On 9/29/26 06:29, Paolo Bonzini wrote:


Il mar 29 set 2026, 04:22 Andrey Polivoda <[email protected] <mailto:[email protected]>> ha scritto:

    Hi Paolo,

    In its current form, this patch makes VROUNDSS (and possibly other
    instructions) use 256-bit operand size when VEX.L=1


Thanks for the test case. I will take a look. The ordering of validate_vex() and decode_op_*() is tricky, I should fix it once and for all...

I take it I should discard this and expect another PR?


r~

Reply via email to