C
Hi Alex — I didn't see your response earlier, but I reached the same conclusion about a half hour after you, and I can now explain what's going wrong. Happily (for me, as the implementor of the swap), it's not a bug in the swap, but rather a subtle issue in the user code—though I'm wondering whether perhaps we could've done more to protect you. It really is a subtle case. Consider the following lines, running on a 10x10 array mapped to 2 locales, such that locale 0 owns {1..5, 1..10} and locale…
C
Branch: rewrite-printchplenv-help Revision: 764531e4c479a0ede5362c3a34b8b7ecf13dca2a Author: jabraham17 Merged at: 2026-11-03T02:01:51Z Link: Rewrite printchplenv help by jabraham17 · Pull Request #29598 · chapel-lang/chapel · GitHub Log Message: Rewrite printchplenv help Rewrites the printchplenv help to take advantage of argparse This has the advantage of keeping the help text and the arguments in the same place, rather than at different points in the file (annoying maintenance burden)…
C
Hi Brad, The problem also goes away if I use a constant boundary, e.g. T[i,cols+1] = 80.0; instead of a varying one T[i,cols+1] = 80.0*i/rows; so it seems that the bug might be related to indexing of ghost cells. Thanks, Alex.
C
Branch: more-pylints Revision: 4585670013e1be1976ccf1d6429751e8faf87c9e Author: jabraham17 Merged at: 2026-11-03T00:59:46Z Link: Resolve a few more pylints in chplenv by jabraham17 · Pull Request #29637 · chapel-lang/chapel · GitHub Log Message: Resolve a few more pylints in chplenv Fixes a few more pylints in chplenv, mostly related to the linter seeing variables "used before defined", although none of these cases could happen in practice (without --ignore-errors) [Reviewed by @benharsh]…
C
Alex — Interesting, thanks for sharing your code. That definitely looks like a bug to me. Interestingly, it seems to work better with some locale counts than others. -nl 2, 3, and 5 all produce bad results for me, but others work fine. Taking a deeper look. -Brad
C
The following GitHub Actions jobs failed: full-checks / arkouda-main-smoke-test / arkouda-smoke-test, step: "install Arkouda deps (retrying)" Ran 805 total jobs at commit e33f106aa85f306c283ddf6f5df1b2111c147b9a "More CI fixes after update to Ubuntu 26 (#29624)". GitHub Actions run info: SHA: e33f106aa85f306c283ddf6f5df1b2111c147b9a Run ID: 41285216498 Run attempt: 1 Run number: 108 Triggering event name: schedule
C
Hi Alex — Huh… I would not have expected the swap operator to change the behavior relative to your version using whole-array assignments back. There could be a bug in its implementation, but just to clarify its behavior before assuming the worst: The swap operator is not defined in the language as a pointer swap, but rather as a swap via assignments. So T = Tnew should be equivalent to const Temp = T; T = Tnew; Tnew = Temp; or forall (t1,t2) in zip(T, Tnew) { const temp = t1; t1 = t2; t2 =…
C
The following GitHub Actions jobs failed: full-checks / arkouda-main-smoke-test / arkouda-smoke-test, step: "install Arkouda deps (retrying)" full-checks / portability-tests / portability-tests-apptainer (almalinux-8), step: "build Apptainer image for config almalinux-8 (retrying)" full-checks / Chapel paratest with GASNet and AddressSanitizer / Test test/runtime +13 more (gasnet-asan), step: "run tests" full-checks / Chapel paratest on darwin mulitilocale tests with gasnet (segment everything)…
C
Hi Brad, I tried the pointer swap T <=> Tnew and -- of all approaches -- it indeed produces the fastest solution: with BlockDist: At 50000x50000 after 10 iterations reached delta=1.94288 in 0.419369 seconds At 100000x100000 after 10 iterations reached delta=1.94296 in 1.41929 seconds At 200000x200000 after 10 iterations reached delta=1.943 in 5.38291 seconds with StencilDist: At 50000x50000 after 10 iterations reached delta=5.48573 in 0.16537 seconds At 100000x100000 after 10 iterations reached…
C
Hi Damian — Just came back across this. I didn't answer it immediately because I wasn't aware of anything but thought I'd see if others answered first. But I think no answer at this point means "not that anyone following is aware of." If you end up giving it a try, let us know what you find. Have a good weekend, -Brad
C
@razoumov : My subconscious must've still been chewing on this topic this morning, because in the wee hours, I remembered something else you should try: I believe that if, rather than assigning one array back to the other, you use Chapel's swap operator, <=>, to swap the two arrays, it should be implemented as pointer swaps rather than deep copies of the array since they share the same domain. This should get you similar performance as the unroll-by-two approach I suggested above, but without…
C
Branch: more-ci-fixes Revision: e33f106aa85f306c283ddf6f5df1b2111c147b9a Author: jabraham17 Merged at: 2026-10-30T21:35:20Z Link: More CI fixes after update to Ubuntu 26 by jabraham17 · Pull Request #29624 · chapel-lang/chapel · GitHub Log Message: More CI fixes after update to Ubuntu 26 Builds on Fix Docker CI image by jabraham17 · Pull Request #29623 · chapel-lang/chapel · GitHub Adds more CI fixes caused by Ubuntu 26 breaking some CI checks fixes an issue in a script where we relied on…
C
Jade noticed we didn't originally get a mail sent out when this merged yesterday. We aren't sure why, but it looks like none of the workflows that should have been triggered on a push to main were triggered (lints etc), so the problem seems to be on GitHub's side. I took this opportunity to test out the manual dispatch mode I added to the emailer workflow in Handle push of simultaneous PR merges in emailer by arifthpe · Pull Request #29555 · chapel-lang/chapel · GitHub, which resulted in this…
C
Branch: improve-email Revision: 7d9f11a843d8876e24d24c411056fe8424716ef5 Author: jabraham17 Merged at: 2026-10-30T01:56:20Z Link: Improve paratest output and the nightly test emailer by jabraham17 · Pull Request #29618 · chapel-lang/chapel · GitHub Log Message: Improve paratest output and the nightly test emailer Improves paratest output and the nightly test emailer to be more useful with less clicking Improves the paratest output to directly show the failing test output. Prior to this, you…
C
Yep! [filler text for Discourse character minimum :)]
C
Ooh! Fancy upgrade to the mail output, very cool! Was this due to Jade's #29618? -Brad
C
All of these failures will be fixed by More CI fixes after update to Ubuntu 26 by jabraham17 · Pull Request #29624 · chapel-lang/chapel · GitHub, except the arkouda main test, which will require arkouda changes. See More CI fixes after update to Ubuntu 26 by jabraham17 · Pull Request #29624 · chapel-lang/chapel · GitHub for details -Jade
C
The following GitHub Actions jobs failed: core-checks / check_dyno_linters, step: "run dyno linters" extended-checks / arkouda-release-smoke-test / arkouda-smoke-test, step: "install Arkouda deps (retrying)" full-checks / arkouda-main-smoke-test / arkouda-smoke-test, step: "install Arkouda deps (retrying)" full-checks / Chapel paratest on linux64 aarch64 / Aggregate results (linux64-aarch64), step: "aggregate and report results" 10 failed test(s) in linux64-aarch64: [Error matching compiler…
C
Branch: fix-ci2 Revision: fabe2965630d5750df2eb1333ae36bcbcd084077 Author: jabraham17 Merged at: 2026-10-29T22:36:20Z Link: Fix Docker CI image by jabraham17 · Pull Request #29623 · chapel-lang/chapel · GitHub Log Message: Fix Docker CI image Fixes failures caused by update to use Ubuntu 26 in CI image The issue was caused by other packages bringing partial installs of GCC 16 (default is 15), so clang would try and use GCC 16. But, without a full C++ GCC 16 install, clang would be broken and…
C
Hi Brad, You are absolutely right: it is the copy-back with large array slices T[1..rows,1..cols] = Tnew[1..rows,1..cols]; that was slowing down the code! Following your comment, I first tried T = Tnew; (without modifying the boundary, that gives a different solution for BlockDist, but that's easy to correct) and the timings went down to: with BlockDist: At 50000x50000 after 10 iterations reached delta=1.33821 in 0.539635 0.529417 seconds At 100000x100000 after 10 iterations reached…
C
Branch: gpu-export Revision: a7d2bd658749bc553b8698e87bf0e457ba31182d Author: benharsh Merged at: 2026-10-29T19:40:37Z Link: Do not create GPU specializations for exported functions by benharsh · Pull Request #29616 · chapel-lang/chapel · GitHub Log Message: Do not create GPU specializations for exported functions Exported functions are intended to be called from external code, and until we decide on how to rename specializations, for clarity's sake we should not make specialized GPU copies. I…
C
Hi Alex — At a glance, the use of the stencil distribution looks appropriate to me, and I would agree that the delta between the block and stencil distributions' performance seems surprisingly small. I hope to do some more digging, but thought I'd at least confirm that much for now before lunch and some other things I should be prioritizing. One thing I wonder about is the copy-back using array slices within the loop–specifically whether that's bogging down the computation in some way, and…
C
Branch: resolve-pylint Revision: 6e99015823e19ff790c1f2c988e509eddc522352 Author: jabraham17 Merged at: 2026-10-29T19:12:19Z Link: Resolve many pylint issues by jabraham17 · Pull Request #29591 · chapel-lang/chapel · GitHub Log Message: Resolve many pylint issues Resolves the pylint issues revealed by add CI to check that basic python scripts will work with python 3.5 by jabraham17 · Pull Request #29590 · chapel-lang/chapel · GitHub Many of the changes in this PR are pretty superfluous and just…
C
Branch: use-latest Revision: 4114e8fc56cb6b45ee4c0661e3146755b6d6c7b7 Author: jabraham17 Merged at: 2026-10-29T19:06:02Z Link: Use ubuntu:latest for the CI image by jabraham17 · Pull Request #29556 · chapel-lang/chapel · GitHub Log Message: Use ubuntu:latest for the CI image Uses ubuntu:latest for the CI image [Reviewed by @arifthpe] Compare: Use ubuntu:latest for the CI image by jabraham17 · Pull Request #29556 · chapel-lang/chapel · GitHub Diff: M util/packaging/docker/github-ci/Dockerfile…
C
29619, "bradcray", ""Play" button in Chapel Codespaces runs previous binary on failed compilation", "2026-10-29T18:30:47Z" Open issue chapel-lang/chapel "Play" button in Chapel Codespaces runs previous binary on failed compilation opened 06:30PM - 29 Oct 26 UTC bradcray type: Bug area: Tools While using Chapel in a GitHub Codespace today, I found that if I did the follow…ing actions: * launch a codespace * use the "Play" button to compile and run a correct program * introduce a bug into the…
C
Branch: add-noRegex Revision: 07c99946d7ae7a4ef62186921d08b85998cffa4e Author: jabraham17 Merged at: 2026-10-29T18:15:44Z Link: Compile Socket with noRegex to work on quickstart by jabraham17 · Pull Request #29615 · chapel-lang/chapel · GitHub Log Message: Compile Socket with noRegex to work on quickstart Adds -sonoRegex to Socket tests so they work with quickstart. While there, I also added missing test files for ipv6 Socket tests (I am assuming these tests are skipped in places where they are…
C
Branch: add-more-tests Revision: 16abeb570658660067faf45b0839cefb80139bac Author: jabraham17 Merged at: 2026-10-29T18:10:08Z Link: Add many more CI tests + fixes to make them work by jabraham17 · Pull Request #29574 · chapel-lang/chapel · GitHub Log Message: Add many more CI tests + fixes to make them work Adds many more CI tests expands Python interop tests to use LIB_PIC=pic and also test gasnet adds CHPL_ATOMICS tests adds --incremental tests adds mpicc tests refactors logs for GCC and LLVM…
C
Branch: freebsd-disable-caching Revision: 9d1318171c4058a509ea96e915c4df56a6e8bad5 Author: arifthpe Merged at: 2026-10-29T17:54:53Z Link: Disable caching in the FreeBSD VM job by arifthpe · Pull Request #29617 · chapel-lang/chapel · GitHub Log Message: Disable caching in the FreeBSD VM job Free up some cache space by disabling the default caching in the FreeBSD VM (GitHub - vmactions/freebsd-vm: Test your code in a real FreeBSD virtual machine on GitHub Actions -- QEMU-powered, with SSH and…
C
The gasnet-fast failed tests are functions/deitz/iterators/leader_follower/test_leader_follower5 and functions/deitz/iterators/leader_follower/test_recursive_leader3. The first failed with no output, the second with *** FATAL ERROR: Master got an xSocket: error in sendAll() - connection closed (32:Broken pipe). * The quickstart failures are from the new Socket tests, pending fix in Compile Socket with noRegex to work on quickstart by jabraham17 · Pull Request #29615 · chapel-lang/chapel ·…
C
The following GitHub Actions jobs failed: full-checks / Chapel paratest with GASNet fast segment / Aggregate results (gasnet-fast), step: "aggregate and report results" full-checks / Chapel paratest with quickstart configuration / Aggregate results (quickstart), step: "aggregate and report results" Ran 786 total jobs at commit 9fb570b56650af81ae07d89b26538a8d08c3b45a "Fix bad extern calls in Socket (#29614)". GitHub Actions run info: SHA: 9fb570b56650af81ae07d89b26538a8d08c3b45a Run ID:…