Revert accidental squash merge of enable_symm#4
Closed
annayesy wants to merge 2 commits into
Closed
Conversation
* lu stabilisation * deserialise pivotmethod as type, value * remove stabilisation from diagonal blocks * diag factors stabilisation * rlst type fix * return factor norms * op + alpha*I type of stabilisation * modify results path when stabilising * fix results path format * addition of SRRQR algorithm * rrqr added to filenemes * max entry id logged * debug print * debug print * prints removed * fixed max rank * fixed max rank * remove inactive indices after batching * alternative * removal of inactive indices undone * debug prints * undo print * remove innecessary conj * conj reg * undo * remove left perm trans * undo * no conj * conj array * fix * u_arr and comp factor public * remove inactive indices in between batches * manage level=1 * joint order implemented * undo remove inactive inds * fixed joint multiplication * remove inactive indices * save samples * use max number of samples * fix min num samples * debug print * debug print * appending data fixed * reduce oversampling * alternative oversampling * full use of samples * fix to save * fixed col major * don't load shape * fix col major rlst way * bug fix * fix corrupted shape * macro implementation for loading samples * small bug fix * reduce local_oversample * max sample for local_oversample * num threads limited * pools of threads * fixes to parallelism * use mis for batch generation * remove inactive batches * disable mis * disable greedy coloring: * remove duplicate definitions * debug prints * remove debug prints * minimum samples added * minimum samples added * disable parallel in id and lu * undo serial batches * modify batching * batching fixed * return to old batching * sort before batching * record update id time * fix non-symmetric lu * record level and mv times * fix non-symmetric sampling * changes to level effort * leaf count added * effective dofs * extra stats * update stats * remove empty boxes * fix tree indexing * undo removal * remove inactive neighbours * remove empty * init * fixed errors * documentation of base factors * more documentation * debug print * transpose direct inversion * modify transpose * non symmetric block extraction * remove trans * add trans * fix save samples * fixes to non symmetric * fix trans for arrayvectorspace * conj right * right * right * fixed * fixed trans product * Precompute fixed rank nsamples and update inactive indices after each color eliminated on a level * fix rust formatting * Install HDF5 in tests and pin cargo-upgrades in CI * Use git crates.io index for cargo-upgrades in CI * Fix CI checks for warnings * Reduce RSRS memory usage and add diagnostics * Record RSRS memory snapshots in statistics * Fix Hermitian matvec products and CI coverage * samples management * add samples incrementally/no sample copying * fixed sampling * Optimize sample storage and restore fixed-rank projection * Restore direct LU extraction for fixed-rank runs * Separate sample loading time and load samples in place * Speed up fixed-rank sample budget estimation * Speed up diagonal extraction for tiny fixed-rank boxes * simplify index calculation for block ext * Prepare ifp_tests changes for enable_symm merge * complex non-symmetric bug fixed * Fix CI style checks * Restore fixed-rank sample budget estimation * Restore per-level fixed-rank activation * force min_num_samples when its more than budget * Fix fixed-rank sample budgeting on adaptive trees The fixed-rank estimators capped carried coarser boxes by the requested rank before those boxes were processed. On adaptive meshes this undercounted the runtime near-field sketch size, so per-level and constant modes could start with too few samples and only discover the larger requirement at a later level, leading to late sample growth and LAPACK/subview failures. Estimate the runtime active-sample upper bound with carried boxes at full size, use it directly for per-level budgets, and floor constant budgets by it. Keep runtime targets from falling below level minima and add a regression test for carried-box capping. * Modified PerLevel and Constant sample use * Box coloring for parallel processing and launch mask * Store lu(X_RR) from joint id lu step for symmetrized X_RR and store factorized LU only --------- Co-authored-by: Ignacia Piccardo <ignacia-fp@eduroam-int-dhcp-97-151-180.ucl.ac.uk> Co-authored-by: Ignacia Piccardo <ignacia-fp@ignacias-air.home> Co-authored-by: Ignacia Piccardo <ignacia-fp@Ignacias-MacBook-Air.local> Co-authored-by: Ignacia Piccardo <ignacia-fp@Mac.lan> Co-authored-by: Ignacia Piccardo <ignacia-fp@138-38-193-234.eduroam.bath.ac.uk> Co-authored-by: Anna Yesypenko <annayesypenko@gmail.com>
This reverts commit 323dddf.
Collaborator
Author
|
Hi @mscroggs, could you help with this? @ignacia-fp and I accidentally used squash merge when we meant to do a regular merge and preserve the branch’s commit history. |
Member
|
@annayesy I reverted the merge into the main branch and force pushed that. I have opened #5 which should make the same changed and set it to auto merge (not squashed) once the tests pass. force pushing is a little dangerous as it can lead to accidentally deleting things, so I made a local backup copy before of the full repo before I did anything. Let me know if anything doesn't look right and I can get files from that backup. In general, I prefer squash merging for most commits as it keeps the main branch history tidy. But for larger changes like #3, it can be preferred not to. I updated the repo settings to allow non-squash merges, but squash merges are still the default. Let me know if you'd like me to change that (after all, this is your and @ignacia-fp's project not mine so I can't impose my preferences on you). |
Collaborator
|
Thank you very much, Matthew. We needed this just for a particular reason,
but it is good to have the option to not squash.
Best,
Ignacia
El lun, 27 abr 2026 a las 15:58, Matthew Scroggs ***@***.***>)
escribió:
… Closed #4 <#4>.
—
Reply to this email directly, view it on GitHub
<#4 (comment)>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/APLJXP54FSJUIWRXOK3MK7D4X5YS3AVCNFSM6AAAAACYFPRCCSVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMRUHEYTCMJQGI4DSMQ>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Collaborator
Author
|
Thank you for helping to resolve this @mscroggs ! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This reverts commit 323dddf.