-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi there i'm currently trying to set up a port for a simple RISC (actually it is *some* version of DLX; the binutils-2.1x one).
Since i'm inexperienced with GCC (messed around with SUIF/Machine-SUIF for some time) i have some questions. First of all, some basic information of the port follows: - - i'm using as a base an existing port. My first thought was MIPS but it is a rather complex port (since it has to support the entire MIPS family over several generations). The most DLX-friendly port that I have found is "nios2", which is not in the mainline but in a gnupro version of gcc sources for Altera Nios-2 (versions 5.0, 5.1 and 6.0 of the nios2-gnupro sources). - - at first, there will be no floating-point operations. - - if possible i would like to disable support for (long) and (long long) integer operands (DI and TI types that is). For example, i don't want to do anything for a "muldi3" define_insn. If it is ever encountered in given C application code, the compiler (cc1) should fail. - - as of today, xgcc and cc1 are produced, but the build process halts when gcc tries to find a way to properly deal with "muldi3". A few small application benchmarks that I use produce DLX assembly, but I have to really produce "elf32-dlx" binaries that I could on a simulator. My questions (for now): 1) as of current, xgcc and cc1 are produced, but the build process halts when gcc tries to find a way to properly deal with "muldi3". How can i disable integer long and long long arithmetic for the target? 2) How should I make "gcc" not to expect that the target will deal with floating-point operations? Not even with "SoftFloat" or included in the "libgcc" library? Thanks in advance Nikolaos Kavvadias -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (MingW32) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFFITXiMPiy0tCWlz4RApMaAKCWeSE1i018h5RaS5snsBqcotgHzACgnOxx zUZcTx6KqAGhGpQ2tivw97Y= =Hob8 -----END PGP SIGNATURE-----
