antti....@googlemail.com
unread,Aug 16, 2009, 3:01:08 AM8/16/09Sign in to reply to author
Sign in to forward
You do not have permission to delete messages in this group
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to Antti-Brain
Hi
maybe i need to work on the name (Advanced Brain Machine), but the
concept seems to be finalized for a small, optimized soft processor
core that is high level compiler and multi threading friendly.
some pre-release infos:
---------------------------------
* 8 bit Alu/Datapath
* Windowed Register File, 16 Registers in each context
* optimized to work from single block RAM
- minumum 1 BRAM with 256x16 + 512*8 aspect ratios needed (address can
be larger)
* Not using distributed RAM (friendly for Altera, Actel, Silicon
Blue..)
* Fast context switch
- nested interrupts and/or
- thread switching
* Instruction cycle 4 clocks
* All Instructions execute in one cycle
yes, there is a compromise, it is not executing 1 CPI (clock per
instruction), but wasting 4 clocks.
I tried to use only 2 clocks per instruction but it really isnt
possible.
sure it possible to have soft core processors to run 1 clock per
instruction, but for the goal it
really isnt possible, there are delays for fetching code and
registers, and if all has to be streamed
from/to single BRAM there is a need for more then 1 clock per
instruction.
more details to follow.. :)
Antti