Fix 748 - Updates RAM32X1S property to correct default

[clavin-xlnx]

Fixes #748 by updating unisim_data.dat with the appropriate setting for IS_CLK_INVERTED.

[eddieh-xlnx]

Looking back at #748, what should RapidWright do if we instantiate a RAM32X1S_1 instead here? I'm seeing that Vivado emits this:

create_cell -reference RAM32X1S_1 inst
WARNING: [Coretcl 2-1024] Primitives of type 'RAM32X1S_1' are not supported by the current part and have been retargeted to 'RAM32X1S'.

However, RapidWright doesn't retarget (not reasonable to expect it to retarget to RAM32X1S?) but it might be reasonable to expand this directly to the end result (where inst/SP is a RAMS32) with the IS_CLK_INVERTED set correctly?

I'll push my modified test; we can revert if it's out of scope of this PR.

[clavin-xlnx]

However, RapidWright doesn't retarget (not reasonable to expect it to retarget to RAM32X1S?) but it might be reasonable to expand this directly to the end result (where inst/SP is a RAMS32) with the IS_CLK_INVERTED set correctly?

RapidWright doesn't support deprecated primitives that require re-targeting. It doesn't have the retargeting information necessary to do it properly. It also doesn't contain an exhaustive list of primitives that require retargeting.

Usually, this doesn't cause an issue as primitives are retargeted automatically in Vivado. It would be a significant task to add this support and handle this more elegantly but its unclear that there is a significant benefit.

I'll push my modified test; we can revert if it's out of scope of this PR.

The best we could do for now is just detect these specific cases and throw an error, but it wouldn't be exhaustive for all other retargeted prims.

[eddieh-xlnx]

That's reasonable, let's revert my test. In this case, would we expect the IS_CLK_INVERTED property of a RAM32X1S_1 instance created by/loaded into RapidWright to be unreliable then?

[clavin-xlnx]

That's reasonable, let's revert my test. In this case, would we expect the IS_CLK_INVERTED property of a RAM32X1S_1 instance created by/loaded into RapidWright to be unreliable then?

That is correct.

[eddieh-xlnx]

Changed unsupported macros into a canary test.