home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.msdos.programmer
- Path: sparky!uunet!usc!sdd.hp.com!zaphod.mps.ohio-state.edu!uwm.edu!linac!att!cbnewsk!cbnewsj!davet
- From: davet@cbnewsj.cb.att.com (Dave Tutelman)
- Subject: Re: COMPILING SPEED
- Organization: AT&T Bell Labs - Lincroft, NJ
- Date: Wed, 11 Nov 1992 12:55:14 GMT
- Message-ID: <1992Nov11.125514.1465@cbnewsj.cb.att.com>
- Summary: Pascal was slow, but C was slower
- References: <1dj8miINN6sl@matt.ksu.ksu.edu> <1992Nov10.160940.5958@exu.ericsson.se>
- Lines: 54
-
- In article <1992Nov10.160940.5958@exu.ericsson.se> exuhag@exu.ericsson.se writes:
- >Dan Odom writes:
- >>Yes, it is. Go write a compiler for Pascal. Then write a compiler for
- >>C++. You'll see.
- >
- >But until Turbo came along, MS-DOS Pascal compilers were total
- >dogs.
-
- Well, sort of. The competitors to Turbo in 1983 were compiled Pascals by
- Microsoft and IBM (the same product, I thought), and an interpreted Pascal
- by (if memory serves) P-systems. The ads in Byte for Turbo (I still have a
- copy somewhere) stressed that it:
- - Compiled much faster than IBM/Microsoft.
- - Ran much faster than P-systems.
- - Ran in less RAM / disk than either.
- - Was MUCH cheaper than either.
-
- One of those "too good to be true" ads, but I figured for $49 I could take a
- chance, and it WAS true. (Still have Turbo Pascal 1.0 floppy, Ser #1027.)
- Besides, I was talked into it by their EXCELLENT, free tech support.
- (Apropos another discussion in this newsgroup.)
-
- ON THE OTHER SIDE OF THE FENCE ......
- Did you ever try a C compiler of the same era? MUCH SLOWER than even IBM
- Pascal in compiling. Much worse.
- The biggest speed problem with IBM Pascal was not just compute speed,
- but juggling the 3-4 floppies you needed for
- a floppy-only machine -- which was most PCs back then. Most of Turbo's
- speed advantage came from doing the whole job in RAM.
-
- The C compilers of the day also required floppy-juggling. But you COULD
- take a coffee break between handling floppies.
-
- >What's troubling here is that people are blindly accepting
- >the "fact" that C++ compilers _have_ to be slow.
-
- Borland C++ runs REAL fast on my home machine; I have no complaints at all
- about speed. But that's because they haven't forgotten their roots. They
- use as much extended memory as they can eat, and I've found that compile
- speed is very sensitive to RAM (both extended and lower) available to
- BC++. They're STILL doing it by keeping as much as possible in core.
- Evidence:
- - A home machine memory upgrade gave a manifold increase in compile speed.
- I went from 1Meg to 4Meg, giving almost a 10X speed gain.
- - My office machine, with even more extended memory but with less available
- space in the lower 640K, compiles fast enough for a while (about 1200
- lines at a clip), but then stops and waits for a while... an annoying
- while. On a few occasions, such hiccups end with an "out of memory"
- fatal error. I'll unload a TSR or two, and recompile successfully,
- and much faster.
-
- Dave
- ---
- Not associated with Borland, but a usually-satisfied customer.
-