Add 32bit support for GreenFunctions #238
Merged
Conversation
This file contains 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
pablosanjose
force-pushed
the
typestable_g
branch
from
f75631c to
3aaa3b8
Compare
fix omega coversion cleanup sanitize_eigen also on vectors; allow non-BlasComplex test new deterministic dual phi algorithm tighten version for 32-bit test skip skip ci cache for the moment ci fix? remove ubuntu-latest x86 from ci matrix exclude x86 altogether (broken on nightly due to new Memory) ComplexF32 fixes to rest of GreenSolvers type-stability fixes fix graphene model fix graphene (redo)
pablosanjose
force-pushed
the
typestable_g
branch
from
8c18a89 to
ba36bed
Compare
Codecov ReportAttention:
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master JuliaLang/julia#238 +/- ##
==========================================
- Coverage 92.78% 92.66% -0.12%
==========================================
Files 34 34
Lines 5500 5538 +38
==========================================
+ Hits 5103 5132 +29
- Misses 397 406 +9 ☔ View full report in Codecov by Sentry. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Closes JuliaLang/julia#236
Although Quantica is coded for a generic
T<:AbstractFloat, the GreenFunction part was not tested forT == Float32, only forT == Float64. Doing so revealed a number of issues, which are fixed in this PRGreenSolvers.SparseLU, which relies on UMFPACK, does not supportComplexF32eltypes yet, so we must internally work withComplexF64, taking care to returnComplexF32if needed without making extra copies. Tim Davis mentioned on slack thatComplexF32support is in the short-term to-do list for UMFPACK. At that point, the corresponding part of this PR could be simplified and generalized.GreenSolvers.Schurbuilds onSparseLUfor some of its internals, so it needs adjustmentsGreenSolvers.Bandsneeded a conversion of the Brillouin zone mesh to the Hamiltonian eltype to properly propagate the ComplexF32 typeGreenSolvers.Bandsrelied on a random choice of dual numbers for degeneracy resolution. This randomness exactly cancelled out in all our tests at 64bit precision, so the results were deterministic. But on 32bit, the lower precision made the cancellation inexact, and the result would vary from run to run, or even yield divide-by-zero errors. In collaboration with Adrián Prada we devised a new algorithm to choose the optimal dual number that is deterministic and retains maximal precision