Summary: | virtual/blas,cblas,lapack,lapacke: make eselect-ld-so non optional | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Aisha Tammy <gentoo> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | CONFIRMED --- | ||
Severity: | normal | CC: | gentoo, gentoobugs, sam |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 747136 |
Description
Aisha Tammy
2020-10-07 11:19:29 UTC
Also, I should have clarified, this is not just for blas, but also cblas,lapack,lapacke and all others who may be hidden. The USE switch was inspired by soap's message to give users a least surprising default. > 99% of Gentoo users will be happy with just linking against OpenBLAS/reference-lapack. https://archives.gentoo.org/gentoo-dev/message/57049f92ce5386cb77b75f6482a36d2a I like to see eselect-ldso becoming non-optional. As long as David agrees, go for it. Thankfully with the eselect-blas/lapack, we are not using eselect-alternatives so we don't need to worry about symlinks :) (I think soap and I had a tangential comment about alternatives being irrelevant now and probably should be retired, but thats an aside, its part of a much bigger intel parallel studio problem). So far, I am able to get it non optional wherever possible (like mkl, as it has no overhead), but full non-optional behaviour is harder. But I am in no hurry, so we can take our time thinking of a solution (In reply to Aisha Tammy from comment #3) > Thankfully with the eselect-blas/lapack, we are not using > eselect-alternatives so we don't need to worry about symlinks :) (I think > soap and I had a tangential comment about alternatives being irrelevant now > and probably should be retired, but thats an aside, its part of a much > bigger intel parallel studio problem). Glad to hear that. > So far, I am able to get it non optional wherever possible (like mkl, as it > has no overhead), but full non-optional behaviour is harder. Not that much harder. Version bumps with USE flag updates and profit. > But I am in no hurry, so we can take our time thinking of a solution |