The Core crate decodes and executes instructions, but does not fetch instructions from memory.
In order to fetch instructions, a core needs to know where the code is actually loaded in memory.
This could be overwritten by the still missing executable format, but I propose that the default code section starts at 0x100 in memory. Unless overwritten, $RIP should be initially set to that value, and this is where the first instruction will live (a.k.a. as the reset vector).
The fetch-decode-execute loop should then be as simple as:
- Read 32 bits from $RIP, that's the next instruction to decode and execute
- Increase RIP by 4 (next instruction, 32 bits further)
- Decode the instruction
- Execute the instruction
- goto step 1
Note that the instruction executions step (step 4) could modify RIP as well (branches and jumps).
The Core crate decodes and executes instructions, but does not fetch instructions from memory.
In order to fetch instructions, a core needs to know where the code is actually loaded in memory.
This could be overwritten by the still missing executable format, but I propose that the default code section starts at 0x100 in memory. Unless overwritten,
$RIPshould be initially set to that value, and this is where the first instruction will live (a.k.a. as the reset vector).The fetch-decode-execute loop should then be as simple as:
Note that the instruction executions step (step 4) could modify RIP as well (branches and jumps).