HUBEXEC: TIPS & TRAPS - All about the restrictions of running code from hub ram.....

Comments

• Cluso99 Posts: 18,069

  2014-03-26 19:52 edited 2014-03-26 19:52

  1. Self-modifying code is not possible for hub code (SETD/SETS/SETINST/SETCOND cannot modify hub code).
  
  While there may be more complex ways to achieve this, it is more than likely to fail.
  It is also possible to use HUBEX code from multiple cogs/tasks/threads concurrently (typically known as re-entrant code) and this would be broken if self-modifying code were used!
• Bill Henning Posts: 6,445

  2014-03-26 20:11 edited 2014-03-26 20:11

  2. Be careful about referring to orgh hub labels and not org cog labels
  
  p.s.
  
  p2 docs call it "hub execution", on the threads we've shortened it to "hub exec" and even "hubexec"
  
  I don't like "hubex" as it does not clearly imply execution.
• cgracey Posts: 14,133

  2014-03-26 22:11 edited 2014-03-26 22:11

  LOCBASE/LOCBYTE/LOCWORD/LOCLONG/LOCINST D,@ and LOCPTRA/LOCPTRB @ are best used by code executing from the hub, since relativity is maintained. When these instructions are executed from the cog, the absolute address of the @ must be known in order to properly compute PC+@, since PC is now (absolute) $000.$1FF. Also, the @ must always refer to hub memory space (ORGH) and not cog space (ORG).
• potatohead Posts: 10,254

  2014-03-26 22:34 edited 2014-03-26 22:34

  I like HUBEXEC too. I've typed it both ways, and always think of FEDEX...
• Sapieha Posts: 2,964

  2014-03-26 23:59 edited 2014-03-26 23:59

  Hi.
  
  For real Addresses use for ORGH.
  
  Real Address divided by 4
  
  orgh $E00 /4 '$380 = 18-bit load address $E00
  orgh $1000 /4 'some hub code at $1000
  
  
  Else it will not be address that You maybe was thinking to use