Age | Commit message (Collapse) | Author | |
---|---|---|---|
2021-09-08 | current_process_memory_usage: Add system memory stats. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-08-21 | map_handle cache now looks to be working well. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-08-21 | Reimplemented current_process_memory_usage() for Linux so it's a lot faster ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
for very large programs. Debugged the map handle cache somewhat, but it's still failing. Tomorrow! | |||
2021-08-04 | Try once again on passing CI on Mac. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-08-03 | Try fix last commit on Mac OS. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-08-03 | Add utils::current_process_cpu_usage(). | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-05-25 | Fix CI failures in last commit. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-05-25 | Fix bug where native Linux dynamic_thread_pool_group would schedule delayed ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
work items inverted to correctness. | |||
2021-05-11 | Fix preceding commit. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-04-15 | Merge https://github.com/ned14/llfio/pull/77 into my current working tree ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
and push the bits which are safe to push for now. | |||
2021-03-19 | Reenable the llfio::dynamic_thread_pool_group::io_aware_work_item unit test. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-03-16 | Fix failure to execute with priority on Linux native dynamic_thread_pool_group. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-03-16 | more wip | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-03-16 | More wip | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-03-16 | Make the dynamic thread pool work scheduling not quite as strict in the ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
native Linux implementation, now we only examine higher priority groups once per group cycle. This produces behaviours much more consistent with the Win32 thread pool and Grand Central Dispatch. | |||
2021-03-16 | wip native linux threadpool implementation for dynamic_thread_pool_group. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-03-16 | Implemented and debugged a Grand Unified Dispatch backend for ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
dynamic_thread_pool_group. Works surprisingly nicely on Linux, haven't actually tested it on Mac OS nor FreeBSD, but no reason it shouldn't work just fine. | |||
2021-03-16 | Implement dynamic_thread_pool_group::io_aware_work_item. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2021-03-16 | Add unit test to ensure nested dynamic thread pool groups work as expected. ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Still Windows only implementation. | |||
2021-03-16 | Add statfs_t::f_iosinprogress and statfs_t::f_ioswaittime which | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
can be used to monitor how congested the underlying hardware device for a handle is. | |||
2021-03-15 | Fix inability to open directory junctions on Windows using `symlink_handle` ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
(issue #73). | |||
2021-03-06 | Fix lots of bugs due to newer kernel 5.8 in Ubuntu 20.04, and OpenZFS v0.8: | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
1. clone_extents_to() didn't handle partial clones, and now it does. 2. OpenZFS has a bug in SEEK_DATA which caused empty clones, we use a hacky workaround to fix it. 3. mapped_file_handle now fails loudly if it fails to allocate the map due to VMA exhaustion. This closes off a ton of unpleasant surprise in heavily loaded applications. 4. If your Linux kernel was old enough, or you are on ZFS, renaming a mapped_file_handle caused internal state corruption which led to lots of bad things, including `bad_file_descriptor` errors. 5. On Windows, renaming a mapped_file_handle now tears down and restores the maps, as you cannot rename a file with open maps. | |||
2021-01-29 | Port from AFIO v1 `normalise_path()`, whose API I have improved and made ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
more directed as `to_win32_path()`. | |||
2020-11-24 | Disable one of the MSVC Experimental Filesystem emulations, because it is ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
literally insanity on a plate. | |||
2020-11-24 | Add path view special casing for VS2017 Experimental Filesystem. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-11-20 | path_view tests now pass on at least MSVC and libstdc++. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-11-20 | Fix unit test failures, and regress Windows CI testing to VS2017 as VS2019 ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
latest ICEs. | |||
2020-11-19 | Fix without_trailing_separator() removing root. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-11-19 | Add path_view::without_trailing_separator(). | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-11-19 | Also fix path_view::remove_filename(), which was not returning a trailing slash. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-11-19 | Fix bug in path_view::parent_path() where "/x.txt" was not returning "/" as ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
the parent path. | |||
2020-11-12 | Rework how path_view::c_str's constructors are implemented to work around ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
this weird bug only seen on clang with libstdc++ where the c_str() constructor never delegates. Now no c_str() constructor delegates to any other, which surely will fix this bug. | |||
2020-11-11 | More Windows install fixes, disable issue0009 test on 32 bit, yet more debug ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
printing on clang inside path_discovery. | |||
2020-11-10 | again | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-11-10 | again | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-10-30 | Implement the latest round of changes to path_view as per WG21 LEWG guidance. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-10-13 | Don't commit stuff when sick and at 10.30pm at night! Fix previous commit. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-10-13 | Extend issue0009 timeout on Mac OS, it looks like it's a touch slow to tear ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
down a 2^40 sized file map. | |||
2020-10-12 | Fix issue0009 which had been refactored to match Windows, and does not work ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
on POSIX. | |||
2020-10-12 | P1030 refactored path_view appears to be mostly working. Tests all pass on ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Windows (after fixing the long broken pipe_handle coroutines test). On GCC7 on Linux, quite a few tests appear to have broken, we shall see what the CI says. | |||
2020-10-09 | More wip path_view refactor to meet latest P1030R4, now working on MSVC, not ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
yet working on POSIX. | |||
2020-10-08 | Long, long round of refactoring path_view to match latest P1030 draft ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
normative wording. Not done yet however, so this is a wip commit. | |||
2020-09-07 | Fix the poor performance of the previous commits fixing the race condition ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
in tear down of mapped files on Windows. | |||
2020-09-04 | Fix a subtle race condition in map_handle on Windows where if many threads ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
are currently allocating and freeing maps, it could rarely occur that one thread would free the map of another thread. The newly rewritten v3 of win32_maps_apply() we really, really hope is the last time! | |||
2020-07-13 | On POSIX, non-atomic renames were losing path tracking. Also, renameat2() ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
was never being used on Linux. Both fixed. Finally, stat_t now compares each member item individually, instead of using memcmp(). | |||
2020-07-03 | Fix bug in reduce() on POSIX which was causing random failure to reduce. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-07-03 | Yet more travis and appveyor bashing | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-07-03 | More travis bashing, more appveyor debug printing | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-07-03 | Add more appveyor debug printing | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2020-07-03 | Fix perhaps the cause of appveyor failing. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |