Klibs (x86_64): add User-Managed Concurrency Groups #1996
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
This change moves the initialization of global Make variables from rules.mk to a new vars.mk file. This allows specifying that a klib or a test should be built or run only in certain environments or configurations (for example, only for a given architecture). The RUNTIME_TESTS target in the main Makefile has been modified by adding the missing symlink program and making the vsyscall program conditional to the x86_64 architecture.
The existing code in thread_wakeup() assumes that this function is only called from the syscall context associated to the thread to be woken up. This change allows this function to be called from any context; this will allow implementing kernel features that can be used for user-mode threading.
This new header file contains various utilities for running test cases. Duplicated code across the test suite has been removed in favor of using the code in this header file.
francescolavra
force-pushed
the
feature/umcg
branch
from
be7ba1f
to
30d71cf
Compare
This changeset adds a new klib that implements UMCG (the Linux kernel feature proposed by Google in order to support user-mode threading). The implementation in Nanos is compatible with the Linux kernel code at https://www.spinics.net/lists/kernel/msg4740112.html. A new runtime test has been added to exercise this feature. Beside various test cases, this program can be run in demo mode (with the `-d` command line option), which is inspired by the code at https://lore.kernel.org/lkml/[email protected]/. In addition, it contains a set of performance tests with which the performance of thread switching with UMCG can be evaluated and compared with traditional threading implementations: - '-u' option: UMCG-based performance test - '-f' option: futex-based performance test - '-c' option: condition variable-based performance test The performance test creates a worker thread and continuously switches between the main thread and the worker thread, measuring how many main->worker->main thread switching cycles can be executed per unit of time. The test by default runs for 10 seconds, but the duration can be modified with the `-t` command line option. Performance tests run in a KVM instance with a single vCPU showed the following figures: - UMCG: 209995 cycles/sec - futex: 154419 cycles/sec - condition variable: 100965 cycles/sec In the generic Unix code, a new blockq_init() function has been added in order to be able to use blockq structs embedded in other structs.
francescolavra
force-pushed
the
feature/umcg
branch
from
30d71cf
to
11ce06b
Compare
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.
This changeset adds a new klib that implements UMCG (the Linux kernel feature proposed by Google in order to support user-mode threading).
The implementation in Nanos is compatible with the Linux kernel code at https://www.spinics.net/lists/kernel/msg4740112.html.
A new runtime test has been added to exercise this feature. Beside various test cases, this program can be run in demo mode (with the
-dcommand line option), which is inspired by the code at https://lore.kernel.org/lkml/[email protected]/.In addition, it contains a set of performance tests with which the performance of thread switching with UMCG can be evaluated and compared with traditional threading implementations:
-uoption: UMCG-based performance test-foption: futex-based performance test-coption: condition variable-based performance testThe performance test creates a worker thread and continuously switches between the main thread and the worker thread, measuring how many main->worker->main thread switching cycles can be executed per unit of time. The test by default runs for 10 seconds, but the duration can be modified with the
-tcommand line option.Performance tests run in a KVM instance with a single vCPU showed the following figures:
The first 2 commits refactor existing code in order to allow defining arch-specific klibs and runtime tests, and to remove code duplication in the test suite.