Run the capture thread with real-time priority #116
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
Previously, the variables used for synchronisation between threads during capture were declared as volatile. While this will work on some platforms and compilers (it's usually safe on x86), it's not strictly correct for the C++ memory model: you need to update the variable using an operation with a defined ordering to ensure that other changes to memory (e.g. data in the buffer) are visible to other threads. Use atomics instead, with the default sequential-consistency semantics, which should do the right thing on all platforms.
This sets the scheduling class of the USB capture thread to SCHED_RR while a capture is in progress. This should give the capture thread a better chance of reacting quickly to transfers completing, especially if there are other regular userspace processes competing with it. If setting the priority fails (e.g. because the system ulimit for real-time priority isn't high enough), it will print a warning and continue as usual.
If mlock fails (e.g. because the corresponding ulimit is set too low), print a warning and continue. (Note that on some platforms, mlock requires the base address to be page-aligned; the code doesn't currently try to achieve that because it's not required on Linux.)
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Outdated suggestions cannot be applied.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 uses a couple of techniques common in Linux audio real-time programming: if possible, run the latency-sensitive capture thread with the
SCHED_RRscheduling policy so it will preempt other userspace processes, and usemlockto ensure that the disk buffers stay in main memory rather than being paged out.I tested this on my oldest USB3-capable machine (a Lenovo X220), and it significantly reduces the incidence of test-mode errors when the machine's under deliberately heavy load. On a current machine running a PREEMPT-RT kernel, I ran an 8h test-mode capture without errors while doing large compiles in the background.
This only affects the USB capture side of things - it would be worth thinking about increasing the number of disk buffers as well by default, since the current buffer is only about 3 seconds' worth of capture, and I expect most machines with USB3 can spare a couple of gigs of memory...