Fix GPU hangs in SiftGPU#56
Open
stotko wants to merge 1 commit intoniessner:masterfrom
Open
Conversation
|
Awesome! This fix should definitely be accepted. |
Open
|
I have fixed this problem for one week，in the environment of "Windows10 CUDA10.1 VS2017 GTX1650",I comment all __syncthreads() , it didn't work ,then I try again in the environement of "Windows10 CUDA8.0 VS2013 GTX1650",it works well ,thank you! |
i am wondering that did you comment all __syncthreads() in ProgramCU.cu or just the ones in ComputerOrientation_Kernel() functions？ |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Running BundleFusion on a RTX graphics card causes the GPU to hang due to thread synchronization bugs (see #51). For some reason, this problem does not occur on previous graphics cards.
Fix the synchronization bugs in the
__global__ void ComputeOrientation_Kernel(...)by replacing the conditional__syncthreads()calls with unconditional ones. Thread synchronization requires that all threads reach the call, otherwise the GPU hangs and waits for the remaining ones, which will of course never happen due to theifcondition.