[comp.lang.pascal] Performance of TP on the MAC

mglacy@lamar.ColoState.EDU (04/26/91)

I just had my first experience converting a TP 4.0 application from an
IBM version to a MAC (TP 1.1) version.  To my surprise, the MAC version
appears to be substantially slower.  The program involves no graphics, a lot
of floating point arithmetic, a sort of a large array, and a large
number of calls to Random (actually RandomX on the MAC).  The MAC 
I've used (SE) appears to take something like 3 times as long as
my IBM XT to run the program.  Is this peculiar?  Anybody with
similar experiences?

Mike Lacy

bobb@vice.ICO.TEK.COM (Bob Beauchaine) (04/27/91)

In article <14533@ccncsu.ColoState.EDU> mglacy@lamar.ColoState.EDU () writes:
>The MAC 
>I've used (SE) appears to take something like 3 times as long as
>my IBM XT to run the program.  Is this peculiar?  Anybody with
>similar experiences?
>
>Mike Lacy

In college, I had to write a program to do character based mathematics
(i.e., no floating point data types allowed).  As an informal benchmark,
a friend and I calculated 10^10,000 using nearly identical routines for
exponentiation.  My machine (an 8Mhz AT) would solve the problem in 
about 30 seconds.  His, a Mac II something-or-other (I believe it had
a 68020) took 2 minutes.  I wasn't impressed.

/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\ 

Bob Beauchaine bobb@vice.ICO.TEK.COM 

C: The language that combines the power of assembly language with the 
   flexibility of assembly language.

christer@cs.umu.se (Christer Ericson) (04/28/91)

In article <14533@ccncsu.ColoState.EDU> mglacy@lamar.ColoState.EDU () writes:
>I just had my first experience converting a TP 4.0 application from an
>IBM version to a MAC (TP 1.1) version.  To my surprise, the MAC version
>appears to be substantially slower.  The program involves no graphics, a lot
>of floating point arithmetic, a sort of a large array, and a large
>number of calls to Random (actually RandomX on the MAC).  The MAC 
>I've used (SE) appears to take something like 3 times as long as
>my IBM XT to run the program.  Is this peculiar?  Anybody with
>similar experiences?

You shouldn't use TP on the Mac. It hasn't been updated since 1986(87?), nor is
Borland supporting it in any way. Use THINK Pascal (or MPW Pascal) instead.
TP didn't generate any good code either, it's main advantage over its
competitors was the fast linkage (but it didn't remove unused code though).

>Mike Lacy


| Christer Ericson                            Internet: christer@cs.umu.se |
| Department of Computer Science, University of Umea, S-90187 UMEA, Sweden |

drc@claris.com (Dennis Cohen) (04/28/91)

bobb@vice.ICO.TEK.COM (Bob Beauchaine) writes:

>In article <14533@ccncsu.ColoState.EDU> mglacy@lamar.ColoState.EDU () writes:
>>The MAC 
>>I've used (SE) appears to take something like 3 times as long as
>>my IBM XT to run the program.  Is this peculiar?  Anybody with
>>similar experiences?
>>
>>Mike Lacy

>In college, I had to write a program to do character based mathematics
>(i.e., no floating point data types allowed).  As an informal benchmark,
>a friend and I calculated 10^10,000 using nearly identical routines for
>exponentiation.  My machine (an 8Mhz AT) would solve the problem in 
>about 30 seconds.  His, a Mac II something-or-other (I believe it had
>a 68020) took 2 minutes.  I wasn't impressed.

This is a typical example of people confusing the quality/capabilities of
a particular compiler with the system on which it operates.  I have spent
a fair amount of time comparing performance of generated code from Pascal
and C compilers (among others) on a variety of platforms and the one thing
that can be said is that the range of results precludes there being any
really useful conclusions drawn.

In the case originally discussed, Turbo Pascal on the Mac (v1.1 was the last
offered by Borland) vs. Turbo Pascal (assumed v4.0 or later) on the PC, even
less can be said since they share almost no common code, their i/o and
memory management routines are vastly different, and without seeing the
routine in question one is hard-pressed to draw any useful conclusion.

As a counter-example (and almost equally meaningless), compare the performance
of FoxBase Plus/Mac vs the PC implementation.  Sorts, indexes, and all the
other "performance" benchmarks have a 16MHz 68020 as vastly faster than a
20MHz 80286.  This is equally meaningless.

I can take the same source and run it through MS-Pascal and TP6.0 and get
vastly different performance _on the same hardware_ and can rig the tests
with Turbo C vs MS C 5.0x to show that either is faster than the other.  I
can do the same thing on the Mac to "prove" than one compiler is better
than another.

It just ain't that simple, folks!  Anecdotal evidence of this sort just
isn't conducive to drawing intelligent conclusions.

-- 
| Dennis Cohen     drc@claris.com   COHEN2   AFC DCohen    71076,1377
|                  Internet       AppleLink  AmerOnline    CompuServe
| Disclaimer:  Any unattributed opinions expressed above are _MINE_!