From: H. P. A. <hp...@zy...> - 2005-02-28 19:33:40
|
John Coffman wrote: > The move to 64-bit support is really not that hard. A well thought out > type definition file, with appropriate conditionals to define the obvious: > > u8, s8 bytes & chars > u16, s16 16-bit ints > u32, s32 32-bit ints > u64, s64 64-bit ints ; support varies for 'long long' > > Similar defs for floats. Some environments just won't be able to > support all types, unless a compatibility library is built. Ugh!! That's true for C89. For C99, 64-bit types are required and standard defines are in <inttypes.h>. > Are you aware that the DOS 16-bit compiler does not support any SSE > instructions? This was necessary to fit the instruction table into a > 64k segment of its own. Depends on which version of the compiler. > One approach I would suggest: for compilers that don't support 'long > long', define 's64' to be the same as 's32'. This will impose some > limitations on the resulting assembler, but "sizeof(u64)>sizeof(u32)" > conditionals can point out limitations if any are actually tickled. I suggest lets stop pretending to support ancient systems *as compile hosts*. It's just a big waste of time. -hpa |