https://bugs.exim.org/show_bug.cgi?id=2430
Bug ID: 2430
Summary: Severe performance decrease in (8-bit)
case-insensitive mode
Product: PCRE
Version: 10.32 (PCRE2)
Hardware: x86-64
OS: Linux
Status: NEW
Severity: bug
Priority: medium
Component: Code
Assignee: p...@hermes.cam.ac.uk
Reporter: andreas.bergm...@cyren.com
CC: pcre-dev@exim.org
I noticed a performance decrease in (8-bit) case-insensitive mode by a factor
of 100 when searching large buffers, which feels like a defect.
At least it should go into the documentation.
The pattern is "(?i)abc" vs. "aA][bB][cC]", the test data is 1 MB of random
text like "|w67i0sao-t032-tju-aeo-pk829a4vflh3\r\n".
------------------------------------------------------------------------
$ perf record ./TestPCRE
1048576 bytes 0.016975 sec (?i)abc
Samples: 66 of event 'cycles:ppp', Event count (approx.): 53979758
Overhead Command Shared Object Symbol
89.89% TestPCRE libc-2.24.so [.] memchr
5.59% TestPCRE [kernel.kallsyms] [k] clear_page_c_e
4.41% TestPCRE [kernel.kallsyms] [k] filemap_map_pages
0.10% perf_4.9 [kernel.kallsyms] [k] native_irq_return_iret
0.01% perf_4.9 [kernel.kallsyms] [k] native_write_msr
------------------------------------------------------------------------
$ perf record ./TestPCRE
1048576 bytes 0.000111 sec [aA][bB][cC]
Samples: 9 of event 'cycles:ppp', Event count (approx.): 6553514
Overhead Command Shared Object Symbol
41.23% TestPCRE [kernel.kallsyms] [k] filemap_map_pages
39.25% TestPCRE libc-2.24.so [.] __memset_avx2_erms
18.47% TestPCRE [kernel.kallsyms] [k] vma_compute_subtree_gap
1.00% perf_4.9 [kernel.kallsyms] [k] perf_event_addr_filters_exec
0.05% perf_4.9 [kernel.kallsyms] [k] native_write_msr
------------------------------------------------------------------------
--
You are receiving this mail because:
You are on the CC list for the bug.
--
## List details at https://lists.exim.org/mailman/listinfo/pcre-dev
