From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=0.6 required=5.0 tests=BAYES_05,INVALID_DATE,
	MSGID_SHORT autolearn=no autolearn_force=no version=3.4.4
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!caip!lll-crg!seismo!nbires!hao!hplabs!pesnta!peora!ucf-cs!novavax!houligan!bseymour
From: bseymour@houligan.UUCP (Burch Seymour)
Newsgroups: net.lang.ada
Subject: Ada Compile Rates
Message-ID: <83@houligan.UUCP>
Date: Wed, 2-Jul-86 10:51:16 EDT
Article-I.D.: houligan.83
Posted: Wed Jul  2 10:51:16 1986
Date-Received: Fri, 4-Jul-86 08:38:15 EDT
Organization: Gould Electronics, Ft. Lauderdale, Florida.
List-Id: <comp.lang.ada>


>>>>> food for the line eaters of the world <<<<<

A recent posting was discussing the relative compile speeds of two Ada
compilers. While the relative speeds, given similar benchmarks, may be
of interest, let's not forget that the absolute line rate in the
absence of information about what was compiled is pretty useless. For
example (and assuming no fumbles with the calculator keys):

Compiling a 4 stmt procedure (proc is, begin, null, end) required 9.1
cpu seconds. This I will call the minimal overhead which includes
compiler loading, elaboration(this is a self hosted Ada compiler),
and set-up. This works out to 26 stmts/minute (cpu).

Compiling a 500 stmt program with all stmt's very easy to compile 
(i := 0;) took 20.1 cpu sec for a compile rate of 1493 stmts/cpu minute. 
If we adjust by the overhead time and consider the "at speed" rate
this works out to 2727 stmts/cpu minute.

Compiling a similar 500 stmt program, the difference being 10 of the
assignment statements were replaced by 10 with's, the cpu time went
to 35.1 sec's. This yields 854 stmts/cpu minute. Adjusting for
overhead yields 1154 stmts/cpu minute.

So, if I didn't like my compiler I could say it only compiles at 26
statements/cpu minute, and technically be accurate. If I liked my
compiler I could probably say it "runs" at over 2700 statements/ cpu
minute and also be correct. A difference of a factor of 100 depending
on how I measure the compile rate. Even with the two programs of
similar length, which is a more realistic comparision, the speed
was nearly a factor of 2 different.

The point of all of this is to be careful how you interpret compile
rates. We have agreed with our compiler vendor that all compile rates
will be measured using a pre-agreed upon set of (10 I think) ACVC
sources since these are written by a third party and are not designed
to either enhance or bust compile rates.


-- 
-------------------------------------------------------------------------
  "A nation that beats its swords into plowshares generally ends up
    doing the plowing for one that has kept it swords."  Anon

Burch Seymour -Gould C.S.D. at   ....mcnc!rti-sel!gould!bseymour
-------------------------------------------------------------------------