• d-bugmail puremagic.com (31/31) Dec 22 2017 https://issues.dlang.org/show_bug.cgi?id=18114

d-bugmail puremagic.com writes:

https://issues.dlang.org/show_bug.cgi?id=18114

          Issue ID: 18114
           Summary: dmd 2.078-beta1: regex performance regression
           Product: D
           Version: D2
          Hardware: x86
                OS: Mac OS X
            Status: NEW
          Severity: enhancement
          Priority: P1
         Component: phobos
          Assignee: nobody puremagic.com
          Reporter: jrdemail2000-dlang yahoo.com

My standard performance benchmarks show a regex regression moving from DMD
2.077.1 to DMD 2.078.0-beta1. The benchmark is the one described here:
https://github.com/eBay/tsv-utils-dlang/blob/master/docs/Performance.md#regular-expression-filter-benchmark.
No other performance regressions were seen, leading me to believe it is likely
specific to regex, rather than some other element of the program.

This program takes a user supplied regex as a command line argument. It creates
a Regex!char instance from the command line arg. Then it reads a file
line-by-line, searching each line for presence of the regex using matchFirst.
It's a fielded-search, so the matchFirst call is limited to a portion of the
line.

The benchmark times (MacBook Pro, OS X, 16GB ram, SSD drives):
- DMD 2.077.1:        11.40 seconds
- DMD 2.078.0-beta1:  15.10 seconds

Approximately a 30% increase. I ran it about 10 times, with very consistent
results.

If it would be helpful, I could produce a small standalone program to validate
the behavior.

--