Hello Andrius, Thanks. Some other minor issues have emerged with some file cleanups during install and a fix will be pushed soon. > Yes, at least for the ones with white background in the buildd table. > Otherwise these archs should be excluded. Looks like mmseqs2 is really reliant on either AVX2 or SSE 4.1 so I will exclude everything apart from amd64 and kfreebsd. > By the way, build on kfreebsd fails due to hardcoded triplet path, could > you investigate this as well? I overlooked the fact that this directory changes depending on the arch :/. I will also fix this shortly. Kind regards, Shayan Doust On 18/10/2019 06:57, Andrius Merkys wrote: > Hi Shayan, > > On Thu, 17 Oct 2019, 23:09 Shayan Doust, <hello@shayandoust.me > <mailto:hello@shayandoust.me>> wrote: > > > This is a well known issue, I will fix this tommorrow > Thanks! > > > Done now. > > It seems like mmseqs2 strictly relies on various SSE extensions that are > not found on other architectures apart from that of amd64. I will see > what I can do and if no specific SSE-reliant code is used, maybe I can > rely on conditions in cmake to build without these extra expected > extensions. > > > This would be great. > > Generally, is it important for all builds across all architectures > to pass? > > > Yes, at least for the ones with white background in the buildd table. > Otherwise these archs should be excluded. > > By the way, build on kfreebsd fails due to hardcoded triplet path, could > you investigate this as well? > > Best, > Andrius
Attachment:
signature.asc
Description: OpenPGP digital signature