Re: [openrisc] OR16 ISA

[jimmy87@sunplus.com.tw]

I got it. I have missed communication system for a long time.
But communication problems are kind of different from insn encoding.
Although those insns are seldom used, we still need to use it. And those
insns still occupy 16-bit rather than just opcodes. One insn contains
both opcode and operands. We can widen the opcode of
the seldom used insn, but it still has operands...operands can not be
Huffman encoded...
kind of trouble.
Best regards, Jimmy





"Damjan Lampret" <lampret@opencores.org> 於 2000/05/26 06:06:24 PM

請回應 給 openrisc@opencores.org

收件人：  openrisc@opencores.org
副本抄送： (副本密送： jimmy87/Sunplus)

主旨：    Re: [openrisc] OR16 ISA



Insns that have more operands and are more often used should have less
opcode bits. And it would be easier to squeez these insns into 4-bit opcode
w/o insns that are not often used (test, set etc.). You can run dhry
benchmark in or1ksim and see which insns are more commonly used. You will
see that these are l.lw, l.sw, l.lbz, add insns and compare/branch insns.

--damjan

>
> I don't get it.
> how to squeeze the following 8-bit opcode into 4-bit?
> Best regards, Jimmy
>