Re: [Qemu-devel] [PATCH 2/7] target-alpha: Turn CPU definitions into subclasses

["Andreas Färber" <afaerber@suse.de>]

Am 31.10.2012 05:57, schrieb Richard Henderson:
> On 2012-10-31 14:03, Andreas Färber wrote:
>> +static const AlphaCPUInfo alpha_cpus[] = {
>> +    { .name = "ev4",     .initfn = ev4_cpu_initfn },
>> +    { .name = "ev5",     .initfn = ev5_cpu_initfn },
>> +    { .name = "ev56",    .initfn = ev56_cpu_initfn },
>> +    { .name = "pca56",   .initfn = pca56_cpu_initfn },
>> +    { .name = "ev6",     .initfn = ev6_cpu_initfn },
>> +    { .name = "ev67",    .initfn = ev67_cpu_initfn },
>> +    { .name = "ev68",    .initfn = ev68_cpu_initfn },
>> +    { .name = "21064",   .initfn = alpha_21064_cpu_initfn },
>> +    { .name = "21164",   .initfn = alpha_21164_cpu_initfn },
>> +    { .name = "21164a",  .initfn = alpha_21164a_cpu_initfn },
>> +    { .name = "21164pc", .initfn = alpha_21164pc_cpu_initfn },
>> +    { .name = "21264",   .initfn = alpha_21264_cpu_initfn },
>> +    { .name = "21264a",  .initfn = alpha_21264a_cpu_initfn },
>> +};
> 
> The "2*" names are aliases of the "ev*" names.  There's no need for so
> much duplication.  And for that matter, "ev68" is no different from "ev67"
> at the level for which we emulate.  In hw, it was more cache and a faster
> multiply implementation.

Clearly I know little to nothing about Alpha CPU models. :)
Regarding ev68, we'll need to carry it for backwards compatibility; can
we assume that the Alpha ISA is dead? Then I could drop this shrinking
array and make, e.g., ev68 a trivial subclass of ev67.

The name scheme we are heading towards now looks like <name>-alpha-cpu.
Did I understand you correctly that we would want, e.g., ev4-alpha-cpu
as type and have "21064" map to it? Or the other way around?

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg