On Wed, Mar 10, 2010 at 1:13 AM, Philipp Klaus Krause <[email protected]> wrote: > Do you mean that BitC shall no longer compile to C, but to something > else, such as Java bytecode instead?
Compiling to C was always a temporary measure. It was clear from the beginning that BitC needed to go straight to native instructions. The reason we went to C initially was that we wanted to preserve the ability to bootstrap the compiler in an environment that had a C compiler and then rebuild the compiler in BitC. > So far I assumed that BitC was intended for low-level embedded systems > development, where efficiency is essential, and sometimes the only > "high-level" language supported by tools is C. That is certainly where the work started, but there are a surprising number of applications that benefit from mandatory data layout, and the capabilities of current bytecode languages in this regard are not very strong. Also, few of those languages (notable exception: F#) make any real use of modern type systems. In general, I don't think that you can build a ground-up programming language today with only one application in mind. You need to have a clear set of initial applications, but if the language is *only* good for those then it is just too much work to build something like BitC. There is no reason *not* to be able to target CLI/JVM. Jonathan _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
