[vcpkg] Make RealFilesystem::remove_all much, much faster, and start benchmarking #7570
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Windows benchmark: |
|
WSL: |
|
macOS: |
ras0219-msft
reviewed
Aug 7, 2019
I added benchmarks to measure how fast the parallel remove_all code was
-- it turns out, about 3x slower than stdfs::remove_all. Since this was
the case, I removed all of the parallelism and rewrote it serially, and
ended up about 30% faster than stdfs::remove_all (in addition to
supporting symlinks).
In addition, I did the following three orthogonal changes:
- simplified the work queue, basing it on Billy O'Neal's idea
- Fix warnings on older versions of compilers in tests, by splitting
the pragmas out of pch.h.
- Ran clang-format on some files
In fixing up remove_all, the following changes were made:
- On Windows, regular symlinks and directory symlinks are distinct;
as an example, to remove directory symlinks (and junctions, for that
matter), one must use RemoveDirectory. Only on Windows, I added new
`file_type` and `file_status` types, with `file_type` including a new
`directory_symlink` enumerator, and `file_status` being exactly the
same as the old one except using the new `file_type`. On Unix, I
didn't make that change since they don't make a distinction.
- I added new `symlink_status` and `status` functions which use the
new `file_status` on Windows.
- I made `Filesystem::exists` call `fs::exists(status(p))`, as opposed
to the old version which called `stdfs::exists` directly.
- Added benchmarks to `vcpkg-test/files.cpp`. They test the
performance of `remove_all` on small directories (~20 files), with
symlinks and without, and on large directories (~2000 files), with
symlinks and without.
strega-nil
added a commit
to strega-nil/vcpkg
that referenced
this pull request
May 5, 2021
I added benchmarks to measure how fast the parallel remove_all code was
-- it turns out, about 3x slower than stdfs::remove_all. Since this was
the case, I removed all of the parallelism and rewrote it serially, and
ended up about 30% faster than stdfs::remove_all (in addition to
supporting symlinks).
In addition, I did the following three orthogonal changes:
- simplified the work queue, basing it on Billy O'Neal's idea
- Fix warnings on older versions of compilers in tests, by splitting
the pragmas out of pch.h.
- Ran clang-format on some files
In fixing up remove_all, the following changes were made:
- On Windows, regular symlinks and directory symlinks are distinct;
as an example, to remove directory symlinks (and junctions, for that
matter), one must use RemoveDirectory. Only on Windows, I added new
`file_type` and `file_status` types, with `file_type` including a new
`directory_symlink` enumerator, and `file_status` being exactly the
same as the old one except using the new `file_type`. On Unix, I
didn't make that change since they don't make a distinction.
- I added new `symlink_status` and `status` functions which use the
new `file_status` on Windows.
- I made `Filesystem::exists` call `fs::exists(status(p))`, as opposed
to the old version which called `stdfs::exists` directly.
- Added benchmarks to `vcpkg-test/files.cpp`. They test the
performance of `remove_all` on small directories (~20 files), with
symlinks and without, and on large directories (~2000 files), with
symlinks and without.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
It turns out that parallel file remove (#7228) is not good for performance -- generally getting a 3x slowdown. Since I added the catch2 testing framework in #7315, it's now easy to benchmark, and so I did. Results in a follow up comment.
Sorry for the broken code, I need to leave to feed my cat and will fix it at some point before EOD tomorrow.