Looking at the data above and earlier findings (that running for text with both 600+ chars takes 1500ms+) I would go with threshold (switching tofastDiff()) when shorter string/array has above 400 elements and the length sum for both is above 1300 or if the length sum is above 2000 ...
So I did the closest thing; I found the CUDA.jl file in ~/.julia/packages/CUDA/M4jkK/src/, cd'd to it, and ran the commands that you specified: julia> using Artifacts julia> platform = Base.BinaryPlatforms.HostPlatform() Linux x86_64 {cxxstring_abi=cxx11, julia_version=1.7.0, li...
"name": "skatejs-dom-diff", "requires": true, "lockfileVersion": 1, "dependencies": { "@semantic-release/commit-analyzer": { "version": "2.0.0", "resolved": "https://registry.npmjs.org/@semantic-release/commit-analyzer/-/commit-analyzer-2.0.0.tgz", "integrity": "sha1-kk0eLDAWf...
"is-shared-array-buffer": "^1.0.2", "is-string": "^1.0.7", "is-weakref": "^1.0.2", "object-inspect": "^1.12.2", "object-keys": "^1.1.1", "object.assign": "^4.1.4", "regexp.prototype.flags": "^1.4.3", "safe-regex-test": "^1.0.0", "string.prototype.trimend":...