Re: [Qemu-devel] Re: [PATCH] CONFIG_MMU_MAP powerpc host support

[Daniel Serpell <daniel_serpell@yahoo.com>]

Hi!

El Mon, Dec 20, 2004 at 11:12:31PM +0100, Magnus Damm escribio:
>
[...]
> But if we could combine the most popular
> guest instructions into one micro operation (which is not so micro
> anymore) then the compiler could rearrange things to fully take
> advantage of the host cpu.
> 
> It all boils down to some table-based generic guest opcode matching
> code that does a longest prefix match and supports masking of
> bitfields...
> 
> I think it would be very interesting to collect opcode statistics for
> certain guest operating systems. Or maybe someone already has done
> that?
> 

I have done that, it's not so difficult. I added a fixed static two
dimensional table (int ins_usage[MAX_INS][MAX_INS]), an then a global
variable "int last_ins".

Also, I moddified the "ops*.h" and "op.c" files, adding at the
beggining: "ins_usage[last_ins][THIS_INS]++; last_inst=THIS_INS;",
with THIS_INS initialized to some unique constant.

The problem is, I then realized that I measured the instructions
usage but not the time spent in each instruction, and the instructions
doing memory-access where spending almost all the emulated time.

So, I presumed that accelerating memory-access were a better project.

If you are interested, I could send the patches.

        Daniel.