On 7/10/25 05:43, Philippe Mathieu-Daudé wrote:
On 8/9/25 14:26, Mads Ynddal wrote:
On 3 Sep 2025, at 12.06, Philippe Mathieu-Daudé <[email protected]>
wrote:
Signed-off-by: Philippe Mathieu-Daudé <[email protected]>
---
target/arm/hvf/hvf.c | 40 +++++++++++++++++++++-------------------
1 file changed, 21 insertions(+), 19 deletions(-)
It'd be nice with a commit message justifying the need for the change. I
assume performance?
Performance is a super positive side effect. Actually I had issues in
the outer vCPU loop because it is only aware of EXCP_* return values,
and the code handling unknown '0' retval is dubious. IOW I'd rather have
the outer loop assert for unknown values, as a programming error.
(Also Richard later noted flush_cpu_state() should be called out of the
inner loop).
