On 3/25/21 7:43 PM, Richard Henderson wrote: > On 3/25/21 5:59 AM, Claudio Fontana wrote: >>> Isn't tcg/tcg-* redundant? >> >> I considered that, and at some point I had "cpu.c" for x86 too. After >> working on it for a while, I noticed how >> it got really confusing in practice to have files called just "cpu.c" when >> working on them, just too many files are called "cpu.c". It was confusing. >> >> I also like the extra emphasis on the accel for this: >> >> kvm/kvm.c >> kvm/kvm-cpu.c >> kvm/kvm-stub.c >> >> tcg/tcg-cpu.c >> tcg/tcg-stub.c > > But then you go and invent > > cpu-sve.c > kvm/cpu-sve.c > tcg/cpu-sve.c > > etc. So, please make up your mind. >
Inconsistencies are all over already, and the two are very unrelated things. tcg-cpu comes from accel-cpu, where accel- is replaced by tcg- in this case. cpu-sve is a specialization of "cpu" where we currently put all the properties and the like. cpu-sve-props is probably a better name.
