| openpowerbot | [mattermost] <lkcl> confirmed, that is the "voodoo magic incantation" needed to be added to Memory-declarations to get them to be treated as BRAMs by yosys ecp5 rules | 01:44 |
|---|---|---|
| openpowerbot | [mattermost] <lkcl> it works on cacheram, and because i've converted tagmem to Memory declarations, works on those too | 01:45 |
| openpowerbot | [mattermost] <lkcl> no idea if it will work as an attribute on registers (next door to where "ram: distributed" is put in various places) but it's worth trying | 01:46 |
| lkcl | ha. adding this *increased* resource utilisation for icache.vhdl: | 10:23 |
| lkcl | attribute syn_ramstyle of cache_tags : signal is "block_ram"; | 10:23 |
| lkcl | but adding this *decreased* resource utilisation: | 10:24 |
| lkcl | attribute syn_ramstyle of itlb_tags : signal is "block_ram"; | 10:24 |
| lkcl | attribute syn_ramstyle of itlb_ptes : signal is "block_ram"; | 10:24 |
| lkcl | because yosys was able to turn itlb_tags/ptes into $mem_v2. | 10:24 |
| lkcl | somehow | 10:24 |
| openpowerbot | [mattermost] <lkcl> oh hang on $mem_v2 means it doesn't match known patterns | 15:55 |
| openpowerbot | [mattermost] <lkcl> $ | 15:55 |
| openpowerbot | [mattermost] <lkcl> $mem_v2 is an abstract high-level concept: if it doesn't get mapped down to BRAM, SRAM, or (last resort DFFs) that's a bad sign | 15:56 |
| openpowerbot | [mattermost] <lkcl> icache is weird because it's a combinatorial lookup for the ptes and tlb (same clock cycle) | 15:56 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!