Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Buffer Manager] Create volatile region for virtual memory management #2624

Open
wants to merge 98 commits into
base: master
Choose a base branch
from

Conversation

nikriek
Copy link
Contributor

@nikriek nikriek commented Oct 25, 2023

This PR depends on #2618. VolatileRegion is used to create and manage the virtual memory region using various syscall e.g free(PageID) to call madvise(MADV_DONTNEED).

nikriek and others added 30 commits October 12, 2023 11:34
Implements a pattern that removes the DISTINCT clause if the result is unique by definition or previous GROUP BY clauses.
Integrate the Star Schema Benchmark (SSB) as a Hyrise benchmark
executable
…ts (#2536)

Estimate query plans rewritten by the `JoinToPredicateRewriteRule` as if they were still (semi-) joins to place the rwritten predicates correctly in the query plans.
Treat uncorrelated subqueries just like normal inputs in Optimizer::validate_lqp(...)`.
Ensure that the TPC-DS and SSB benchmark binaries crash when invoked with unfitting SFs and `--verify`.
#2571)

Handle `SELECT DISTINCT` statements correctly if there are further operations on the selected columns.
This PR uses a `std::stringstream` to encapsulate writing the
default-type encoding message to `std::cout` to prevent messages from
multiple threads being mangled.
Small changes to support builds with clang 16.
…#2577)

Usage of linker LLD is not required anymore, but a warning is shown if the default linker is used on Linux. Added the option to use another linker by specifying `HYRISE_LINKER` (for example `mold`) - `CMAKE_LINKER` was actually already set to `/bin/ld`.
Adding a test that recursively adds tasks to the scheduler (a simple merge sort).

This test triggered a minor issue when shutting down the scheduler. This
issue should be fixed with the PR.
This PR prepares the codebase for subquery pruning.
Simply replaces emplace_back(RowID{a,b}) with emplace_back(a,b)

Moreover, we increase the timeouts for the Python `pexpect` scripts as they often timed out when the load on the CI server was high.
Updates the `plot_performance_breakdown.py` script to get rid of some warnings, generated PDF instead of PNG and improve legend placement.
Add tests for  `Difference` operator and the `<<` operator of the `OperatorScanPredicate`.

---------

Co-authored-by: Tarik Alnawa <43680469+T4rikA@users.noreply.github.com>
Co-authored-by: Paula <67064988+postmartem@users.noreply.github.com>
Co-authored-by: Tarik Alnawa <talnawa@outlook.de>
Co-authored-by: Bouncner <martin+github@boissier.de>
Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
Add tests for LQP nodes and operators.

Co-authored-by: Finn Schoellkopf <schoellkopf.finn@gmail.com>
Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
Bouncner and others added 28 commits November 2, 2023 20:36
Simply replaces emplace_back(RowID{a,b}) with emplace_back(a,b)

Moreover, we increase the timeouts for the Python `pexpect` scripts as they often timed out when the load on the CI server was high.
Updates the `plot_performance_breakdown.py` script to get rid of some warnings, generated PDF instead of PNG and improve legend placement.
Add tests for  `Difference` operator and the `<<` operator of the `OperatorScanPredicate`.

---------

Co-authored-by: Tarik Alnawa <43680469+T4rikA@users.noreply.github.com>
Co-authored-by: Paula <67064988+postmartem@users.noreply.github.com>
Co-authored-by: Tarik Alnawa <talnawa@outlook.de>
Co-authored-by: Bouncner <martin+github@boissier.de>
Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
Add tests for LQP nodes and operators.

Co-authored-by: Finn Schoellkopf <schoellkopf.finn@gmail.com>
Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
Add tests for table and dummy table node.
Co-authored-by: Fabian Mahling <FabianMahling@web.de>
Co-authored-by: rothaarlappen <roessler.paul@web.de>
Co-authored-by: Fabian Mahling <45626086+Tratori@users.noreply.github.com>
Co-authored-by: Bouncner <martin+github@boissier.de>
Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
This commit tells Jenkins to repeat the x86 macos step several times before failing.
Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
Co-authored-by: Bouncner <martin+github@boissier.de>
Enables Apple's clang version as it has sufficiently caught up with LLVM.
Fix CI issues with concurrent SSB data generation and `StoredtableNode`'s cached output expressions.

Co-authored-by: Bouncner <martin+github@boissier.de>
Add the representation of SQL:2003 window functions in the LQP, including the translation in the SQLTranslator.
This PR adjusts the `IndexScan` to use the new `TableIndexes` introduced
in PR #2448 .

Co-authored-by: martin.boissier <martin.boissier@hpi.de>
Co-authored-by: Bouncner <martin+github@boissier.de>
…eping (#2548)

Using a new concurrent queue in the NodeQueueScheduler and semaphores instead of conditionals that workers wait on (less awaking, lower CPU usage when idling).
Replace the manually created `IN` list with a join, see #2602.
Sets the maximal begin and end commit ID whenever an Insert or Delete operator commits.
…2606)

Fixes performance and optimizer breakdown plots for cached queries.
Prunes tables using Predicates that contain uncorrelated subquery results. The information about such Predicates is added to StoredTableNodes by the ChunkPruningRule and later to GetTable operators by the LQPTranslator. Deep copies of LQPs and PQPs preserve the information.
Ensure C2x-compatibility, which clang-16+ warns about.
Replace TPC-DS with TPC-C, hoping for faster generation times without timeouts. Also, add pruning with between exclusive predicates to RangeFilter.
No longer create TaskQueues for NUMA nodes without any workers and make scheduler shutdown mechanism more fail safe.

---------

Co-authored-by: Daniel Lindner <27929897+dey4ss@users.noreply.github.com>
@nikriek nikriek changed the title [Buffer Manager] Create volatile region [Buffer Manager] Create volatile region for virtual memory management Nov 6, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

9 participants