2019-07-10 14:35:10 -07:00
|
|
|
#pragma once
|
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
#include <vcpkg/base/checks.h>
|
|
|
|
|
2019-07-10 15:42:13 -07:00
|
|
|
#include <condition_variable>
|
2019-07-10 14:35:10 -07:00
|
|
|
#include <memory>
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
#include <vector>
|
2019-07-10 14:35:10 -07:00
|
|
|
|
2019-07-11 18:16:10 -07:00
|
|
|
namespace vcpkg
|
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
template<class Action>
|
2019-07-11 18:16:10 -07:00
|
|
|
struct WorkQueue
|
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
WorkQueue(LineInfo li) : m_line_info(li) {}
|
|
|
|
WorkQueue(const WorkQueue&) = delete;
|
2019-07-10 14:35:10 -07:00
|
|
|
|
2019-07-11 18:16:10 -07:00
|
|
|
~WorkQueue()
|
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
auto lck = std::unique_lock<std::mutex>(m_mutex, std::try_to_lock);
|
|
|
|
/*
|
|
|
|
if we don't own the lock, there isn't much we can do
|
|
|
|
it is likely a spurious failure
|
|
|
|
*/
|
|
|
|
if (lck && m_running_workers != 0)
|
2019-07-11 18:16:10 -07:00
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
Checks::exit_with_message(
|
|
|
|
m_line_info, "Internal error -- outstanding workers (%u) at destruct point", m_running_workers);
|
2019-07-11 15:01:29 -07:00
|
|
|
}
|
|
|
|
}
|
2019-07-10 14:35:10 -07:00
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
template<class F>
|
|
|
|
void run_and_join(unsigned num_threads, const F& tld_init) noexcept
|
2019-07-11 18:16:10 -07:00
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
if (m_actions.empty()) return;
|
2019-07-11 18:16:10 -07:00
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
std::vector<std::thread> threads;
|
|
|
|
threads.reserve(num_threads);
|
|
|
|
for (unsigned i = 0; i < num_threads; ++i)
|
2019-07-11 18:16:10 -07:00
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
threads.emplace_back(Worker<decltype(tld_init())>{this, tld_init()});
|
2019-07-11 18:16:10 -07:00
|
|
|
}
|
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
for (auto& thrd : threads)
|
2019-07-11 18:16:10 -07:00
|
|
|
{
|
2019-07-10 14:35:10 -07:00
|
|
|
thrd.join();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// useful in the case of errors
|
|
|
|
// doesn't stop any existing running tasks
|
|
|
|
// returns immediately, so that one can call this in a task
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
void cancel() const
|
2019-07-11 18:16:10 -07:00
|
|
|
{
|
2019-07-10 14:35:10 -07:00
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
auto lck = std::lock_guard<std::mutex>(m_mutex);
|
|
|
|
m_cancelled = true;
|
|
|
|
m_actions.clear();
|
2019-07-10 14:35:10 -07:00
|
|
|
}
|
|
|
|
m_cv.notify_all();
|
|
|
|
}
|
|
|
|
|
2019-07-11 18:16:10 -07:00
|
|
|
void enqueue_action(Action a) const
|
|
|
|
{
|
2019-07-10 14:35:10 -07:00
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
auto lck = std::lock_guard<std::mutex>(m_mutex);
|
|
|
|
if (m_cancelled) return;
|
2019-07-11 15:01:29 -07:00
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
m_actions.push_back(std::move(a));
|
2019-07-10 14:35:10 -07:00
|
|
|
}
|
|
|
|
m_cv.notify_one();
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
template<class ThreadLocalData>
|
2019-07-11 18:16:10 -07:00
|
|
|
struct Worker
|
|
|
|
{
|
2019-07-10 14:35:10 -07:00
|
|
|
const WorkQueue* work_queue;
|
|
|
|
ThreadLocalData tld;
|
|
|
|
|
2019-07-11 18:16:10 -07:00
|
|
|
void operator()()
|
|
|
|
{
|
2019-07-10 15:42:13 -07:00
|
|
|
auto lck = std::unique_lock<std::mutex>(work_queue->m_mutex);
|
2019-07-11 18:16:10 -07:00
|
|
|
for (;;)
|
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
const auto& w = *work_queue;
|
|
|
|
work_queue->m_cv.wait(lck, [&w] {
|
|
|
|
if (w.m_cancelled)
|
|
|
|
return true;
|
|
|
|
else if (!w.m_actions.empty())
|
|
|
|
return true;
|
|
|
|
else if (w.m_running_workers == 0)
|
|
|
|
return true;
|
|
|
|
else
|
|
|
|
return false;
|
|
|
|
});
|
|
|
|
|
|
|
|
if (work_queue->m_cancelled || work_queue->m_actions.empty())
|
2019-07-11 18:16:10 -07:00
|
|
|
{
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
/*
|
|
|
|
if we've been cancelled, or if the work queue is empty
|
|
|
|
and there are no other workers, we want to return
|
|
|
|
immediately; we don't check for the latter condition
|
|
|
|
since if we're at this point, then either the queue
|
|
|
|
is not empty, or there are no other workers, or both.
|
|
|
|
We can't have an empty queue, and other workers, or
|
|
|
|
we would still be in the wait.
|
|
|
|
*/
|
2019-07-11 18:16:10 -07:00
|
|
|
break;
|
2019-07-10 14:35:10 -07:00
|
|
|
}
|
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
++work_queue->m_running_workers;
|
2019-07-10 14:35:10 -07:00
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
auto action = std::move(work_queue->m_actions.back());
|
2019-07-10 15:42:13 -07:00
|
|
|
work_queue->m_actions.pop_back();
|
2019-07-10 14:35:10 -07:00
|
|
|
|
2019-07-10 15:42:13 -07:00
|
|
|
lck.unlock();
|
2019-07-15 18:51:03 -07:00
|
|
|
work_queue->m_cv.notify_one();
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
std::move(action)(tld, *work_queue);
|
2019-07-10 15:42:13 -07:00
|
|
|
lck.lock();
|
2019-07-11 18:16:10 -07:00
|
|
|
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
const auto after = --work_queue->m_running_workers;
|
|
|
|
if (work_queue->m_actions.empty() && after == 0)
|
|
|
|
{
|
|
|
|
work_queue->m_cv.notify_all();
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
2019-07-10 14:35:10 -07:00
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2019-07-11 18:16:10 -07:00
|
|
|
mutable std::mutex m_mutex{};
|
|
|
|
// these are all under m_mutex
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
mutable bool m_cancelled = false;
|
2019-07-11 18:16:10 -07:00
|
|
|
mutable std::vector<Action> m_actions{};
|
|
|
|
mutable std::condition_variable m_cv{};
|
[vcpkg] Make Filesystem::remove_all faster #7570
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.
2019-08-02 16:49:45 -07:00
|
|
|
mutable unsigned long m_running_workers = 0;
|
2019-07-10 14:35:10 -07:00
|
|
|
|
2019-07-11 15:01:29 -07:00
|
|
|
LineInfo m_line_info;
|
2019-07-10 14:35:10 -07:00
|
|
|
};
|
|
|
|
}
|