The

caravel_core.v

GL netlist doesn't match the

caravel_core.spef

😑 It'd be nice to have the actual SPEF extraction for whatever actually ends up on the chip.

Do you have an example of a difference?

I'm looking at the caravel repo on https://github.com/efabless/caravel/ And if you look for instance in the GL netlist at

net9796

it goes from a

sky130_fd_sc_hd__clkdlybuf4s18_2

to a

sky130_fd_sc_hd__clkbuf_16

But in the SPEF it's between two

sky130_fd_sc_hd__clkdlybuf4s15_2

... And they've been uploaded in two different commits. The netlist and matching spef were uploaded, then two weeks later a new commit saying it was re-hardened and the netlist changed but the new SPEF wasn't added ...

How does TinyTapeout know if the caravel github repo is the source of the macro that was actually used for tapeout, I did not see it in any of the caravel macro project on the Tinytapeout side. It seems to have been picked up magically during the efabless tapeout process. So how do we know where it obtained its copy from ? What is the sha1 of the gds it used ? Then and only then might we compare things to the one on the repo link. So this is why I ask if there are additional downloads in efabless portal that were next to the final OAS file that might contain the actual caravel macro used.

The tapeout submission thing has a tag

mpw-9e

so this should match the tag in the caravel repo with what was actually taped out, at least that's my understanding.

@Passant Do you know who’s responsible for the caravel spef updates?