics) for the atmosphere and ocean. It's never created for operational uses of predicting climate. Instead, it's configured for academic purposes of understanding some specific dynamical processes, in an idealized setting.
Fortran dinosaurs crying, losing hoap
Reminder: Julia depends on LLVM, which is written in C++.
Julia --> C++ --> assembly language --> hardware
Fortran --> assembly language --> hardware
Have you heard of compilers? Optimization?