PASM question -- HUBOP implementation in the instruction set
Wossname
Posts: 174
I'm back with another awkward edge-case question about something nobody else will ever care about 
(might be one for @cgracey if he has a free second during a lunch-hour, hehe)
In a nutshell, why are...
CLKSET, COGID, COGINIT, COGSTOP, LOCKNEW, LOCKRET, LOCKSET, and LOCKCLR *
...considered to be "HUB operations", whereas...
RDBYTE, RDWORD, RDLONG, WRBYTE, WRWORD and WRLONG
...are not described in the same manner?
All the above take need HUB sync and take between 8 and 23 system clocks to run.
I realise that the first 8 all share the same opcode number ('iiiiii' field in the 32 bits mapping are all %000011). And that the three LSBs of the source field are used to determine which HUBOP is executed. But there why not use the 4 LSBs (it looks to be available(?)) instead and just incorporate the latter 6 opcodes into the same HUPOB %000011 umbrella?
I'm merely curious, I noticed this just now while writing my simulator.
*Ahh, the Oxford comma :P
(might be one for @cgracey if he has a free second during a lunch-hour, hehe)
In a nutshell, why are...
CLKSET, COGID, COGINIT, COGSTOP, LOCKNEW, LOCKRET, LOCKSET, and LOCKCLR *
...considered to be "HUB operations", whereas...
RDBYTE, RDWORD, RDLONG, WRBYTE, WRWORD and WRLONG
...are not described in the same manner?
All the above take need HUB sync and take between 8 and 23 system clocks to run.
I realise that the first 8 all share the same opcode number ('iiiiii' field in the 32 bits mapping are all %000011). And that the three LSBs of the source field are used to determine which HUBOP is executed. But there why not use the 4 LSBs (it looks to be available(?)) instead and just incorporate the latter 6 opcodes into the same HUPOB %000011 umbrella?
I'm merely curious, I noticed this just now while writing my simulator.
*Ahh, the Oxford comma :P
Comments
whereas the other hub ops only use a dst register and can share an op-field value.
This is the sort of compromise you always see in real instruction sets (except perhaps in
some VLIW or microcoded architectures).
Note that the RDLONG and WRLONG operations actually have the same op-field, the r-flag
determines which one is meant...
The opcode DOCs do expand this a little, some are control commands and some are RAM (or FIFO) commands
CLKSET Hub Control - Clock
COGID... Hub Control - Cogs
LOCKNEW... Hub Control - Locks
WRBYTE... Hub RAM - Write
RDFAST Hub FIFO - New Read
RFWORD Hub FIFO - Read
I'm not sure the Control group even need 'Hub' word in there, as they are more P2 Chip Config & Control, than anything 'HUB memory' related