Crispin Cowan wrote: > Al Eridani wrote: >> On 11/9/06, Crispin Cowan <[EMAIL PROTECTED]> wrote: >> >>> Prior to Java, resorting to compiling to byte code (e.g. P-code back in >>> the Pascal days) was considered a lame kludge because the language >>> developers couldn't be bothered to write a real compiler. >>> >> "Post-Java, resorting to compiling to machine code is considered a lame >> kludge because the language developers cannot be bothered to write a >> real optimizer." >> > I don't see what a bytecode intermediate stage has to do with "real > optimizer". Very sophisticated optimizers have existed for native code > generators for a very long time. > > Bytecode interpreter performance blows goats, so I'm going to assume you > are referring to JIT. The first order effect of JIT is slow startup > time, but that's not an advantage either. So you must be claiming that > dynamic profiling (using runtime behavior to optimize code) is a major > advantage. It had better be, because the time constraints of doing your > optimization at JIT time restrict the amount of optimization you can do > vs. with a native code generator that gets to run off-line for as long > as it needs to. > > But yes, dynamic profiling can be an advantage. However, its use is not > restricted to bytecode systems. VMware, the Transmeta CPU, and DEC's > FX86 (virtual machine emulation to run x86 code on Alpha CPUs) use > dynamic translation to optimize performance. It works, in that those > systems all do gain performance from dynamic profiling, but note also > the reputation that they all have for speed: poor. > > And then there's "write once, run anywhere." Yeah ... right. I've run > Java applets, and Javascript applets, and the latter are vastly superior > for performance, and worse, all too often the Java applets are not "run > anywhere", they only run on very specific JVM implementations. > > There's the nice property that bytecode can be type safe. I really like > that. But the bytecode checker is slow; do people really run it > habitually? More important; is type safety a valuable property for > *untrusted code* that you are going to have to sandbox anyway? > > So I give up; what is it that's so great about bytecode? It looks a > *lot* like the Emperor is not wearing clothes to me. > > Crispin > Considering that on VMS, due to the use everywhere of a single calling standard and a linker that can understand things, you can link any language with any other language and get things to run. A single app with some pieces in Fortran, C, Pascal, BASIC, and assembly language, all in one program, is perfectly feasible. I have worked with such. Everything's compiled, everything compiles to optimised machine code, and there is no interpreter needed.
So why does anyone consider that multilingual development requires some kind of interpretive runtime? The old P-system was current a LONG time ago now, always did have speed problems (and took down some pretty decent apps with that ship in its day because they couldn't run as fast as native code ones). If there is some construct that NEEDS to be interpreted to gain something, it can be justified on that basis. Using interpretive runtimes just to link languages, or just to achieve portability when source code portability runs pretty well thanks, seems wasteful to me. Anybody know why .net uses a runtime of the sort it does? Glenn Everhart _______________________________________________ Secure Coding mailing list (SC-L) SC-L@securecoding.org List information, subscriptions, etc - http://krvw.com/mailman/listinfo/sc-l List charter available at - http://www.securecoding.org/list/charter.php
