NodeJs( V8 ) or C / C++ for heavy CPU operations ( > 100 Mlsec )?

Question

Node.js

Is it normal practice to write some hard-cpu piece of code with execution time, say, over 100 milliseconds for NodeJS in C/C++? Well, further integration with the node by creating a child process from the node / creating an addon.
Especially, for example, if the task is well parallelized, then one could even use CUDA.
As an example: suppose that for some end-point it is required to make a breadth-first traversal in a fairly large branched graph, the graph is parameterized by GET parameters, i.e. for each request, you will have to repeat the traversal (it is not possible to count 1 time in n seconds and cache the result somewhere).
So here's the question , so does anyone do it or are there any other approaches to solve similar problems?