Update GNU assembler bug detection #278
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
|
@manodeep This is also ready for review (and should be quick). The failure seems to be due to astropy bot not picking up the milestone. |
|
Yup - looks good. Checked on my AVX512 cluster with the old binutils and AVX512 got disabled, while with the new binutils AVX512 remained enabled. |
manodeep
approved these changes
Sep 15, 2022
lgarrison
added a commit
that referenced
this pull request
Oct 7, 2022
* Attempt to enable Rmax comparable to half boxsize by removing (unnecessary?) duplicate cell checks * Comment out unused var * Add test implementing Manodeep's example * Restore duplicate cell pair check, but only count a pair as a duplicate if the wrap value is identical. * Add pragmas for CI (will revisit) * Update GNU assembler bug detection (#278) * Update GNU assembler bug detection * Cosmetic enhancement to suppress spurious warnings during GAS bug test * Fix test error code * Add another test of large Rmax, comparing against brute-force * Add const qualifiers * Apply @manodeep's fix for large Rmax test against brute-force, and require nmesh>=2 to avoid duplicate cell pairs * Make boxsize non-trivial in test. Remove extra print statement. * Add comments on array broadcasting to test * Changed variable name for clarity Co-authored-by: Manodeep Sinha <manodeep@gmail.com>
manodeep
added a commit
that referenced
this pull request
Dec 7, 2022
* Implement non-cubic periodic box for theory.DD * Use boxsize tuple when calling low-level python bindings * Changelog * pep8 * Fix gridlink for non-cubic * Implement turning off periodicity per-dimension by specifying boxsize=-1 in that dimension * Change binsize calculation to use particle extent, not periodic wrap * Port per-dimension periodicity and boxsize to other modules * Fix calling CPython modules * Extend simple boxsize tests to other modules * pep8 * Only apply minimum cell criterion in a dimension when the periodic wrap makes wrap-crossing pairs possible * Assign particles to cell 0 when they are all in a plane * Add warning when all particles fall in a plane * Add comments and parens. More explicit particle positions in narrow extent test. * WIP: enable larger Rmax (up to half the box size) (#277) * Attempt to enable Rmax comparable to half boxsize by removing (unnecessary?) duplicate cell checks * Comment out unused var * Add test implementing Manodeep's example * Restore duplicate cell pair check, but only count a pair as a duplicate if the wrap value is identical. * Add pragmas for CI (will revisit) * Update GNU assembler bug detection (#278) * Update GNU assembler bug detection * Cosmetic enhancement to suppress spurious warnings during GAS bug test * Fix test error code * Add another test of large Rmax, comparing against brute-force * Add const qualifiers * Apply @manodeep's fix for large Rmax test against brute-force, and require nmesh>=2 to avoid duplicate cell pairs * Make boxsize non-trivial in test. Remove extra print statement. * Add comments on array broadcasting to test * Changed variable name for clarity Co-authored-by: Manodeep Sinha <manodeep@gmail.com> * Fix missing docstring escapes * Add tests for anisotropic boxes against brute-force. Allow more boxsize arg types, and tests for those arg types. Remove old large Rmax test. * Fix Rmax>L/2 for non-periodic. Greatly expand brute-force tests. Fix passing boxsize=None to the Python API. Co-authored-by: Manodeep Sinha <manodeep@gmail.com>
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.
The fix for the GNU assembler bug (#196, https://sourceware.org/bugzilla/show_bug.cgi?id=23465) has been backported (https://bugzilla.redhat.com/show_bug.cgi?id=1869401), so we shouldn't rely on version numbers to detect the presence of the bug. Instead, we should just assemble a micro-program (
vmovaps 64(,%rax), %zmm0) and check that it produces the correct result.This will re-enable Corrfunc AVX-512 on some platforms (notably, my desktop at work!).