[Jikesrvm-core] Re: Port to 64-bit AMD (Opteron/Athlon)

[kris v. <kri...@el...>]

Hi Eliot,

sorry for the late respons, holidays !

We had a student (Frederik De Schrijver) working on this port. But now 
the work is stopped unfinished and we have no one to continue the work.
Frederik had adjusted the c code for startup and added infrastructure 
for the extra AMD64 instructions.
Some changes were already made in the baseline's VM_Compiler, but I 
believe that that migth be not 100% complete.
The work was stopped at the testing of bytecodes during startup.

I also agree that the remainder of the work to be done should be 
straightforward, because of the existence of the PPC64 port.

Kris Venstermans
Ghent Univerity, Belgium


>From: Eliot Moss <mo...@cs...>
>Date: Sat, 30 Jul 2005 17:53:26 -0400
>To: jik...@li...
>Subject: [Jikesrvm-core] Port to 64-bit AMD (Opteron/Athlon)
>Reply-To: jik...@li...
>
>Having heard nothing in reply, I am trying again before I meet with my
>contacts on Monday :-).
>
>Since last Monday I have also looked at the AMD 64 arch books and find that
>the extensions are pretty straightforward. They add 8 more GP regs (total
>16), with all regs 64 bits long. They scarf a new set of prefix
>instructions which give the high-order bits for registers numbers (when all
>are not 0) and control the operand width (when it is not the default,
>generally 32 bits). Both the greater number of registers and the fact they
>one can pretty much use them in a "flat" way seem good. In fact, one could
>do a 32-bit-address version that takes advantage of 64-bit registers/ops
>for longs and exploits the larger register set, and also a full 64-bit
>version.
>
>Enough for now .... Eliot
>  
>