Shop OBEX P1 Docs P2 Docs Learn Events
Why not more COG memory? - Page 2 — Parallax Forums

Why not more COG memory?

2»

Comments

  • Bill HenningBill Henning Posts: 6,445
    edited 2010-11-13 17:22
    Good catch!

    Obviously, this needs some more thought - for example, waitxxx and cogxxx need to be examined for potential bad effects...

    Fortunately, the 23 bits available with all top 9 bits being 0 shifted left two positions due to long alignment give a 32MB (8MLong) address space, and I am sure we can still recover some bits from the top bits of instruction encoding.

    Perhaps a better method would be to use 0000 for the condition bits, then by using

    ROL address,#11

    we would have a 1GB (256MLong) address space... with all conditions zero, none of those pesky WRxxx could cause a problem.

    Mind you, this could be a problem on Prop2 if Chip used some of the NOP patterns...
    kuroneko wrote: »
    Careful with that one. wrxxxx is encoded with R = 0 (rdxxxx uses R=1) and that may have drastic results.
  • Cluso99Cluso99 Posts: 18,069
    edited 2010-11-13 19:03
    It could well be possible to just ensure the top 4 bits are "0001" which gives 28 bit address space (256MB IIRC without working it out). This would decode to the unused prop instructions mul/muls/enc/ones. Obviously this needs checking to ensure the prop does indeed ignore these op codes. I am sure Chip would answer if this is needed.
  • kuronekokuroneko Posts: 3,623
    edited 2010-11-13 21:25
    Cluso99 wrote: »
    This would decode to the unused prop instructions mul/muls/enc/ones. Obviously this needs checking to ensure the prop does indeed ignore these op codes.

    Unless nr is specified those instructions clear the destination register.
Sign In or Register to comment.