New: ddrForwardClock

[DigitalBrains1]

This PR was inspired by the clash-cores PR #7 Add RGMII core which uses this construction but keeps the output clock as a Signal domDDR Bit. With the new ddrOutClock, we can correctly declare the thing to be a clock.

There's a qualitative difference between Clock and Signal _ Bit. In a nutshell, a signal is relative to a clock, it has setup and hold constraints and carries data. All of these don't apply to a clock. You could say a clock leads and a signal follows. The resulting output on the pin is no different, it's just about correctly declaring what something is in the Haskell world.

The new ddrForwardClock function gets a DDR output primitive as an argument. If you look at clash-prelude, the only directly fitting one is Clash.Explicit.DDR.ddrOut (if you apply the reset value first). However, in a PR that I will create soon, the API for Clash.Intel.DDR and Clash.Xilinx.DDR is changed to accomodate the signature introduced in this PR. Even with the current code, it's possible to adapt the vendor primitives to fit ddrForwardClock, but soon it will be a natural fit.

Still TODO:

• Write a changelog entry (see changelog/README.md)
• Check copyright notices are up to date in edited files

[DigitalBrains1]

GHC 8.10.7 should get its priorities straight:

    • Expected kind ‘Nat’,
        but ‘DomainPeriod domIn ~ 2’ has kind ‘Constraint’
    • In the first argument of ‘(*)’, namely ‘DomainPeriod domIn ~ 2’
      In the type signature:
[...]
                        DomainPeriod domIn ~ 2 * DomainPeriod domDdr) =>
[...]

Why is this same expression no longer a problem in clash-cores PR #7 Add RGMII core? I couldn't get it to work here. I did notice that code had fewer KnownDomains than I expected; if you look at our docs, a use of DomainPeriod is documented as needing KnownDomain in its example, but apparently you don't?

I'll remove some KnownDomains and add parentheses to the expression.

[rowanG077]

Yes this is very nice! Thank you!

[rowanG077]

Perhaps ddrForwardClock is a more appropriate name as this technique is generally referred to a clock forwarding, or at least that's what I know it as and is googleable.

[DigitalBrains1]

I've changed the name from ddrOutClock to ddrForwardClock.