Feature #10326: [PATCH] optimize: recv << "literal string"

• Status: Rejected
• Priority: Normal
• Assignee: -
• Target version: 2.2.0

[normalperson (Eric Wong)]

Concatenating literal strings to existing strings seems fairly common
practice, so avoid extra garbage when doing it. I don't know if escape
analysis is coming to Ruby, but this seems easier-to-do for now:

vm2_strcat

i = 0
str = ""
while i<6_000_000 # benchmark loop 2
  i += 1
  str << "const"
  str.clear
end

trunk	1.020890216
trunk	1.018797116
trunk	0.992393447
built	0.626914306
built	0.627910138
built	0.644469095

---

benchmark results:
minimum results in each 3 measurements.
Execution time (sec)

name |trunk |built
--------+-------+------
loop_whileloop2 |0.160 |0.159
vm2_strcat* |0.833 |0.468

Speedup ratio: compare with the result of `trunk' (greater is better)

name |built
--------+-------
loop_whileloop2 |1.006
vm2_strcat* |1.779

Future optimizations:

• recv << "interpolated #{dstr}"
  (immediate free + force recycle)

• elide allocation for equality comparisons:
  str == "literal" and "literal" == str

• optimize << recv is IO << (str|dstr) (maybe)

[normalperson (Eric Wong)]

• Related to Feature #10329: [PATCH 2/1] optimize: foo == "literal string" added

[normalperson (Eric Wong)]

[e-mail] wrote:

Future optimizations:

• recv << "interpolated #{dstr}"
  (immediate free + force recycle)

Tried, not worth it from what I can tell:
http://80x24.org/spew/m/7c7dd29f45920fb8fc9d2c72858334c60e0a9c58.txt
(however, a good learning experience for me)

• elide allocation for equality comparisons:
  str == "literal" and "literal" == str

See [Feature #10329] and [Feature #10333]

• optimize << recv is IO << (str|dstr) (maybe)

I doubt it's worth the effort because of syscall overhead. This
(#10326) is fairly minor compared to the "==" optimizations in #10329
and #10333

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)

[normalperson (Eric Wong)]

• Status changed from Open to Rejected

obsoleted by Feature #10423