| 19 Feb 2024 |
 sorear | i've been calling the situation "new golden age of processor errata" for years, i just don't see a need to single out t-head and I'm paranoidly wondering whether they get held to a different standard because of geopolitics | 13:30:48 |
| sorear | yes (lp4a=th1520) | 13:30:56 |
 Pratham Patel (you can mention me) | probably | 13:31:17 |
 Steven Keuchel | In reply to @sorear:matrix.org
t-head has a userspace erratum where the FP underflow flag is incorrectly not set if a multiply rounds away from zero to produce the smallest normal number, AFAIK the only thing this affects is the glibc regression test suite thanks for that! I am going to read up on it. | 13:44:25 |
| sorear | In reply to @skeuchel:matrix.org
thanks for that! I am going to read up on it. https://github.com/revyos/revyos/issues/17 fwiw | 13:46:22 |
 Alex | In reply to @thefossguy:matrix.org
So far, the only machines I trust for building nixpkgs are the HiFive Unmatched and the VisionFive 2. You should also count the Star64, because it uses the same SoC as the VF2 (unless you have a reason for excluding it too?) | 14:33:01 |
| Pratham Patel (you can mention me) | Ah no reason to exclude it, just forgot about it! :D | 14:33:26 |
| Alex | On the topic of implementation errata, I wonder if anyone's doing any exhaustive testing on the various RISC-V processor designs... | 14:34:10 |
| Pratham Patel (you can mention me) | there are a million (hyperbole) Indian startups doing exactly that! | 14:34:52 |
| Pratham Patel (you can mention me) | the last tenstorrent event almost a year ago in BLR was exactly that :) | 14:35:15 |
| sorear | it's hard to exhaustively test anything of that size, especially when the specs themselves are nondeterministic (memory ordering, counters) or just plain ambiguous | 14:37:22 |
