To do what others still refused, I just give an answer:
Both, x86 and PPC PCs are more simmilar than different. So they're using rather the same architecture (PCI, ...). The different instruction sets are just two different devices named CPU. Therefore we use different compilers or call gcc with another architecture flag. The devices are the same on each plattform. PCI-cards are exchangeable and since PCI is PCI, the dev's adresses remain the same. So a well written driver will work on either architecture by just being recompiled. Normally. But there's still the Endian-problem (how order the multiple bytes of a large number have). x86 uses little endian, while ppc natively uses big endian. Up to G4 there's a switch to make the processor think different, but G5 misses this flag. So we seem to be forced to use big endian for our build.
What did M$ use in former times, when winnt 3.51 PPC was available?
Rick Langschultz wrote:
I was reading the email on free PowerPC computers available. I wanted to work on a port for powerpc a long time ago, but was shot down in the IRC channel. I want to use pearpc and a native powerpc with FreeBSD or linux in it to compile the project. Also since PowerPC and x86 and x64 have different instruction sets, how will ReactOS implement the NT kernel on non-intel architectures? I would imaging PPC-ata and X86-ata have different addresses. And not all adapters work from x86 on ppc. Maybe someone could work closely with Darwine to get some of this handled.
Ros-dev mailing list Ros-dev@reactos.com http://reactos.com:8080/mailman/listinfo/ros-dev