[bitc-dev] BitC 0.20: LLVM vs CLI
aleksi.nurmi at helsinki.fi
Sat Mar 13 03:34:54 PST 2010
2010/3/11 Jonathan S. Shapiro <shap at eros-os.org>:
>> >From the marketing point of view, CLI isn't really that good...
> You mean: it's not that good if we can't do the full CTS, or to you mean
> that CLI isn't that good in general?
I think I was mainly thinking about the full CTS thing, but there are
other reasons: I am probably wrong, but I remain concerned that due to
the abundance of safe CLI languages, there's no place for BitC. The
main advantage of BitC over F# is representation control, but I don't
see a lot of .NET developers _that_ interested in the good
old-fashioned constant factor performance. Abstractions for
concurrency and parallelism seem more important for success in that
environment. This is of course speculation.
Is part of the reason to begin with a CLI backend that, in order to
create a new type system compatible with CTS, you want to try it out
in practice? That would make sense even if the primary target is
eventually going to be LLVM and native code.
Mono's LLVM backend doesn't really help as it uses JIT compilation and
the executables require the Mono runtime anyway. General CLI -> native
compilation isn't really possible due to reflection.
Do you consider to separate the interop part of the type system?
More information about the bitc-dev