2022-04-04 21:15:06 <KipIngram   > This has some interesting ideas in it, especially for Forth hardeware:
2022-04-04 21:15:09 <KipIngram   > https://www.youtube.com/watch?v=kSQ3BpPJEH4
2022-04-04 21:15:59 <KipIngram   > He introduces an idea where literals are built up in sections.  This  lets you have a one-byte wide instruction bus, but the data section can be any width.
2022-04-04 21:16:25 <KipIngram   > You build up literals however wide you need by a succession of literal bytes
2022-04-04 21:17:09 <KipIngram   > And I THINK where he's going here is that jump targets PRECEDE the jump opcode.  I assume the target gets pushed to the stack, like any other literal, and then jump takes the target from the stack.
2022-04-04 21:17:32 <KipIngram   > It really does make the instruction stream flawlessly single-width, and all instructions are one instruction word wide.
2022-04-04 21:17:47 <KipIngram   > Interesting idea.
2022-04-04 21:18:25 <KipIngram   > Distant jumps would take longer than near jumps - it would take longer to build the literal.
2022-04-04 21:19:02 <KipIngram   > If an instruction word has the MSB set, all the rest of the bits are literal bits.  If not, it's an instruction.
2022-04-04 21:24:26 <KipIngram   > I don't see that as terribly useful for a software implementation, but it looks like it could be quite interesting for a Forth processor.
2022-04-04 21:33:11 <KipIngram   > In his hardware work, the benefit that gave is that "jump value" is all one instruction - is a multicycle instruction, and required a longer period of not getting interrupted.  Whereas "literal; jump" takes no more time (two cycles) but is two single cycle instructions, alloweing interrupts in between.
2022-04-04 21:34:15 <KipIngram   > I'm not quite sure yet how he distinguishes between a multi-component single literal and a string of separate literals.