Closed WebFreak001 closed 4 years ago
note: these are also on a much more powerful machine, but I believe most speed reduction between OSs comes from fewer system wide import paths
Old (Debug) | New (Debug) | Old (Release) | New (Release) | |
---|---|---|---|---|
10525 | 3306 | 9673 | 1516 | |
10518 | 3268 | 9839 | 1525 | |
10468 | 3261 | 9636 | 1552 | |
average | 10504 | 3278 | 9716 | 1531 |
relative speed | 100% | 320% | 100% | 635% |
Old (Debug) | New (Debug) | Old (Release) | New (Release) | |
---|---|---|---|---|
84 | 20 | 93 | 7 | |
116 | 23 | 91 | 11 | |
85 | 23 | 91 | 11 | |
average | 95 | 22 | 92 | 10 |
relative speed | 100% | 432% | 100% | 948% |
LDC 1.22.0 Target: x86_64-pc-linux-gnu Host CPU: znver1 lots of available RAM fast NVMe SSD
I would make a patch release to be compatible with the other newly released dsymbol for practical reasons now:
import foo.bar
with matching import paths containing ["a/foo/bar.d", "b/foo/bar.d", "c/foo/bar.d"]
would returnc/foo/bar.d
a/foo/bar.d
resolveImportLocation
function.d
matches inside import paths now immediately return on the first import path instead of returning at the end of the function being equal to the path inside the last matching import pathThis gives a big speed up improvement in DCD on Windows:
Duration of first completion (ms)
Duration of later completions (ms)
LDC 1.19.0 Target: x86_64-pc-windows-msvc Host CPU: ivybridge very little available RAM (<500 MB) some SSD