Text preview for : 19771103_More_Mesa_3.0_Statistics.pdf part of xerox 19771103 More Mesa 3.0 Statistics xerox sdd memos_1977 19771103_More_Mesa_3.0_Statistics.pdf
Back to : 19771103_More_Mesa_3.0_St | Home
Inter-Office Memorandum
To John VVeaver Date November 3, 1977
From John Wick Location Palo Alto
Subject More Mesa 3.0 statistics Organization SOD/SO
XEROX SDD ARCHIVES
XEROX I have read and understood
Pages _ _ _ _To _ _
Filed on: [MAXC]MOREMESASTATS30.BRAVO
Reviewer
---------Date-------
# of Pages Ref .. 11.5>0 D. =3 ~ I
This memo contains the source-to-object expansion ratios for Mesa 3.0 and the previous
release. It also attempts to estimate some productivity figures. The old numbers are those
reported in W. Shultz's memo of June 6, 1977 (reflecting status as of April, 1977). The new
numbers are copied from Mesa 3.0 Statistics, October 17, 1977; most of these (except frame
sizes) are reproduced below.
Source Object Object per
Code Code Source Line
Project (lines) (bytes) (bytes)
Compiler (April) 23659 85782 3.63
Compiler (3.0) 26234 96466 3.68
Change + 2575 +10684 +0.05
Binder (April) 0 0 nfa *
Binder (3.0) 6218 24886 4.00
Change + 6218 +24886 nfa
System (April) 13089 30432 2.33
System (3.0) 17611 51164 2.91
Change + 4522 +20732 +0.58
Debugger (April) 6838 26490 3.87
Debugger (3.0) 7798 31354 4.02
Change + 960 + 4864 +0.15
Bootmesa (April) 4564 17958 3.93
Bootmesa (3.0) 2838 9998 3.52
Change - 1726 - 7960 -0.41
Utilities (April) 1481 5598 3.78
Utilities (3.0) 7173 31108 4.34
Change + 5692 +25510 +0.56
Totals (April) 49631 166260 3.35
Totals (3.0) 67872 244976 3.61
Change +18241 +78716 +0.26
More Mesa 3.0 Statistics 2
* The old binder is included in the April system figure.
Productivity in this case is hard to measure because we have no figures on the amount of
code deleted from the old system, and therefore no figures on the total amount of new code
added. The following calculations are based on the net change only (the utilities have not
been included; they are viewed as overhead):
time: 4/1 - 10/15 = 28 weeks
people: 6 - 1 (support) = 5
person weeks: 140
source lines: 18241 - 5692 = 12549
productivity: 89.64 loc/pw = 4660 loc/py
This figure includes all activity except support functions, which I have estimated to occupy
about one sixth of our time. Note that the staff size includes one full time person from
CSL (Ed Satterthwaite somewhat less than full time, with Jim Mitchell making up the
difference).
To get some handle on the total amount of new code written, we can try to estimate the
amount of code that was removed in one component (say, the system), and then apply this
factor to the other projects. An examination on a module basis indicates that about 900
lines were removed from the runtime system.
Old code size 13089
Deletes (estimate) - 895
Adds (computed) 5417
New code size
---
17611
This calculation indicates that we should add about 20% to the net change to get the total
amount of code written:
source lines: 12549 * 120% = 15059
productivity: 107.56 loc/pw = 5590 loc/py
Copy: Mesa Group