Hi Willy, this is v2 of the series, addressing your two remarks on v1.
Changes since v1: - 1/4: after an escape, check whether the next byte is a '%' before calling memchr(), as you suggested. On a 5 kB path made only of escapes this saves 6.9 us per request (616.1 -> 609.2 us, -1.1%), on a 2 kB UTF-8 path 4.4 us (575.3 -> 571.0 us, -0.8%), which is about 2 to 4 ns per memchr() call avoided. An 8 byte path does not change. - 2/4: the '=' lookup in the collection loop uses iststop() instead of memchr(). v1 had only been measured with 6 character names; with 1-2 character names (467 parameters in 20 kB) the byte loop is 1.0% faster, with the 6 character ones 0.7%, consistent over five runs each. Small, but in the direction you expected, and the code is shorter. The numbers are in the commit message. - 3/4: unchanged. It splits on '#' with istsplit(), so the first remark does not apply there. - 4/4: only drops a duplicated Signed-off-by line. The output is unchanged. v1, v2 with only the first change, and v2 with both were run over 300k generated inputs (escapes, UTF-8 runs, short names, destination sizes on both sides of the truncation point, strict and non-strict): 3.59M comparisons of output and error code, no difference. reg-tests/http-rules/normalize_uri.vtc passes with all three, and with the series applied on current master (3e758361), where src/uri_normalizer.c has not changed since 3.2. The 32-bit note on 2/4 from v1 still stands (12 instead of 8 bytes per parameter, so the ERR_ALLOC bound drops by a third there); I can rework it if you prefer the old bound. Best regards Youngkwang Lee (4): OPTIM: uri_normalizer: bulk-copy runs between percent signs OPTIM: uri_normalizer: compute query parameter names only once OPTIM: uri_normalizer: bulk-copy the parts around encoded fragments REGTESTS: uri_normalizer: cover long URIs and many query parameters reg-tests/http-rules/normalize_uri.vtc | 38 +++++- src/uri_normalizer.c | 165 ++++++++++++++++--------- 2 files changed, 142 insertions(+), 61 deletions(-) base-commit: 3e758361c327bdf9e15ef3e2db99f1e9166a9763 -- 2.45.1
