Re: [PATCH 6/8] RFC: accel/tcg: only poll for interrupts in blocks that can close a cycle

On 8/18/26 10:42, Matt Turner wrote:
> RFC because:
> 
> - The soundness argument depends on every goto_tb destination passing
>    through translator_use_goto_tb(). No target in the tree bypasses it
>    today, but nothing enforces that.
> - System mode interrupt latency now depends on guest control flow rather
>    than on block count. The bound is one straight-line run between cycles,
>    which should be fine, but timer-driven guests deserve a closer look than
>    I can give them.

Yeah, I don't think this is going to wash.  There is plenty of guest code where 
straight-line code is massive.  You need to be able to guarantee that we perform the exit 
check regularly, without icount enabled.

The proper way to attack this is to build larger translation blocks with multiple guest 
basic blocks, which in turn requires more TCG infrastructure.  We'll still perform the 
exit check at the beginning of each TB, but the intent is that there would be fewer of 
them by some small-to-moderate multiple.


r~