CAROZZONI@TOPS20.RADC.AF.MIL (11/20/89)
>Then perhaps you could explain the total dominance of C in the >mini/micro world, the only completely healthy segment of the computer >market? The last figures I've read indicated that 65 percent of all >software development in this world was being done in C, and the next >highest figure for any other language was around six percent. Are >Borland, MicroSoft, Lotus, Ashton-Tate, WordPerfect, and all of >those companies just that stupid? Anyone can pick a metric to justify their favorite software, but does it really solve anything? A COBOL advocate can justify using COBOL because there are more lines of it than anything else. A MS-DOS advocate can justify MS-DOS, can 35 million MS-DOS users be wrong? I'm not interested in using software just because my neighbor does. Also, a large (in Lotus 123 case almost 100%) portion of the s/w is written in assembler. Are they stupid for using assembler instead of C? Not in this case. I spend a lot of time (too much) writing s/w for MS-DOS machines. The biggest problem is the 640K limit (and 64K segments). I think you will find the super fast Borland and Microsoft products are written almost entirely in assembler. But neither you nor I would advocating the use of assembler. On the plus side for HOL's (C, Ada, etc), look at the classic Word Processing war between Microsoft and Word Perfect. MS Word wins all the benchmark's in terms of speed since it is written using MS MASM. Word Perfect is slower since it is written in C - but look who has ported their software to so maaaany different machines! If this were a race for market share, the fastest (MS Word) has actually lost. >IBM was and is committed to PL/1, the PC-Jr, OS/2, MCA... Apparently, >they have simply added Ada to their list of big-time losers with which >(temporarily) there is a profit to be made. > Put simply, "IBM should be committed". Just kidding. IBM chooses what they think will reap the greatest profits. OS/2 and MCA could have just as well as been UNIX and NuBus, etc. If IBM is indeed going to use Ada in a big way, it's because their accountants told them to. >The best compile times I've ever heard of either on a VAX class machine >or a 386 is around 1000/2000 lines/minute; Turbo C is around 15000 on a >fast 286, Zortech C++ about the same, Turbo Pascal around 40000 on a fast >286; God knows what they run at on a 386. We're talking about at least >a two-order-of-magnitude differential. The problem is that they are all MS-DOS hosted, what about us VMX/UNIX users. And as stated above, these products are written in assembler and forever doomed to 8088 architecture. Most of the Ada compilers are written in Ada - so what is actually being compared is an Ada compiler written in Ada VS a C compiler written in assembler. >Ada started in 79; taking 6 years to write a validated compiler doesn't >speak favorably of your efforts. World War II was fought in less time. The final Ada spec was 1983, the first compiler to pass validation was 1984/5. My primary reason for reading these various BB's are for information exchange. It's unfortunate that so many people get off on bashing a different point of view. I own a CBM Amiga and read the Amiga BB, but have to read through countless (and senseless) lines of IBM bashers. Years ago, the AI BB spent half their time putting C down, now they spend half their time on the transition from LISP to C. And on it goes. -- joe -------
harold@harvax.UUCP (Harold Rabbie) (11/21/89)
OK, let's get this issue out of the way regarding Ada tasking. Mr. Holden states the following: >> Another article in the same issue describes the use of Ada in connection >> with a real-time embedded digital signal processing application. Since >> Ada tasking could not be made fast enough for such work, the engineers >> adapted a commercial run-time system, VRTX, and informed all programmers: >> "Thou shalt not use Ada tasking, but rather VRTX tasking. >> "Thou shalt not use Ada dynamic memory allocation... >> "Thou shalt not use Ada generics; too slow... 'Tain't true that Ada tasking can't be made fast enough. How do you measure speed? You can buy a superset of VRTX called ARTX which implements Ada tasking, and, amazingly, it has the SAME context switch time for Ada tasks as VRTX tasks. So it's not Ada tasking that's broken, it's just some of the implementations. Also, Ada dynamic memory allocation - there are implementations that are slow, don't support UNCHECKED_DEALLOCATION, and raise STORAGE_ERROR just as soon as look at you. But then there are some that don't, and that give you fixed execution time for allocation and de-allocation. May the forces of free-market capitalism prevail! Ada generics too slow? Most implementations that I'm aware of just spit out multiple copies of the same code. Too large, maybe, but not too slow. I never understood this belief that generics somehow aren't safe Ada. After all, they're only glorified macros. Exceptions on the other hand ........ are just glorified GOTOs. --------------------------+-------------------------------------------- select | Harold Rabbie "Kernel Sanders" when REAL_TIME => | Ready Systems "We do real time right" accept READY_SYSTEMS; | PHONE: (408)522-7329 or | FAX: (408)736-3400 terminate; | UUNET: {sun!pyramid,hplabs}!harvax!harold end select; | ARPA: rabbieh@ajpo.sei.cmu.edu --------------------------+--------------------------------------------