iss: Processor start, firmware, startup definitions

[Jozott00]

I’m a bit confused about the relationship between the start, firmware and startup definitions in a processor definition.

My understanding is that the firmware defines how the ROM is written, which is essentially the reset vector of the processor. So when starting the full system emulation, the very first thing that gets emulated is the written reset vector.

Consider the following rv64im.vadl example that confirms to the spike machine in QEMU.

micro processor CPU implements RV64IM with ABI = {
  // we will jump to lowest emitted firmware address by default.
  // if this is not correct, we can override it with
  // [start: 0x1000]
  firmware = {
    MEM<4>(0x1000 as Bits<64>) := 0x00000297  // auipc t0, 0x0
    MEM<4>(0x1004 as Bits<64>) := 0x02828613  // addi a2, t0, 40
    // this processor has no zicsr extension
    MEM<4>(0x1008 as Bits<64>) := 0x00000013  // addi x0, x0, 0
    MEM<4>(0x100c as Bits<64>) := 0x0202b583  // ld   a1, 32(t0)
    MEM<4>(0x1010 as Bits<64>) := 0x0182b283  // ld   t0, 24(t0)
    MEM<4>(0x1014 as Bits<64>) := 0x00028067  // jr   t0
    // store start_addr in memory (0x80000000)
    MEM<4>(0x1018 as Bits<64>) := 0x80000000  // lo32(start_addr)
    MEM<4>(0x101c as Bits<64>) := 0x00000000  // hi32(start_addr)
    // we do not yet support a fdt
    MEM<4>(0x1020 as Bits<64>) := 0x00000000  // lo32(fdt_addr)
    MEM<4>(0x1024 as Bits<64>) := 0x00000000  // hi32(fdt_addr)
  }
}

Here we write instructions to the ROM that should be the first thing that is executed when starting the machine. For us the only interesting thing is that it will load 0x80000000 from memory and jump to it, which is the location where the actual bios/bare-metal program starts.

However, a VADL processor might also set start and startup, where start is an expression and startup is a procedure setting things in the environment.

Questions

1. What does the result value of start represent?

2. Is it an alternative to the firmware, so instead of specifying the firmware, one can just set the start address to define where emulation should start?

3. In this case, can the user specify only one of the two, but not both?

4. When is the startup procedure executed?

5. Is it another alternative to start and firmware?

6. Does the firmware read access to memory and general access to registers?
   I would assume no, as the firmware is only supposed to write constant data to memory.
   So resource access is limited to memory writes.

[Jozott00]

We do a complete rework of the processor definition. This is discussed in #138.