Hi Stefan Thanks for the reply. Sure, I will look into TCG for improvement as a long term plan. But for now, I have to solve this memcpy performance issue.
Regarding calculating host virtual address from guest virtual address, I think, QEMU must be doing this somehow. I would like to know how QEMU is handling the guest virtual address? It will be very much helpful if you can throw some light on this. Regards Guna On Wed, Mar 9, 2011 at 12:30 AM, Stefan Hajnoczi <stefa...@gmail.com> wrote: > On Tue, Mar 8, 2011 at 5:30 PM, Gunasekaran Dharman <guna.s...@gmail.com> > wrote: > > From guest application, Iam passing some virtual addresses to QEMU > through > > device I/O operation. Now I want to convert these guest virtual addresses > to > > host virtual addresses so that I can perform some operation using them. > In > > QEMU, is there any macro or function available to convert guest virtual > > address to host virtual address? > > My objective is to perform guest's memcpy operation in host so that it > will > > be faster. > > That's not really possible. On architectures with a software-managed > TLB you don't know the layout of the virtual memory structures > (because they are defined in software and not standardized). > > I suggest learning about and optimizing TCG instead of inventing a > memcpy device because it can benefit all code and does not require a > custom device and guest drivers. > > Stefan >