3 ms·
Wow, that's really cool. I feel like DiffEq is one of those libraries that seems to leverage every one of Julia's cool features that I can think of. Making dif
by eigenspace 7y ago
Wow, that's really cool. I feel like DiffEq is one of those libraries that seems to leverage every one of Julia's cool features that I can think of.
Making differential equations 'just work' on a GPU is super cool. I feel like if this were almost any other language, it'd be impossible to get this level of code-reuse where a user can just bring their own differential equation that may not even know about GPUs and then have the differential equation library solve it on the GPU by just giving a GPUarray as the initial condition.
________________________________
Also, since the blog post doesn't seem to link to it, here's a link to the actual DifferentialEquations.jl repo: https://github.com/JuliaDiffEq/DifferentialEquations.jl https://github.com/JuliaDiffEq/DifferentialEquations.jl
- ViralBShah 7y agoYeah, it is amazing to see how one can really rethink a core mathematical library with new tools and ideas. My mind continues to be blown every time I hear Chris talk about Measurements.jl and DifferentialEquations.jl just work, and you get sensitivity analysis for free (and neither package had planned for the other).
- The_rationalist 7y agowell openmp or openacc are still simpler.
- gugagore 7y agoDo you mean they are simpler to use? Because those are just models of parallel computation. A little like saying Turing machines are simpler than python.