| *** tpb has joined #vtr-dev | 00:00 | |
| mithro | daveshah: Over here now :-P | 14:00 |
|---|---|---|
| mithro | daveshah: One thing which kind of sucks with the pb_type specification format is that you have to specify the parent name in the interconnect stuff | 14:00 |
| daveshah | mithro: yeah, most of the problems are more down the the limits of pb_type | 14:01 |
| mithro | daveshah: So you can't just do a <pb_type name="xxx"><xi:include "blah" xpointer="xpointer(pb_types/child::node())" /></pb_type> | 14:02 |
| mithro | daveshah: I feel like there is probably something clever you could do with xi:include or xslt to make it work.... | 14:03 |
| daveshah | mithro: yeah, not sure if that is worth it vs generating with Python though | 14:04 |
| mithro | daveshah: Yeah | 14:04 |
| daveshah | there are probably similar issues when trying to infer a num_pbs from a `for generate` | 14:04 |
| mithro | daveshah: I actually did like jhol's idea of creating a decent Python API | 14:05 |
| daveshah | mithro: Personally I'd rip out the entire pb_type system and use an alternative API to get the arch into VPR in the first place | 14:05 |
| mithro | daveshah: So, I noticed that you were in favor of strong APIs and weak data formats when talking about libtrellis | 14:06 |
| mithro | daveshah: And we should have a discussion about why I think strong file formats frequently work out better | 14:07 |
| mithro | Actually, I just had an idea... | 14:14 |
| mithro | Like most of my ideas, it's probably a bad one -- but I'll see | 14:15 |
| mithro | holy crap -- this is horrible that you can do this..... | 14:41 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!