On 01.03.2013 10:50, Henry Vermaak wrote:
On Fri, Mar 01, 2013 at 09:23:29AM +0000, Mark Morgan Lloyd wrote:
Sven Barth wrote:
An llvm target will move the optimisation burden away from fpc, which
would be very interesting.
While we would welcome a LLVM backend it is basically a consent in
the development team that this would only be an additional
alternative to the normal backends FPC provides.
LLVM's target list doesn't look particularly brilliant compared with
FPC's :-/
How do you mean? It supports more architectures than FPC, as far as I
can see (http://llvm.org/Features.html). They also have a C backend
that you can use for targets that they don't support.
Targets supported by FPC and LLVM:
i386
x86_64
PowerPC
PowerPC64
ARM (including Thumb)
Sparc
MIPS (in how far does LLVM support LE and BE and 32 and 64 bit variants
here?)
Targets supported by LLVM only:
Alpha
CellSPU
MSP430
SystemZ
XCore
Targets supported by FPC only:
AVR
(AVR32 [in a branch])
M68k
(JVM [in brackets, because JVM uses a high level code generator like
LLVM would do])
And you should also not forget that LLVM needs to be able to generate
code that's fitting for the target operating system. Think about the
more exotic ones supported by FPC: DOS, OS/2, GBA, NDS, etc. Also there
are only Windows binaries provided for Win32 and these are considered
experimental...
But even if LLVM would support all targets that FPC supports the core
developers don't *want* to make LLVM the default.
Regards,
Sven
_______________________________________________
fpc-pascal maillist - fpc-pascal@lists.freepascal.org
http://lists.freepascal.org/mailman/listinfo/fpc-pascal
