Conversion of '0 and '1

[NilsGraf]

'0 and '1 are currently converted to 1'sb0 and 1'sb1, resp. The latter is not accurate for multi-bit assignments (because '1 actually means that all bits are set to 1, not just the LSB). And the former causes LEC issues when 1'sb0 drives a multi-bit input of a module (resulting in the upper input-bits being undriven).

Would it make sense to use the following translations instead?

• 1'sb0 to 'd0 (this works fine in LEC for multi-bit inputs)
• 1'sb1 to -'sd1 (note that signed -1 is all-ones)

See also this workaround. Thanks!

[zachjs]

Handling these "unbased unsized" literals has proven deceptively complex. In self-determined contexts (i.e., {'1, 'x}), they should behave as though they are a single bit. Otherwise, they fill the context-supplied size.

As far as I can tell 'd0 is exactly the same as 0, and -'sd1 is exactly equivalent to -1 (but not -'sb1). I recently added a hack to deal with the sizing associated with ternary expressions. Your suggestion of -'sd1 thankfully eliminates the need for that hack. However, I think to make this conversion work properly, it must be context aware, selecting either the sized or unsized replacement depending on where it is used..

[NilsGraf]

Awesome. I think mapping '0 and '1 onto 0 and -1 should also work fine.

[zachjs]

I think 8cfd05de1a821cc91986643b48249e637882b81f should be closer to what you're looking for. Please let me know if you find any issues.

[zachjs]

Has this issue been resolved?

[NilsGraf]

Yes, thanks!