Age | Commit message (Collapse) | Author | |
---|---|---|---|
2017-09-08 | Made a start on mapped_file_handle | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-09-05 | Added handle flags for disabling or maximising prefetching. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-09-05 | Implemented mmap based small file append. Insertion is now equally fast as ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
fetching on Windows, suggesting that fetching is overwhelmingly slow now. | |||
2017-09-04 | AFIO now creates sparse files by default on Windows. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-09-04 | Added integrity benchmarks for toy key value store for NTFS | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-09-04 | Enabled integrity checking in toy key value store, was fairly amazed to see ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
it worked first time | |||
2017-09-04 | Append-only mode didn't actually work on POSIX, fixed :) | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Toy key-value store now working on Linux. Was quite surprised at the causes of failure, all platform-specific quirks. | |||
2017-09-04 | Hopefully restore working build on Travis | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-09-04 | Fix missing header on Linux, should compile now. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-09-04 | Added io_handle::max_buffers(). | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Toy key value store now implements transaction::commit() using gathered i/o. Toy key value store now compiles on POSIX. | |||
2017-09-03 | Implemented key deletion into the toy key-value store. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Actually handle errors properly now, correctly rolling back and abandoning any transaction. | |||
2017-09-01 | On Windows we could not open byte range locks on an append-only file, fixed. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Toy key-value store actually commits and fetches now. | |||
2017-08-31 | Add the exploratory key-value store. Nobody should use it! | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-27 | Update storage profile database with latest results. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-27 | Fix segfaults on CI | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-24 | Fix bug where we couldn't work with files in the root of a Windows drive. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Test the OS scheduler in storage_profile to see how long it takes to do stuff. | |||
2017-08-21 | Flush all modified data and drop filesystem caches before benchmarking ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
storage in storage_profile. | |||
2017-08-21 | Added algorithm::mapped_view, and barrier_on_close for mapped sections. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-18 | Added results for desktop machine. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-18 | New latency results with the fixed tool | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-17 | Update benchmarks with results from actually working tests. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-17 | Eliminate the 2017 results as having been unreliable. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-17 | Looks like the direct i/o storage_profile tests never actually worked :( | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-17 | Fix incorrect printing of test failure. The former code was a logic error :( | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-17 | Reduce need for temporary files if not running those tests. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-17 | Get the programs compiling on POSIX. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Add fs-probe results for my other Windows machine. | |||
2017-08-17 | Fix programs build when in non-sibling config, and have Travis test those ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
are building from now on. | |||
2017-08-17 | Change how quickccplib dependencies are found. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-11 | Fixed the miscalculation of QD16 latencies in storage_profile. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-11 | Added fs_handle::parent_path_handle() | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
unlink() and relink() simply weren't working on Windows. Fixed. We now request privs to rename and delete when opening a handle with the ability to modify. Fixed bug where path_view was returning zero lengthed c_str on Windows sometimes. Added clock overhead to storage_profile. fs_probe now takes bitfield as args. | |||
2017-08-10 | Added i/o latency benchmarking to storage_profile | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-08-10 | Upgraded programs/* to use Outcome v2. | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
Fixed quite a few bugs which had crept in: - The \\.\ prefix handling code was just wrong. Now is \!!\ - We no longer copy paths if null terminating on Windows if they are going to the NT kernel as it doesn't use null termination. - Failure to fetch inode after file open is no longer a failure. | |||
2017-07-21 | We now use a proper custom NTSTATUS error code category on Windows instead ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
of converting to Win32 codes. Made a start on replacing all usage of make_*() with Outcome v2. | |||
2017-07-14 | First attempt at a port to QuickCppLib and Outcome v2 (it's not working yet) | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-05-03 | Relicenced under Apache 2.0 + Boost licence | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2017-04-28 | Use cmake namespaced target aliases for indicating dependencies between ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
libraries | |||
2016-09-04 | Found a big problem where win_delete_on_last_close was causing opens of such ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
files to no longer be openable after the first close of a handle so marked. Did a substantial refactor of how unlinking and unlinking on close works. Totally untested code, that'll be tomorrow :) | |||
2016-08-27 | Refactored code to handle bitfield and undoer now living in Boost-lite, and ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
child_process now lives in KernelTest. | |||
2016-08-27 | Much to my surprise, algorithm::shared_fs_mutex::memory_map almost nearly ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
worked first time. Performance, as expected, is a full order of magnitude faster than byte_ranges. CppCon will be pleased :) | |||
2016-08-26 | Fix failure to correctly pickup sibling boost-lite. Add cmake for AFIO ↵ | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
programs which are now compiling. Added in the new memory map shared_fs_mutex to benchmark-locking. | |||
2016-08-26 | fs_probe => fs-probe | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2016-08-26 | Cleaned out the old fs probe vcxproj files | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |
2016-08-26 | Move fs_probe into programs | Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com) | |