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.

Sign up for GitHub

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

Jump to bottom

GPU backend branch merge #148

Open
wants to merge 117 commits into
base: master
Choose a base branch
from
Open

GPU backend branch merge #148

wants to merge 117 commits into from

Conversation

AjayBrahmakshatriya
Copy link
Member

@AjayBrahmakshatriya AjayBrahmakshatriya commented Mar 19, 2021 •
edited
Loading

This PR is to merge the GPU backend branch changes into the master branch. Following things have been taken care of -

  1. No merge conflicts
  2. Project builds properly
  3. All tests including c++ tests, python tests (with and without schedules) and gpu_tests are passing
  4. GPU backend code changed to adapt to the functor changes introduced
  5. The CGO'21 AE is added to the repo and all applications are successfully running.

@AjayBrahmakshatriya
SSSP basic compiling
17a90fb
@AjayBrahmakshatriya
SSSP compiled and running correctly. Correctness verified
d46022b
@AjayBrahmakshatriya
Added support for deduplicaiton, unfused_boolmap frontier creation an…
530b472 
…d TWCE
@AjayBrahmakshatriya
Pull direction working with just VertexBased load balance
71692ad
@AjayBrahmakshatriya
Added a src filter. Doesn't affect performance for SSSP
05fb2ab
@yunmingzhang17
reformating and adding new GPU schedule tests
4d79563
@yunmingzhang17
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
b0a3392 
…_backend
@AjayBrahmakshatriya
Working Hybrid operator
505441d
@AjayBrahmakshatriya
Fixed the apply expr lower vardecl->assign conversion only if GPU sch…
c4bc834 
…edule is attached
@AjayBrahmakshatriya
Added bitmap support for PULL direction
cb0c070
@AjayBrahmakshatriya
Added GPU test framework
9d109fb
@AjayBrahmakshatriya
Added runtime library tests to GPU test suite
85fb156
@AjayBrahmakshatriya
Wrapped load balance functions in host wrappers
cde571c
setting up tests for gpu-based priority queue
fe272bf
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
5e2ff33 
…_backend
adding the file for gpu priority queue
6532105
start working on a sssp_delta_stepping gpu runtime library test
4811023
@AjayBrahmakshatriya
Basic Kernel fusion implemented. Compiling for SSSP. Need to get oper…
8b9f8c7 
…ator lowering done
@AjayBrahmakshatriya
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
0ea0af5 
…_backend
@AjayBrahmakshatriya
Added hoisted decls for fused kernel generation and printing support
0439720
adding correctness test for sssp, setting up the first verifier
7198946
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
7ad6cfe 
…_backend
@AjayBrahmakshatriya
Basic Kernel fusion working for SSSP TWCE push unfused
4f29384
@AjayBrahmakshatriya
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
c318d42 
…_backend
@AjayBrahmakshatriya
Changed file name for 4.mtx and relevant test cases
537c5fe
@AjayBrahmakshatriya
Added generating Function declaration before kernel
19ba657
@AjayBrahmakshatriya
Added destination filter to push edge set apply and fixed the bitmap …
ffddce8 
…implementation
moving the sssp runtime lib into a sssp_lp (label propagation) file, …
ed487f1 
…and adding a real delta stepping version
converting a true delta stepping version to use the runtime libraries
57bea68
merging with the latest code
561c9ad
yunmingzhang17 and others added 25 commits October 29, 2019 15:56
@yunmingzhang17
using reference for the frontier
23b8c83
@AjayBrahmakshatriya
Added the EnqueueVertex operator and fixed the implementation for BFS
156a3f3
@AjayBrahmakshatriya
t Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into g…
5a6e067 
…pu_backend
@yunmingzhang17
refactor with updatePriorityMin on the priority queue
437ecb0
@yunmingzhang17
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
6e28e5a 
…_backend
@yunmingzhang17
adding a check to make sure only eager or eager with merge schedule w…
0303539 
…ould lower into ordered processing operator
@AjayBrahmakshatriya
Fixed all codegen for SSSP_delta_stepping
b2cdecc
@AjayBrahmakshatriya
Working code for Priority Queue with kernel fusion
1795db6
@AjayBrahmakshatriya
Typo in codegen for fused kernel
489735b
@AjayBrahmakshatriya
PageRank blocking issue fixed
842f1d8
@AjayBrahmakshatriya
Fixed soem issues for BFS
98c468c
@AjayBrahmakshatriya
Added GPU autotuner
3b854e6
@AjayBrahmakshatriya
Added Pointer jumping version of cc
2dc1827
@AjayBrahmakshatriya
Removed schedule from pjump
6754d68
@AjayBrahmakshatriya
BC changes and autotuner support for GPU backend
08c3ae8
@AjayBrahmakshatriya
Added implementation and test cases for MIRMetadata
8a2e1e0
@AjayBrahmakshatriya
cloning function for mir metadata
63f31f3
@AjayBrahmakshatriya
Added perf tests for GPU
c3be307
@AjayBrahmakshatriya
Update .travis.yml
4c8aac8
@AjayBrahmakshatriya
Small changes
729421d
@AjayBrahmakshatriya
Merge branch 'gpu_backend' of github.com:GraphIt-DSL/graphit into gpu…
09682fc 
…_backend
@AjayBrahmakshatriya
Very minor code generation bug
5974418
@AjayBrahmakshatriya
Update README.md
6cc3eca
@AjayBrahmakshatriya
Merge completed. All CPU + GPU tests passing
42ff6cb
@AjayBrahmakshatriya
Added CGO21 AE to the repo, all tests passing
f8646f1
@yunmingzhang17 yunmingzhang17 marked this pull request as ready for review March 19, 2021 15:04
yunmingzhang17
yunmingzhang17 approved these changes Mar 29, 2021
View reviewed changes
Copy link
Collaborator

@yunmingzhang17 yunmingzhang17 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

generally looks good. It's a big PR, let's test it a bit more.

include/graphit/midend/priority_queue_frontier_reuse.h
@@ -0,0 +1,10 @@
#ifndef PRIORITY_QUEUE_FRONTIER_REUSE_H
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is this an empty file?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is an incomplete feature I was trying to implement. This shouldn't have been added. I will remove it.

src/backend/codegen_gpu/extract_read_write_set.cpp
void ExtractReadWriteSet::visit(mir::StmtBlock::Ptr stmt_block) {
return;
}
void ExtractReadWriteSet::visit(mir::TensorArrayReadExpr::Ptr tare) {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would use a more meaningful variable name, not "tare". In general, it would be easier if we don't use the acronyms.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I agree, I have been a bit liberal with all the names here. I can take a pass and clean up all the variable names used in the GPU backend and the related passes.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@yunmingzhang17 yunmingzhang17 yunmingzhang17 approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@AjayBrahmakshatriya @yunmingzhang17