DLVM: A modern compiler framework for neural network DSLs

Current tally of high-performance, deep-learning-oriented DSLs&#x2F;IRs&#x2F;compilers, in no particular order:<p>- TensorComprehensions (Facebook): <a href="https:&#x2F;&#x2F;github.com&#x2F;facebookresearch&#x2F;TensorComprehensions" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;facebookresearch&#x2F;TensorComprehensions</a><p>- XLA (Google): <a href="https:&#x2F;&#x2F;www.tensorflow.org&#x2F;performance&#x2F;xla&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.tensorflow.org&#x2F;performance&#x2F;xla&#x2F;</a><p>- taco (MIT): <a href="http:&#x2F;&#x2F;tensor-compiler.org&#x2F;" rel="nofollow">http:&#x2F;&#x2F;tensor-compiler.org&#x2F;</a><p>- DLVM (UIUC): <a href="http:&#x2F;&#x2F;dlvm.org&#x2F;" rel="nofollow">http:&#x2F;&#x2F;dlvm.org&#x2F;</a><p>- nGraph (Intel): <a href="http:&#x2F;&#x2F;ngraph.nervanasys.com&#x2F;docs&#x2F;cpp&#x2F;" rel="nofollow">http:&#x2F;&#x2F;ngraph.nervanasys.com&#x2F;docs&#x2F;cpp&#x2F;</a><p>- TVM (DMLC): <a href="https:&#x2F;&#x2F;github.com&#x2F;dmlc&#x2F;tvm" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;dmlc&#x2F;tvm</a><p>Honorable mention to Julia (<a href="http:&#x2F;&#x2F;julialang.org" rel="nofollow">http:&#x2F;&#x2F;julialang.org</a>) as well.

Why are all the neural network DSLs JIT obsessed?

What does it mean by modern?