100 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
andrew	6d8b939a81	Replace CI Docker image with inline apt installs ... Drops the build-image job and the private registry dependency entirely. Each job now installs only the packages it needs, caching /var/cache/apt/archives keyed on the workflow file and ~/.cache/pre-commit keyed on .pre-commit-config.yaml.	2026-06-21 19:21:04 -04:00
andrew	48f9ee46cf	Pin CI steps that use bashisms to bash	2026-06-12 18:05:57 -04:00
andrew	f13a30e8aa	Install biber in the CI image so the paper's bibliography builds	2026-06-12 18:04:49 -04:00
andrew	1394a1a98f	Print test summaries to the step log too ... Gitea doesn't render GITHUB_STEP_SUMMARY yet (https://github.com/go-gitea/gitea/issues/36098), so tee the summary into the log where it's actually visible. Keep writing the summary file so it lights up if Gitea ships the feature.	2026-06-12 17:21:49 -04:00
andrew	c52339a2ba	Add markdown test summaries and a coverage HTML report to CI ... ctest_summary.py renders a Test.xml as markdown for GITHUB_STEP_SUMMARY: a one-liner when everything passes, otherwise the first few failures inline with a link to the full Test.xml on MinIO. It's also usable locally with --all to list every failure from a downloaded Test.xml. The coverage job now also generates and uploads gcovr's html-details report and links it from the step summary.	2026-06-12 17:06:31 -04:00
andrew	3a82d90914	Work around arm64 clang codegen issues breaking CI ... Disable preserve_none under ASan on aarch64: every clang tested (20, 21, trunk 22) miscompiles the continuation chains with that combination, crashing ~97% of fuzz corpus tests. The chains still work with the default calling convention. See #38. Mark the masked Node3 scan results defined for valgrind on aarch64: clang 21+ lowers the in-bounds tests through flags+csel, which memcheck models imprecisely, tainting bits the mask provably clears. See #39. Skip valgrind tests in the arm64 release job, since -DNVALGRIND compiles out the client requests the workaround relies on. Both issues track filing upstream bugs.	2026-06-12 16:35:57 -04:00
andrew	0921d8fedf	Fix arch-specific hardening_check and guard CI package upload globs ... hardening-check tests both CET (x86-only) and branch protection (arm64-only), so pass the ignore flag for whichever doesn't apply to the build arch. The release upload step runs with if: always() so test results are uploaded even when an earlier step fails, but in that case cpack never ran and mc errored on the unmatched package globs. Skip missing packages instead.	2026-06-12 14:45:23 -04:00
andrew	f947d883e7	Only pass -pie when linking executables ... Passing -pie globally made the driver link Scrt1.o into shared libraries, which fails with an undefined reference to main.	2026-06-12 13:58:54 -04:00
andrew	19f430d68f	Add nodejs to the CI image ... The runner executes JavaScript actions (checkout, cache) with node from inside the job container, so custom container images must provide it.	2026-06-12 12:33:48 -04:00
andrew	13e9e88e0e	Pass registry credentials via env instead of script interpolation ... Template interpolation embeds the secret in the generated script file; env indirection keeps it out of argv and off disk.	2026-06-12 12:21:19 -04:00
andrew	12a62a91cf	Track the CI image hash with a label instead of a tag ... Pushing per-hash tags accumulates multi-GB versions in the registry, and package cleanup rules can't distinguish the hash tags that the latest-{arch} tags currently point to from stale ones. Push only latest-{arch} and record the Dockerfile hash as an image label; the skip-rebuild check reads the label back via buildx imagetools.	2026-06-12 12:16:34 -04:00
andrew	cbbb23bf9d	Shrink the CI image ... Replace texlive-full with the texlive packages the paper actually needs (latexmk, latex-extra, pictures for TikZ, bibtex-extra, recommended fonts), and use Ubuntu's valgrind instead of building 3.22 from source - the distro version is newer now. Cuts the image by several GB and removes the slowest step of the image build.	2026-06-12 12:12:10 -04:00
andrew	04d02261e9	Install clang from Ubuntu and push image from buildx ... apt.llvm.org has no repository for the current ubuntu:rolling release, and nothing pins clang 20 - CC=clang already resolved to the distro clang. Drop the llvm.sh step along with its helper packages. docker build now runs under the buildx docker-container driver, which keeps the result in the build cache unless told otherwise, so push directly from the build instead of tagging locally.	2026-06-12 12:06:45 -04:00
andrew	dd8f006d3f	Remove docker socket debug step and unneeded sudo ... Jobs run as root in the job container; the earlier permission errors were SELinux denials on the mounted podman socket, fixed in the runner config with --security-opt label=disable.	2026-06-12 11:58:30 -04:00
andrew	776d06963b	Add temporary docker socket debug step	2026-06-12 11:52:51 -04:00
andrew	e03afe0651	Run docker with sudo in build-image ... The default job image runs as a non-root user that lacks access to the mounted docker socket.	2026-06-12 11:30:28 -04:00
andrew	6eecf6e4ac	Use a registry bot account for the container registry ... Gitea's ephemeral Actions token is not accepted by the container registry, so docker login and image pulls use REGISTRY_USER / REGISTRY_TOKEN secrets (a dedicated low-privilege account and its personal access token with package read/write scope) instead.	2026-06-12 11:16:33 -04:00
andrew	d74a12b5a4	Convert Jenkins CI to Gitea Actions ... Replace the Jenkinsfile with .gitea/workflows/ci.yml. The CI image is built from the Dockerfile and pushed to the Gitea container registry, rebuilt only when the Dockerfile or pre-commit config changes. The aarch64 release build now runs natively on an arm64 runner instead of cross-compiling, so the cross toolchain and qemu are dropped from the image. Artifacts still go to MinIO (via mc, skipped gracefully if credentials are not configured); ccache uses actions/cache instead of a host volume.	2026-06-12 11:13:32 -04:00
andrew	755bbcbe56	Fix gcc 15 build	2025-11-06 13:12:10 -05:00
andrew	2642d453dc	Remove unnecessary branch for interleaved range writes	2024-12-11 21:53:44 -08:00
andrew	7166811387	Fix condition checking for erasing the root	2024-11-21 17:49:54 -08:00
andrew	d68f208d9b	Add to corpus	2024-11-21 17:27:14 -08:00
andrew	81323972aa	Remove "memcpy common section to next node" ... Do it in a simpler, more robust/type safe way	2024-11-21 16:47:05 -08:00
andrew	8694ba8b6a	Remove unused field	2024-11-21 16:35:10 -08:00
andrew	0cea5565b5	Add artificial root parent and stop plumbing impl everywhere	2024-11-21 16:32:13 -08:00
andrew	972f16ed8f	Remove Node::endOfRange ... This should simplify storing leaves directly	2024-11-21 11:55:30 -08:00
andrew	2412684316	childMaxVersion memory needs to be defined ... Also remove some dead code, and only memset children to 0 for Node256. Other nodes detect absence of children other ways.	2024-11-21 10:35:05 -08:00
andrew	8190d2f24e	Update README	2024-11-20 21:55:37 -08:00
andrew	8251631087	Fix some unintentional generic usages of getChildAndMaxVersion	2024-11-20 21:50:48 -08:00
andrew	90fb2a9542	Remove some usages of generic getFirstChild	2024-11-20 21:43:47 -08:00
andrew	7c01f8ba0f	Remove some usages of maxVersion	2024-11-20 21:24:59 -08:00
andrew	0df2db7f8a	Remove some bad unlikely annotations found by -Wmisexpect ... Using a profile from server_bench	2024-11-20 19:06:50 -08:00
andrew	5e975f3b2b	More writes than reads in ServerBench	2024-11-20 17:49:55 -08:00
andrew	bcbae026b2	Update README	2024-11-20 14:53:57 -08:00
andrew	e125b599b5	Remove freeList, min/max capacity tracking ... The freelist doesn't seem to get a good hit rate. Policies other than capacity = minCapacity did not improve the rate we were resizing nodes, but did increase memory usage, so get rid of that too. Add a nodes_resized_total counter.	2024-11-20 14:45:56 -08:00
andrew	3f4d3b685a	More valgrind annotations	2024-11-20 13:36:30 -08:00
andrew	4198b8b090	Some prep for leafs-in-parents	2024-11-20 12:20:11 -08:00
andrew	8757d2387c	Call prefetch within TaggedNodePointer::getType ... Instead of at every call site	2024-11-20 12:07:32 -08:00
andrew	4a22b95d53	Remove state machine transitions "from" Node0 ... Those aren't used	2024-11-20 11:46:45 -08:00
andrew	03d6c7e471	Allocate maxCapacity instead of minCapacity	2024-11-19 16:13:57 -08:00
andrew	ceecc62a63	Disable freelist for macos ... It's faster on my mac m1 without the freelist	2024-11-19 14:28:15 -08:00
andrew	80f0697e79	Fix arm detection on macos	2024-11-19 13:24:13 -08:00
andrew	ce23d3995c	Fix README	2024-11-19 09:02:16 -08:00
andrew	6f899e063b	Update readme	2024-11-18 14:31:56 -08:00
andrew	e5b9c03e77	Restore change that hurt codegen ... I don't understand, but this is necessary for good codegen somehow	2024-11-18 14:29:55 -08:00
andrew	a158d375f5	Add script for updating readme	2024-11-18 14:05:14 -08:00
andrew	ee5a84cd7b	Remove dead stores	2024-11-15 17:03:29 -08:00
andrew	33f14e3d9b	Remove unused header	2024-11-15 16:49:21 -08:00
andrew	77262ee2d3	Fix some grammar in a comment	2024-11-15 16:47:31 -08:00
andrew	9945998e05	Remove unused code in Internal.h	2024-11-15 16:30:20 -08:00
andrew	2777e016ff	Be more consistent about TaggedNodePointer vs Node*	2024-11-15 16:30:05 -08:00
andrew	661ffcd843	Explain purpose for prefetches in getFirstChild	2024-11-15 16:29:42 -08:00
andrew	3a34d3cecb	Minor style improvements	2024-11-15 16:29:17 -08:00
andrew	189c73e3bd	Reduce Node3 size by 8	2024-11-15 15:53:18 -08:00
andrew	35987030fc	Add to corpus	2024-11-15 15:47:32 -08:00
andrew	0621741ec3	Update README benchmarks	2024-11-15 13:12:47 -08:00
andrew	f5ec9f726a	Remove getFirstChildExists ... Once we know the type, for Node3 and higher we know the first child exists anyway	2024-11-15 13:05:11 -08:00
andrew	552fc11c5d	Prefetch second child to improve scan performance	2024-11-15 12:45:01 -08:00
andrew	71ace9cc55	Make more range reads commit in server_bench	2024-11-15 11:35:55 -08:00
andrew	bcf459304f	Improve cpu performance for workload generation in server_bench	2024-11-14 17:16:57 -08:00
andrew	f403c78410	Add test that exercises "too large removal buffer" code	2024-11-14 16:40:47 -08:00
andrew	08958d4109	Remove the reverse step for saving scan search path	2024-11-14 16:27:25 -08:00
andrew	dcc5275ec9	Isolate conflict-set on one thread in server_bench	2024-11-13 22:46:11 -08:00
andrew	c5ef843f9e	Add range reads to server_bench	2024-11-13 14:06:50 -08:00
andrew	b78e817e24	Only set HAS_AVX for x86_64	2024-11-12 18:16:30 -08:00
andrew	9c82f17e20	Enable valgrind annotations for gcc build in jenkins	2024-11-12 18:13:32 -08:00
andrew	665a9313a4	Valgrind annotations for new free list	2024-11-12 18:11:09 -08:00
andrew	6e66202d5e	Add to corpus	2024-11-12 17:50:27 -08:00
andrew	a92271a205	Build with -Wunused-variable	2024-11-12 17:50:27 -08:00
andrew	0dbfb4deae	Detect simd headers in c++ instead of cmake	2024-11-12 17:50:27 -08:00
andrew	6e229b6b36	Skip sorting if already sorted for skip list	2024-11-12 17:50:27 -08:00
andrew	2200de11c8	Point read/write workload for server_bench	2024-11-12 17:50:27 -08:00
andrew	b37feb58dd	Require musttail and preserve_none for interleaved	2024-11-12 17:50:27 -08:00
andrew	94a4802824	Don't do hardening check if cross compiling	2024-11-12 17:50:27 -08:00
andrew	707dbdb391	Build binaries compatible with cf-protection	2024-11-12 17:50:27 -08:00
andrew	bdd343bb57	Use llvm-objcopy if using clang and it's available ... This works around a weird error I was getting when trying to link a translation unit that included Internal.h with libconflict-set-static.a with clang + gnu objcopy	2024-11-12 17:50:27 -08:00
andrew	7b31bd5efe	Use llvm 19 for macos package	2024-11-10 21:47:26 -08:00
andrew	e255e1a926	Add missing symbol imports for macos	2024-11-10 21:05:42 -08:00
andrew	f85b92f8db	Improve next{Physical,Logical} codegen	2024-11-09 13:21:36 -08:00
andrew	3c44614311	Allocate from freelist with min/max capacity constraints	2024-11-08 21:35:13 -08:00
andrew	9c1ac3702e	Move index closer to start of Node{3,16} ... This should slightly improve cache hit rate	2024-11-08 20:54:00 -08:00
andrew	224d21648a	Make server_bench workload point writes of tuple-encoded keys	2024-11-08 20:50:54 -08:00
andrew	33f9c89328	Try reduceLog: true	2024-11-04 17:03:24 -08:00
andrew	12c2d5eb95	Try not logging ctest output to stdout	2024-11-04 15:36:50 -08:00
andrew	db357e747d	Add __mem{cpy,set}_chk to allowed symbol imports	2024-11-04 15:08:25 -08:00
andrew	4494359ca2	Make insert_iterations for interleaved writes more closely match	2024-11-04 14:46:30 -08:00
andrew	f079d84bda	Update README	2024-11-04 14:33:43 -08:00
andrew	724ec09248	Add to corpus	2024-11-04 14:30:12 -08:00
andrew	4eaad39294	Maintain capacity invariant strictly	2024-11-04 13:43:02 -08:00
andrew	891100e649	Add to corpus	2024-11-01 21:32:00 -07:00
andrew	22e55309be	Update benchmarks	2024-11-01 17:50:03 -07:00
andrew	d6269c5b7c	Use interleaved even if preserve_none is missing	2024-11-01 15:12:59 -07:00
andrew	faacdff2d9	Add to corpus	2024-11-01 14:11:43 -07:00
andrew	821179b8de	Remove dead code	2024-11-01 13:48:32 -07:00
andrew	681a961289	Dispatch on type pairs for end iter	2024-11-01 13:45:41 -07:00
andrew	c73a3da14c	Dispatch on type pairs for begin iter	2024-11-01 12:46:36 -07:00
andrew	5153d25cce	Dispatch on type pairs for common prefix iter	2024-11-01 12:06:48 -07:00
andrew	d2ec4e7fae	Fix use of wrong member in endIter	2024-11-01 11:17:26 -07:00
andrew	c7e2358746	Add memcmp to aarch64 imports	2024-10-31 15:36:19 -07:00
andrew	ec1c1cf43f	Run gc less frequently ... But still run it to account for the version potentially increasing	2024-10-31 15:35:00 -07:00