Commit message (Collapse) | Author | Age | Files | Lines | ||
---|---|---|---|---|---|---|
... | ||||||
* | bftw: Use a larger ioq depth | Tavian Barnes | 2023-07-18 | 1 | -22/+12 | |
| | | | | | | | Now that the dirlimit provides backpressure on the number of open directories, we can use a uniformly larger queue depth for increased performance. The current parameters were tuned with a small grid search on my workstation. | |||||
* | bftw: Add a queue of directories to unwrap | Tavian Barnes | 2023-07-18 | 1 | -7/+22 | |
| | | | | | | | | For !BFS_USE_UNWRAPDIR, if a file is still pinned in bftw_closedir(), it has to stay open until its pincount drops to zero. Since this happens in bftw_ioq_pop(), we can't immediately call bftw_unwrapdir() as that adds to the ioq. Instead, add it to a list that gets drained by the next bftw_gc(). | |||||
* | bftw: Add dirs to the end of the queue in bftw_ioq_pop() | Tavian Barnes | 2023-07-18 | 1 | -11/+25 | |
| | | | | | | I tried this before in #105 but it led to performance regressions. The key to avoiding those regressions is to put some backpressure on how many bfs_dir's can be allocated simultaneously. | |||||
* | bftw: Use separate queues for open and closed directories | Tavian Barnes | 2023-07-17 | 1 | -89/+115 | |
| | ||||||
* | bftw: Check that file->fd == bfs_dirfd(file->dir) earlier | Tavian Barnes | 2023-07-17 | 1 | -2/+3 | |
| | | | | | | | This has the potential to fail on at least one known platform: macports with the legacysupport implementation of fdopendir(). Link: https://github.com/macports/macports-ports/pull/19047#issuecomment-1636059809 | |||||
* | bftw: Reserve space in the cache before opening files | Tavian Barnes | 2023-07-17 | 1 | -3/+15 | |
| | | | | | | | This fixes a storm of EMFILE retries observed with -j1 on a very large directory tree. Fixes: 7888fbababd22190e9f919fc272957426a27969e | |||||
* | bftw: Pass the whole bftw_state to bftw_openat() | Tavian Barnes | 2023-07-17 | 1 | -510/+451 | |
| | | | | | This required shuffling a lot of code around. Hopefully the new order makes more sense. | |||||
* | bftw: Add bfs_dir allocation wrappers | Tavian Barnes | 2023-07-17 | 1 | -9/+19 | |
| | ||||||
* | bfstd: Add an ASCII fast path to wordesc() | Tavian Barnes | 2023-07-13 | 1 | -59/+100 | |
| | ||||||
* | alloc: Use a different error code for size overflows | Tavian Barnes | 2023-07-13 | 1 | -0/+10 | |
| | | | | This should help debuggability, and also squelches a GCC warning. | |||||
* | color: Only highlight the trailing slash on ENOTDIR | Tavian Barnes | 2023-07-13 | 1 | -2/+4 | |
| | ||||||
* | color: Don't break up leading and trailing dirs | Tavian Barnes | 2023-07-13 | 1 | -27/+32 | |
| | ||||||
* | color: TTY-escape filenames | Tavian Barnes | 2023-07-13 | 3 | -3/+12 | |
| | ||||||
* | color: Get rid of EXT_MAX | Tavian Barnes | 2023-07-13 | 1 | -10/+30 | |
| | ||||||
* | bfstd: Support wordesc() without allocating | Tavian Barnes | 2023-07-13 | 7 | -70/+175 | |
| | ||||||
* | bfstd: Quote the whole string the same way in wordesc() | Tavian Barnes | 2023-07-13 | 1 | -77/+101 | |
| | ||||||
* | bfstd: Use $'\n' etc. over $'\x0A' | Tavian Barnes | 2023-07-13 | 1 | -5/+35 | |
| | ||||||
* | bfstd: Escape '!' in wordesc() | Tavian Barnes | 2023-07-13 | 1 | -2/+2 | |
| | ||||||
* | parse: Reject -j0 | Tavian Barnes | 2023-07-12 | 1 | -1/+9 | |
| | ||||||
* | ioq: Try harder to avoid setting IOQ_BLOCKED | Tavian Barnes | 2023-07-12 | 1 | -6/+16 | |
| | ||||||
* | ioq: Separate slot and queue operations | Tavian Barnes | 2023-07-11 | 1 | -49/+75 | |
| | ||||||
* | eval: Don't oversubscribe the CPU by default | Tavian Barnes | 2023-07-11 | 1 | -3/+3 | |
| | ||||||
* | bftw: Try to close files asynchronously | Tavian Barnes | 2023-07-10 | 3 | -62/+163 | |
| | ||||||
* | ioq: Implement async close() and closedir() | Tavian Barnes | 2023-07-10 | 3 | -100/+197 | |
| | ||||||
* | bftw: If the ioq is full, try to pop before ioq_opendir() | Tavian Barnes | 2023-07-07 | 1 | -49/+82 | |
| | ||||||
* | ioq: New ioq_capacity() function | Tavian Barnes | 2023-07-06 | 2 | -2/+12 | |
| | ||||||
* | wordesc: Also escape non-printable chars | Tavian Barnes | 2023-07-06 | 2 | -30/+147 | |
| | ||||||
* | Use strcmp() instead of fnmatch() if possible | Tavian Barnes | 2023-07-06 | 3 | -9/+39 | |
| | ||||||
* | diag: Shell-escape expression arguments | Tavian Barnes | 2023-07-06 | 1 | -3/+3 | |
| | ||||||
* | ioq: Don't write to an empty slot in ioqq_trypop() | Tavian Barnes | 2023-07-04 | 1 | -6/+10 | |
| | ||||||
* | list: Fix some parameter docs | Tavian Barnes | 2023-07-04 | 1 | -4/+4 | |
| | ||||||
* | color: Don't print unnecessary reset sequences | Tavian Barnes | 2023-06-30 | 2 | -24/+45 | |
| | ||||||
* | docs: Start preparing for the 3.0 release | Tavian Barnes | 2023-06-30 | 1 | -1/+1 | |
| | ||||||
* | color: Compare values too when deciding to smart-case | Tavian Barnes | 2023-06-30 | 1 | -16/+65 | |
| | ||||||
* | color: Implement smart casing | Tavian Barnes | 2023-06-29 | 1 | -219/+372 | |
| | | | | | | | | | Since coreutils 9.2, ls does case-sensitive extension matching if the same extension is capitalized differently in $LS_COLORS. Implement the same logic. Link: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=33123 Link: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=9086 | |||||
* | parse: Warn about errors parsing $LS_COLORS | Tavian Barnes | 2023-06-29 | 1 | -1/+5 | |
| | ||||||
* | dstring: Add some exact-size utility functions | Tavian Barnes | 2023-06-29 | 2 | -21/+135 | |
| | ||||||
* | dstring: Allow dstreserve(NULL, n) | Tavian Barnes | 2023-06-29 | 1 | -1/+7 | |
| | ||||||
* | thread: Wrap more pthread APIs | Tavian Barnes | 2023-06-26 | 4 | -24/+38 | |
| | ||||||
* | ioq: New ioq_cancel() function | Tavian Barnes | 2023-06-26 | 3 | -7/+29 | |
| | ||||||
* | ioq: Don't check NDEBUG manually in ioqq_trypop() | Tavian Barnes | 2023-06-26 | 1 | -5/+2 | |
| | ||||||
* | bfstd: Add a getprogname() wrapper | Tavian Barnes | 2023-06-24 | 3 | -11/+24 | |
| | ||||||
* | Unify macro naming conventions | Tavian Barnes | 2023-06-24 | 8 | -51/+53 | |
| | | | | | In particular, macros that decide whether to use a particular API/ dependency should be spelled BFS_USE_*, and should be configurable. | |||||
* | diag: New bfs_loc type for source locations | Tavian Barnes | 2023-06-22 | 2 | -14/+48 | |
| | ||||||
* | dir: Arena-allocate directories | Tavian Barnes | 2023-06-20 | 6 | -51/+110 | |
| | ||||||
* | ioq: Arena-allocate ioq_cmd | Tavian Barnes | 2023-06-20 | 1 | -3/+8 | |
| | ||||||
* | trie: Arena-allocate nodes and leaves | Tavian Barnes | 2023-06-20 | 2 | -34/+36 | |
| | ||||||
* | bftw: Arena-allocate struct bftw_file | Tavian Barnes | 2023-06-20 | 1 | -5/+11 | |
| | ||||||
* | alloc: Implement an arena for flexible structs | Tavian Barnes | 2023-06-20 | 2 | -0/+210 | |
| | ||||||
* | alloc: Implement an arena allocator | Tavian Barnes | 2023-06-20 | 2 | -0/+155 | |
| |