Hi Dongli,
> int kvm_arch_init_vcpu(CPUState *cs)
> {
> struct {
> @@ -2237,6 +2247,13 @@ int kvm_arch_init_vcpu(CPUState *cs)
> cpuid_i = kvm_x86_build_cpuid(env, cpuid_data.entries, cpuid_i);
> cpuid_data.cpuid.nent = cpuid_i;
>
> + /*
> + * Initialize PMU information only once for the first vCPU.
> + */
> + if (cs == first_cpu) {
> + kvm_init_pmu_info(env);
> + }
> +
Thank you for the optimization. However, I think it’s not necessary
because:
* This is not a hot path and not a performance bottleneck.
* Many CPUID leaves are consistent across CPUs, and 0xA is just one of them.
* And encoding them all in kvm_x86_build_cpuid() is a common pattern.
Separating out 0xa disrupts code readability and fragments the CPUID encoding.
Therefore, code maintainability and correctness might be more important here,
than performance concern.
> if (((env->cpuid_version >> 8)&0xF) >= 6
> && (env->features[FEAT_1_EDX] & (CPUID_MCE | CPUID_MCA)) ==
> (CPUID_MCE | CPUID_MCA)) {
> --
> 2.39.3
>
