bob@imsvax.UUCP (Bob Burch) (02/22/88)
.....................................................................
Ted Holden
HTE
The Feb. 88 issue of the PC Tech Journal contains a fairly
detailed comparison of several of the top DOS C compilers on the
market today, as well as well as a couple of articles on the
state of C in the micro world today. Julie Anderson, writing in
the "Systems Perspective" column states that "the [MicroSoft v.5]
C optimizing compiler produces the best performing code of all
the compilers we reviewed". I would suggest that Ms. Anderson
either begin reading her own magazine or learn to count.
On page 63 of the article, I count 30 measures of speed for
compiled code, counting the items on either side of slashes as
seperate items, with Turbo C beating MC5 on 15 counts and losing
to MC5 on 14, one draw. The only item which shows a substantial
difference between the two was the unformatted read/write to
disk, which Turbo C lost rather badly, despite winning the
getc/putc comparison. Turbo beat MC5 in all but one measure of
arithmetic speed. TurboC was faster than QuickC in all but one
or two of the tests, often by considerable margins. TurboC
always generated smaller code than either QuickC or MC5.
The natural comparison for performance obviously is between
MC5 and Turbo; QuickC wasn't in the running. A funny thing
about the comparison was that the times given on page 63 for
Turbo were for code generated by the Turbo integrated system and
not by TCC. You'd have to assume code generated by TCC might be
a little faster. And, of course, compile times for Turbo are a
lot faster than for MC5. This translates into such real-world
measures of efficiency as projects done on time and money in the
user's pocket.
TurboC generally checked out as having more features than
MC5, one more memory model, inline code, graphics support (which
QuickC has but 5.0 apparently doesn't), and low-level keyboard
support. TurboC checked out as having slightly better
documentation than MC5 as well. As far as I could tell, reading
the entire article, Turbo C was the winner, pure and simple.
This is all the more remarkable considering the prices of the
compilers in the comparison.
The article pretty much neglected to mention one final point
which I would regard as rather critical. The present version of
Turbo is the second one. There were about ten or twelve basic
kinds of bugs which anyone ever found in Turbo 1.0, only one of
which would ever be seen by Joe Average User. You have to assume
that all such or at least ALMOST all such bugs are gone in Turbo
1.5. Compare this to the kinds of things you keep reading about
MC5 and quickC:
"MC5 destroys hard disk!!!, Woe is me!!!", "MC5 adds numbers
wrong!!", "QuickC involved in fatal conflict with Western-
Digital controller, user can't get refund!!", "MC5, QuickC
don't recognize Hercules Card!!", "MicroSoft compiler fries
FAT table!!!", "Programmer becomes sterile and loses hair
due to Microsoft C compiler!!!", etc. etc. etc.
Who needs it? The bottom line seems to be that Borland is one
whole generation ahead of MicroSoft in C compiler technology.
I say let Microsoft do things they're good at, such as OS's
and assemblers, and buy your high-level compilers from somebody
who's good at producing high-level compilers, such as Philippe
Kahn.
Ted Holden
HT Enterprises
romkey@kaos.UUCP (John Romkey) (02/23/88)
In article <788@imsvax.UUCP> bob@imsvax.UUCP (Bob Burch) writes: > TurboC generally checked out as having more features than > MC5, one more memory model, inline code, graphics support (which > QuickC has but 5.0 apparently doesn't), and low-level keyboard > support > > Ted Holden > HT Enterprises Well, MSC 5.0 does have the graphics support. I've used it with the 5.0 compiler (as opposed to QuickC which I didn't even bother to install) It just doesn't always work quite right.... -- - john romkey ...harvard!spdcc!kaos!romkey romkey@kaos.uucp romkey@xx.lcs.mit.edu
madd@bu-cs.BU.EDU (Jim Frost) (02/23/88)
In article <788@imsvax.UUCP> bob@imsvax.UUCP (Bob Burch) writes: > I say let Microsoft do things they're good at, such as OS's > and assemblers, and buy your high-level compilers from somebody > who's good at producing high-level compilers, such as Philippe > Kahn. This is really good advice (how many people saw MS Fortran?) but, seriously, how good are MS OS's? *Especially* when you have problems.... I think that when applications start coming out for OS/2, we may find that all of the things people have been complaining to uSoft about concerning their languages will show up concerning their new pet operating system. Only time will tell, but I know where I'm putting my money. jim frost madd@bu-it.bu.edu
aja@i.cc.purdue.edu (Miek Rowan) (02/23/88)
In article <788@imsvax.UUCP>, bob@imsvax.UUCP (Bob Burch) writes: > > TurboC generally checked out as having more features than > MC5, one more memory model, inline code, graphics support (which > QuickC has but 5.0 apparently doesn't), and low-level keyboard > support. TurboC checked out as having slightly better > documentation than MC5 as well. As far as I could tell, reading > the entire article, Turbo C was the winner, pure and simple. > This is all the more remarkable considering the prices of the > compilers in the comparison. One thing I have to give usoft credit for is, in my opinion,m ousatanding documentation in 5.0. I use both compilers, often, and have come to never open the tc manuals. They have had a while to get this right, but that doesnt change the fact that it is good docs. And pretty too. :-) Now if the tc integrated debugger turns out as good as codeview ...... miek