File: 503.txt

package info (click to toggle)
storm-lang 0.7.5-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid
  • size: 52,028 kB
  • sloc: ansic: 261,471; cpp: 140,432; sh: 14,891; perl: 9,846; python: 2,525; lisp: 2,504; asm: 860; makefile: 678; pascal: 70; java: 52; xml: 37; awk: 12
file content (39 lines) | stat: -rw-r--r-- 1,327 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

TEST_HEADER
 id = $Id$
 summary = Compare variety.hi/he and variety.wi performance
 language = english
END_HEADER

From variety.hi and variety/he:
|
| .desc: This variety is intended to have as much checking and
| telemetry as possible, but with no significant performance
| overhead as compared to variety.wi.  
|
| .slow: A "significant performance overhead" is a slow-down of
| more than 10% in any typical operation.
|
| .slow.just: The figure 10% was chosen because it is at a level
| that an interactive user would not observe. 
|

To test that varieties hi and he meet these requirements, compare
them with wi by running tests and measuring times. Tests to run:

- client application (Dylan, ScriptWorks &c)
- MMQA_test_function!{77,78,79,80}.c and others

It's not quite clear in the quote above whether "a slow-down of
more than 10%" means a slow down in memory-managament time, or
a slow down in total application time. I take it to refer to a
slow-down in memory-management time, given a typical pattern of
allocation, freeing, &c. Therefore if, say, Dylan goes 5% slower
with variety.hi than variety.wi, I will complain because the
MM takes less than 50% of Dylan's time, so the MM overhead must
be more than 10%.

Test is passed if varieties hi and he are within 10% of performance
of variety.wi.

rit 1998-03-13