perf(parser): Use more performant operations for line parsing #198

Marwes · 2019-03-13T10:57:06Z

take_until_bytes uses memchr to find the next end of line + some
minor changes to use the fact that we know that line contains \r\n
at the end.

decode                  time:   [2.1868 us 2.2129 us 2.2494 us]
                        change: [-30.297% -29.538% -28.706%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  2 (2.00%) high mild
  1 (1.00%) high severe

`take_until_bytes` uses `memchr` to find the next end of line + some minor changes to use the fact that we know that `line` contains `\r\n` at the end. ``` decode time: [2.1868 us 2.2129 us 2.2494 us] change: [-30.297% -29.538% -28.706%] (p = 0.00 < 0.05) Performance has improved. Found 3 outliers among 100 measurements (3.00%) 2 (2.00%) high mild 1 (1.00%) high severe ```

badboy · 2019-03-13T14:37:30Z

First off: really appreciate all this perf work here! Awesome!

This is causing issues at the moment though:

thread 'main' panicked at 'Could not read enough bytes', tests/parser.rs:102:9
thread 'main' panicked at '[quickcheck] TEST FAILED (runtime error). Arguments: (ArbitraryValue(bulk(string-data('""'))), PartialWithErrors { items: [Limited(7)], _marker: PhantomData })
Error: "Could not read enough bytes"', /home/travis/.cargo/registry/src/github.com-1ecc6299db9ec823/quickcheck-0.6.2/src/tester.rs:179:28

Marwes · 2019-03-13T15:15:52Z

Thought I didn't need to add a partial parse test for take_until_bytes specifically since I just generalized an existing parser to implement it. Seems that one lacked a test or two though. Good thing it was caught here at least!

badboy · 2019-03-14T16:38:58Z

Cargo.toml

@@ -26,7 +26,7 @@ sha1 = ">= 0.2, < 0.7"
 url = "1.2"
 rustc-serialize = { version = "0.3.16", optional = true }
 unix_socket = { version = "0.5.0", optional = true }
-combine = "3.8.0"
+combine = "3.8.1"


Wait, what? checks combine commits

src/parser.rs

Markus Westerlind added 2 commits March 12, 2019 12:15

Add a decode benchmark

b1d97ae

Update combine for the take_until_bytes fix

b3d8f6f

badboy approved these changes Mar 14, 2019

View reviewed changes

badboy requested changes Mar 14, 2019

View reviewed changes

src/parser.rs Show resolved Hide resolved

badboy approved these changes Mar 14, 2019

View reviewed changes

badboy merged commit 859c453 into redis-rs:master Mar 14, 2019

Marwes deleted the bench_decode branch March 14, 2019 21:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

perf(parser): Use more performant operations for line parsing #198

perf(parser): Use more performant operations for line parsing #198

Marwes commented Mar 13, 2019

badboy commented Mar 13, 2019

Marwes commented Mar 13, 2019 •

edited

badboy Mar 14, 2019

perf(parser): Use more performant operations for line parsing #198

perf(parser): Use more performant operations for line parsing #198

Conversation

Marwes commented Mar 13, 2019

badboy commented Mar 13, 2019

Marwes commented Mar 13, 2019 • edited

badboy Mar 14, 2019

Choose a reason for hiding this comment

Marwes commented Mar 13, 2019 •

edited