On 5/4/19 5:05 AM, Paolo Bonzini wrote: > The speedup is similar to x86, 120 ns vs 180 ns on an APM Mustang. > > Signed-off-by: Paolo Bonzini <pbonz...@redhat.com> > --- > configure | 2 +- > scripts/qemugdb/coroutine_asm.py | 6 ++++- > util/Makefile.objs | 2 ++ > util/coroutine-asm.c | 45 ++++++++++++++++++++++++++++++++ > 4 files changed, 53 insertions(+), 2 deletions(-)
Reviewed-by: Richard Henderson <richard.hender...@linaro.org> > + "ldr x30, [x1, %[SCRATCH]]\n" /* load destination PC */ \ > + "ldr x1, [x1, %[SP]]\n" /* load destination SP */ \ > + "mov sp, x1\n" \ > + "br x30\n" \ > + "2: \n" \ For future reference, "bti j" (aka hint #36) goes here, for the aarch64 branch target identification extension. r~
