On Tue, 18 Aug 2026 at 10:29, Peter Maydell <[email protected]> wrote: > > On Mon, 17 Aug 2026 at 23:34, Richard Henderson > <[email protected]> wrote: > > > > On 8/17/26 09:37, Peter Maydell wrote: > > >> + tcg_gen_addi_i64(diff, diff, -1); > > >> + tcg_gen_umin_i64(diff, diff, tcg_constant_i64((vsz >> a->esz) - 1)); > > >> + > > >> + /* > > >> + * Since we're bounded, pass as a 32-bit type. > > >> + * Sink the diff += 1 from above into the 32-bit type. > > > > > > Is it worth doing that? Hosts are all 64-bits now, and surely there's > > > not going to be a perf or codesize difference between "add 1" > > > on a 32-bit type vs a 64-bit type ? > > Technically one byte difference on x86_64. > > But we could drop the whole truncation thing it you like. > > I think mostly my feeling was that since the rearrangement of the > umin() expressions above was already quite complicated, moving > the last part of that down into the truncate-to-32-bits was > adding extra complexity on top. Just doing the subtract-one in > 64 bits and keeping the truncate would be fine, as then the > two steps aren't interleaved with each other.
...but thinking more about it I don't think it's worth respinning the series just for this, so I've added it to target-arm.next (with some cc:stable tags). -- PMM
