| *** tpb has joined #vtr-dev | 00:00 | |
| mithro | elms: So is this what your getting... | 17:45 |
|---|---|---|
| mithro | https://www.irccloud.com/pastebin/ym3VzKqR/ | 17:46 |
| tpb | Title: Snippet | IRCCloud (at www.irccloud.com) | 17:46 |
| elms | yep | 17:46 |
| mithro | elms: Probably should be neigh_op_lft_1 | 17:50 |
| mithro | elms: Bet it's because vpr doesn't understand the x_offset / y_offset on the metadata yet | 17:51 |
| mithro | https://www.irccloud.com/pastebin/69jbcQIj/ | 17:51 |
| tpb | Title: Snippet | IRCCloud (at www.irccloud.com) | 17:51 |
| elms | mithro: ok I can start looking at that, unless you already have clear thoughts on it. | 17:53 |
| mithro | I have pretty clear thoughts on that - I'm going to work on it now | 17:54 |
| daveshah | mithro: just an FYI, but it should be safe to treat the neighbour interconnect as configurable interconnect, then ignore the configuration, if shorts aren't working | 18:14 |
| mithro | daveshah: Nah, shorts are working | 18:14 |
| daveshah | Cool | 18:14 |
| mithro | daveshah: The issue is that hlc doesn't have a "global name" for the neighbourhood wires | 18:15 |
| daveshah | Aha | 18:15 |
| mithro | daveshah: Which means a wire which spans multiple tiles only has multiple local names | 18:15 |
| daveshah | Solution: abandon hlc :P | 18:15 |
| mithro | daveshah: I always planned to support multiple local names in different tiles - but hadn't needed to implement it yet | 18:15 |
| mithro | I think elms was running into the problem with the ram tiles -- as vpr only has one object which expands across multiple grid coordinates | 18:16 |
| elms | For the ram (and others that will span multiple grids), it sort of gets messy around mapping between the RAM(block) and RAMB/RAMT(tiles) in VPR. Splitting the RAM into RAMB/RAMT for HLC output is clumsy and needs some additional information about how params get mapped to the different tiles. | 18:26 |
| elms | daveshah: Replacing HLC could be a solution. I'd need more time getting my hands dirty to be able to provide useful input on a good replacement though. | 18:27 |
| daveshah | elms: yeah, I have a similar problem with the parameter split with my ecp5 work | 18:27 |
| daveshah | I think what I will do is just ignore parameters that are not included in the tile under some conditions, so they can be written in all tiles | 18:28 |
| daveshah | IMO a future IL for the iCE40 bitstream would be based on icebox_explain output | 18:28 |
| daveshah | Or JSON based, if we wanted to go that way, but lower level than HLC | 18:29 |
| mithro | elms: | 23:22 |
| mithro | https://www.irccloud.com/pastebin/DXDtQ3B7/ | 23:22 |
| tpb | Title: Snippet | IRCCloud (at www.irccloud.com) | 23:22 |
| mithro | Oh | 23:30 |
| mithro | local_g0_2 -> io_1/D_OUT_1 | 23:30 |
| mithro | elms: Should be an easy fix - testing now | 23:34 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!