On 6 May 2015 at 19:24, Richard Henderson <r...@twiddle.net> wrote: > On 05/06/2015 08:44 AM, Peter Maydell wrote: >> This is making the assumption that what the disassembler sees as >> a "different mode" is the same as what the translator sees as a >> "different mode" for which it needs to end the TB... This happens >> to be true for ARM, at least, but I don't see any particular >> reason why it is necessarily so for everything. > > I'm pretty sure it is, though. Certainly true for x86 and mips. > I might call it a translator bug not to end the TB at a mode change.
Why? The only reason we end the TB at mode changes for ARM/Thumb is that the only way to do it in the architecture happens to require a jump, I think. I grant you it's probably a reasonable pragmatic requirement, though. -- PMM
