On 9/18/23 09:02, Philippe Mathieu-Daudé wrote:
When looking at the arm_cpu_realizefn() method, most of the
code run before the cpu_exec_realizefn() call checks whether
the requested CPU features are compatible with the requested
accelerator. Extract this code to a dedicated handler matching
our recently added CPUClass::verify_accel_features() handler.
How much verification is this intended to do?
Can realize now be required to succeed?
For instance, the error for unset gt_cntfrq_hz just before timer_new could be moved to
within this new function.
r~
