IRC channel logs


back to list of logs

***sebbu2 is now known as sebbu
<civodul>hey there!
<civodul>rekado_: looks like birdsite is back
<civodul>i tried logging in just in case and it worked, no attempt to use 2FA
<civodul>ah no, the account is still marked as "suspended" and cannot send tweets
<rekado_>this is arguably worse than before
<rekado_>at least before people could click on something to show past tweets
<rekado_>now they can’t
<rekado_>everything just appears to be gone
<rekado_>zimoun, I’ll work on wip-r today
<rekado_>I’ll first get myself a list of changes with “guix refresh -t bioconductor” on another branch, so that I can fix problems that might remain on wip-r
<civodul>i looked at the account options but there's nothing obvious that can be done
<civodul>it seems to say that 2FA is disabled
<civodul>there's a footnote linking to
<rekado_>my contact with Twitter was really muddled; at first it was about not being able to access the account, but subsequent steps always were about suspended accounts
<rekado_>so I’m not surprised that it’s now officially in suspended stat
<civodul>it's crazy that it's not immediately obvious that it's suspended when you log in
<civodul>and it doesn't offer a procedure to follow to restore it
<rekado_>we should have known better than to violate their terms of service, which clearly state … *paper rustling* … uhm, nothing about logging in from different locations.
<civodul>it's for our own good though: they noticed "suspicious activity"
<rekado_>ah, true, true. I feel much safer now.
<rekado_>zimoun: r-gqtlbase is broken not only in Guix, but also upstream.
<rekado_>fixed it by patching the NAMESPACE file
<rekado_>I’m trying to build pigx now, because it uses a lot of bioconductor packages; if that works I’ll push my changes and let cuirass take care of the rest.
<rekado_>I made a few changes, removing inputs that are no longer needed; worked off the list of recommended changes that “guix refresh” spat out.
<civodul>hey bavier[m] :-)
<bavier[m]>just looking at your FMV stuff a bit.
<civodul>i'm trying to benchmark things to see if it makes a difference
<civodul>so i thought about GSL, because it's unoptimized in the first place
<civodul>but then we'd need more convincing examples too
<bavier[m]>I was displeased with Clear Linux's fmv script for some reason that I forget, so I wrote my own a while back, for experimentation purposes.
<civodul>oh nice!
<civodul>if you already have experience fiddling with that, that's great
<bavier[m]>I experimented with HYPRE a little
<bavier[m]>and found the "shotgun" approach to be less great
<civodul>you mean adding target_clones attributes to every single vectorizable function?
<bavier[m]>I think gcc even spits out that diagnostic even it it doesn't end up vectorizing or using target-specific instructions.
<civodul>here i'm using -fopt-info-vec on the generic ISA (x86_64)
<civodul>so it just returns functions that can be vectorized on generic x86_64
<civodul>then when you add target_clones on them, clones are not necessarily generated
<bavier[m]>ok, so gcc does additional analysis before creating clones?
<civodul>yes, that's my (limited) understanding
<bavier[m]>anyhow, some more experimentation necessary.
<bavier[m]>I haven't had a chance in a while to do much more.
<bavier[m]>there are certainly benefits that could be gained from having gcov/profile data, but that becomes a multi-step process then.
<civodul>you mean coverage data?
<civodul>to see which specialized function is called?
*civodul goes afk for a bit
<bavier[m]>to create clones for "hot" functions and not for "cold"
<bavier[m]>in case anyone's curious:
<bavier[m]>hi zimoun !
<zimoun>rekado_: thanks for the r-gqtlbase link. How is going the wip-r? I just have tried and it appears that substitutes seems available. But r-diffbind fails because upstream update. Well bug#39885 is becoming high priority in my list. :-)
<rekado_>zimoun: I just finished rebasing; I have the r-diffbind upgrade. Will delete and push the branch.
<rekado_>zimoun: pushed it. Now we just need to wait for enough substitutes to become available.
<rekado_>I could merge wip-r tomorrow or the day after.
<zimoun>rekado_: awesome! Thank you.
<rekado_>thank *you*! All these commits of yours must have been quite a lot of work.