| *** tpb has joined #litex | 00:00 | |
| *** tweakoz has quit IRC | 01:37 | |
| *** tweakoz has joined #litex | 02:21 | |
| *** tweakoz has quit IRC | 04:43 | |
| *** tweakoz has joined #litex | 04:49 | |
| _florent_ | Hi, just for info, the mem_map limitation to 256MB regions has been removed, so this mean we can now allocate smaller or larger regions (@daveshah you should no longer need a hack for the 512MB of the DDR3 on your Trellis board) | 08:56 |
|---|---|---|
| daveshah | _florent_: what about regions that aren't aligned to their size? | 08:56 |
| daveshah | I had to move the CSRs below main memory to get 1GB of space starting at 0xC0000000 | 08:57 |
| daveshah | However, moving the CSRs might be the best solution here anyway | 08:58 |
| _florent_ | daveshah: here is the new decoder: https://github.com/enjoy-digital/litex/blob/master/litex/soc/integration/soc_core.py#L98, it should work with regions that are not aligned to their size | 09:00 |
| tpb | Title: litex/soc_core.py at master · enjoy-digital/litex · GitHub (at github.com) | 09:00 |
| _florent_ | daveshah: i changed the default mem_map: https://github.com/enjoy-digital/litex/blob/master/litex/soc/integration/soc_core.py#L135, but we could still change it if not optimal | 09:01 |
| tpb | Title: litex/soc_core.py at master · enjoy-digital/litex · GitHub (at github.com) | 09:01 |
| daveshah | _florent_: that mem map looks good, pretty much what I had | 09:02 |
| daveshah | Thanks! | 09:02 |
| _florent_ | ok good | 09:03 |
| *** tweakoz has quit IRC | 09:32 | |
| flammit | thanks! | 15:18 |
| somlo | daveshah: I synthesized LiteX+Rocket *with* MMU last night for the ecp5versa last night, using -abc9 | 16:26 |
| somlo | utilization was 103% slices :) | 16:26 |
| somlo | REALLY close... | 16:27 |
| somlo | I'll still need an 85k chip, for now... | 16:27 |
| daveshah | somlo: There are two possible scopes for area reduction still. | 16:28 |
| daveshah | synth_ecp5 -nowidelut might decrease area (but also might not) | 16:28 |
| daveshah | Adding -D 100000 to the abc9 call itself should definitely reduce area (https://github.com/YosysHQ/yosys/blob/master/techlibs/ecp5/synth_ecp5.cc#L279) | 16:29 |
| tpb | Title: yosys/synth_ecp5.cc at master · YosysHQ/yosys · GitHub (at github.com) | 16:29 |
| daveshah | But at the expense of Fmax | 16:29 |
| somlo | daveshah: thanks, I'll try those and report back | 16:32 |
| somlo | daveshah: silly question, if I have yosys -p "synth_ecp5 -nowidelut -abc9 -json ...", how do I add "-D 100000" specifically as a "sub-option" to "-abc9" ? | 16:48 |
| somlo | or would I have to unroll all of synth_ecp5 to be able to do that? | 16:49 |
| daveshah | You'd have to modify synth_ecp5.cc | 16:56 |
| daveshah | We need to work out how to make these options more accessible | 16:57 |
| somlo | ok, I can try that -- but first let's see what I can do with just -nowidelut... | 16:59 |
| daveshah | Sounds good | 16:59 |
| *** tweakoz has joined #litex | 17:47 | |
| *** Dolu1990 has joined #litex | 20:55 | |
| *** Dolu1990 has quit IRC | 22:36 | |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!