Frequency bundling introduced #5
Open
Conversation
This file contains hidden or 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
A couple of fixes
chime branch merged into master
This is a temporary workaround to avoid linking problems with compiling with --enable-shared. The basic issue is that code under Kernel/ can't depend on code under Signal/ under the current dependency scheme.
Updated master branch from demorest
Merging changes to master into chime branch and hoping for the best
…he wrond frequency order
…name() function for various backends
|
I think the change to unpacking to make it cache-friendly accidentally made it unpack waaay more than it was supposed to, so I made a change to BitUnpacker. Also, unloading with subintegrations when using bundles is currently broken, which I'm trying to get working. |
Updating from master
…ed to spaces" This reverts commit 1f88b53.
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Outdated suggestions cannot be applied.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.
It's still buggy, but the ability to set nbundle is in place, and the code (generally) runs from beginning to end and I'm pretty sure it doesn't corrupt data. This makes it possible to run MANY-channel input (for CHIME it's 1024 input channels) in subsets of channels (eg, channels 0-7, then 8-15, then 16-23, ...) allowing for more time samples to be loaded to GPU at once and thus longer FFTs to be performed. In its current form, unpacking on the CPU is the biggest bottleneck. I'm having each thread unpack a complete set of frequency channels, then transfer to GPU etc in bundles until it returns to bundle 0 and unpacks some more. Since streams are launched simultaneously, they all unpack at once which slows down the unpacking, then they all compute at once while no unpacking happens. The unpacking being staggered between threads would solve or largely solve this problem.