riscv: define mvendorid CSR with macro helpers

[rmsyn]

Uses CSR macro helpers to define the mvendorid CSR register.

Related: #229

[romancardenas]

Instead of 0x80, I think there goes the corresponding odd parity bit of offset. We can either include this logic or ignore the parity bit, as done in the mvendorid register. Another option is completely removing jedec_manufacturer and leaving bank and offset

[rmsyn]

Instead of 0x80, I think there goes the corresponding odd parity bit of offset.

I'm not sure I understand what you mean here. The 0x80 is the odd parity bit added back to the offset number.

Another option is completely removing jedec_manufacturer and leaving bank and offset

This is my preferred route. I think we should leave it up to users on how they want to encode the JEDEC manufacturer number.

I'll add some documentation to the offset field about the parity bit.

---

Not sure in practice, but the theoretical max number of continuation bytes is 0x1ff_ffff. We obviously can't output such a number ([0x7f; 0x1ff_ffff]) in an embedded environment (not saying you suggested that, just saying). However, maybe some user wants to return an Iterator of continuation bytes 🤷

I'm not so familiar with the use of JEDEC manufacturer numbers, and how they're usually handled in industry.

[romancardenas]

The 0x80 is the odd parity bit added back to the offset number.

The odd parity bit is 0x80 only when the number of 1s is even (e.g., 0x0a becomes 0x8a). But it should be 0x00 when the number of 1s is already odd (e.g., 0x0b remains 0x0b).

However, maybe some user wants to return an Iterator of continuation bytes

It might be useful for someone to implement an iterator that shows the full JEDEC manufacturer number. If you are in the mood to implement it, then go ahead. Otherwise, we could act lazily and wait for someone who asks for this feature.

[rmsyn]

It might be useful for someone to implement an iterator that shows the full JEDEC manufacturer number. If you are in the mood to implement it, then go ahead. Otherwise, we could act lazily and wait for someone who asks for this feature.

I agree. I've implemented the Iterator solution, however there is a major Git operations outage on GitHub right now. I'll push my changes as soon as I can.

[rmsyn]

The odd parity bit is 0x80 only when the number of 1s is even (e.g., 0x0a becomes 0x8a). But it should be 0x00 when the number of 1s is already odd (e.g., 0x0b remains 0x0b).

I've corrected my implementation to set the parity bit based on the 1s count parity in offset.

For example, an odd offset with an even amount of 1s should have the parity bit set (0x03 -> 0x83)? This appears correct, since this is the JEDEC ID for Fairchild.

[romancardenas]

We can leave it as is or add an iterator for the JEDEC manufacturer number. Whatever you prefer @rmsyn

[romancardenas]

LGTM!

I suggested a small optimization that leaves computing the parity bit when needed. The idea is to reduce the size of the resulting iterator, as it captures one less value.

[romancardenas]

I suggested some niche optimizations to reduce even more the size of the resulting iterator

[romancardenas]

Suggested change

	let mut done = false;
	let mut bank = self.bank();
	let offset = self.offset();
	let parity = ((1 - (offset.count_ones() % 2)) << 7) as usize;
	let mut bank = self.bank() + 2; // add 2 for Some(offset) + None
	let offset = self.offset();

[romancardenas]

Suggested change

	0 if done => None,
	0 => {
	done = true;
	0 => None,
	1 => {
	bank -= 1;
	let parity = ((1 - (offset.count_ones() % 2)) << 7) as usize;