Commit Graph

28 Commits

Author SHA1 Message Date
Geoff Langdale b824ba5d99 Issues #2 and #3 solved and the safety issues raised in my followup comments also resolved. 2018-05-31 13:26:24 +10:00
Geoff Langdale eece51c4fd Dammit switching off DEBUG mode 2018-05-31 12:46:28 +10:00
Geoff Langdale d820829e88 Switching to thicker tape as per https://github.com/lemire/simdjson/issues/4 2018-05-31 12:41:37 +10:00
Geoff Langdale 4ac426b61b Finished code for Unicode escaping; PoC only 2018-05-14 21:08:46 +10:00
Geoff Langdale e28af20593 Add a 'code sketch' for Unicode escapes; incomplete 2018-05-14 16:38:38 +10:00
Geoff Langdale 624b3d8742 Cleanups. 2018-05-14 10:17:00 +10:00
Geoff Langdale 88b0edd60f Partial completion on numbers and string normalization. 2018-05-11 14:45:54 +10:00
Geoff Langdale 750978bef9 Early (rough as guts) implementation of string normalization. 2018-05-08 16:28:54 +10:00
Geoff Langdale 77d8caa332 Allow a -DSQUASH_COUNTERS to suppress counters at build time (e.g. for using other tools like toplev) 2018-05-07 21:33:23 +10:00
Geoff Langdale 1df475a5ea Tiny opt changes to avoid 4K aliasing and loop bound recalc. 2018-05-07 20:45:51 +10:00
Geoff Langdale 8733686fc0 Rotation of our offsets for [] and {} on tapes, to make them navigable. Balancing []/{} moved to stage 4. Prefetch tweaks in stage 1. 2018-05-07 17:26:44 +10:00
Geoff Langdale 0fec86d105 Removed memcmp and replaced with masked compare. 2018-05-05 20:39:43 +10:00
Geoff Langdale 76ad40272e Removed close quotes from structural list; homogenized tape writes sizes (now always 4 or 0). Some tuning. 2018-05-05 14:44:31 +10:00
Geoff Langdale e78e7687d6 Adding a nominal stage 4. 2018-05-03 21:42:16 +10:00
Geoff Langdale 237d83885d Fixed display bug (cycles <-> instructions), tweak performance. 2018-05-01 19:05:31 +10:00
Geoff Langdale 232607278a Tapes now self-describing. Introduced temporary limit of 16MB (2^24) file size (fix later w. streaming) 2018-05-01 10:39:26 +10:00
Geoff Langdale 350aeecf3d Transformed tape machine. Now uses old tape not uptape and puts markers for open brace. Some streamlining. 2018-05-01 10:22:33 +10:00
Geoff Langdale dafcde53e5 Some cleanups and better diagnostic output including number of structural chars. 2018-05-01 09:51:12 +10:00
Daniel Lemire f82cf14c97 Adding stage speeds. 2018-04-27 13:31:28 -04:00
Daniel Lemire da7c811590 Outputting the number of cycles per input byte 2018-04-25 22:41:28 -04:00
Daniel Lemire 484a6c2799 Instrumenting the code to measure cycles and stuff. 2018-04-25 21:36:07 -04:00
Geoff Langdale 211892439e Switched to a simpler but more useful notion of pseudo-structural character; some small fixes. 2018-04-14 22:12:43 +10:00
Geoff Langdale 694942e3cd Merge branch 'master' of https://github.com/lemire/simdjson 2018-04-14 21:47:41 +10:00
Geoff Langdale 03799855df Fixup parsing stage with a new, dual state-machine/tape-machine approach. 2018-04-14 21:46:35 +10:00
Daniel Lemire 3b32b11fa6 Colorful display. 2018-04-06 10:53:51 -04:00
Geoff Langdale 020109b20c Add the high-performance but slightly unprincipled bitmask to index code. 2018-04-06 13:51:52 +10:00
Geoff Langdale e6478e33b3 Updating with recent progress. Adding UNUSED macro and suppressing warning from non-C++ flag. 2018-04-04 14:08:51 +10:00
Daniel Lemire bc1331283a Private research repo. 2018-03-23 00:05:32 -04:00