On 22.02.2013, at 16:11, Alexander Graf wrote: > > On 18.02.2013, at 10:15, Andreas Färber wrote: > >> Hi Alex, >> >> Here's a fixed and extended version, rebased onto ppc-next rebased onto >> master. >> Since my G2leGP3 fix is still missing from ppc-next despite supposedly >> applied, >> it is resent here as prerequisite. Series build-tested on PowerKVM now. >> >> v1 had shown a refactoring of CPU model definitions so that a class_init >> could >> be generated from POWERPC_DEF_SVR() macro. >> >> v2 builds on this and finishes refactoring CPU families into QOM subclasses, >> using a POWERPC_FAMILY() macro inspired by Alex to hide type registration. >> >> Follow-up ideas: >> * Can init_proc callbacks be turned into realize functions? (classes needed) >> * Introduce super-families? (BookE for TARGET_PPCEMB? MMU/Excp/Bus?) >> * Derive host CPU from model instead of copying? (to inherit instance_init) >> >> Available from: >> git://github.com/afaerber/qemu-cpu.git qom-cpu-ppc-types.v2 >> https://github.com/afaerber/qemu-cpu/commits/qom-cpu-ppc-types.v2 > > Thanks, applied all to ppc-next. Any comments I had can easily be addressed > in follow-up patches.
After this patch set, the -cpu ? list is split into 2 segments. I would prefer if we could in the help output revert to the old behavior of only a single list ordered by PVR. Let me show you what I'm thinking of: > PowerPC 601_v1 PVR 00010001 [...] > PowerPC 740_v3.1 PVR 00080301 740 (alias to "PowerPC 740_v3.1) > PowerPC 750cx_v1.0 PVR 00082100 [...] I also think the alias list should go into cpu-models.c. Alex
