Skip to content

Batch and SVM #2

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
Merged
merged 77 commits into from
Oct 18, 2021
Merged

Batch and SVM #2

merged 77 commits into from
Oct 18, 2021

Conversation

@arjunramaswami
Copy link
Collaborator

@arjunramaswami arjunramaswami commented Oct 18, 2021 •
edited
Loading

  • Batched 3D FFT to schedule multiple transformations by overlapping data transfers and FFT computations
  • Using OpenCL Shared Virtual Memory (SVM) for data transfers between FPGA and host

Arjun Ramaswami added 30 commits June 23, 2020 15:07
toggle interleaving using cmd line args, svm init
6d1fd3f
modified tests based on interleaving args
de65c7d
enabled fast emulation
dbc09a8
Working svm host code
6077dd6
malloc zero size check
302844c
Buffer locations attribute in fft3d kernel
d2003e5
Doxygen CMake, removed svm option in data alloc
b122231
Build doxygen only on option
9bd3b08
test simple batch svm execution
683cb21 
two svm fft3d calls within one call
Profile synthesis, fix without fftw
dc58b88
sq: general batch
a60c417
working batch fft3d svm
6f2fd40
cmake fix for ci
7368cdc
fft2d with opt matrix transpose
9731719
working 3D FFT with bitrev diagonal transpose
106182f
loop coalesced fft kernels
efb3e20
fft3d: single loop dbl buffered fetch1
f8dc082
fft3d: fetch_transpose working
a469f65
Buffer location in stores, iteration wise measures
8c090f8 
- buffer location to enable DDR stores in SVM,
- print iteration wise performance measurments
- modify kernel names in host to match the kernel
FFT2d: bram dbl buf and single loop
b7c457f
correct svm buffer flags, init output buffer, fft1d iter
eb91587
clFinish stms in DDR buffer reads
26187f8
FFT3D: separated svm and ddr
6d3ec73
FFT3D: Working opt bram transpose, bool instead of int
2a0176b
FFT3D: renamed bram transpose files, dependencies
724c747
CI: fixed host executable names
8288068
FFT1d: working svm
19c28fe
CI: corrected build, artifact paths
ce22802
Test: fix path to 2d bram
047d23b
CI: Adding release debug to restrict prints
e96acbf
Arjun Ramaswami and others added 28 commits October 8, 2021 17:48
updated tests
fe14f16
removed extern modules
da40391
updated ci
cd03621
updated ci with correct emulation env var
48f2e61
updated ci
8b26390
Update .gitlab-ci.yml
5f795e4
removed common cl files to common dir
fe24e32
Merge branch 'simplebatch_svm' of git.uni-paderborn.de:arjunr/fft3d-f…
ad1aec2 
…pga into simplebatch_svm
gitlab ci update
4fab1ae
updated readme
c06c222
updated changelog
1a97c92
fixed args in fft1d, fft2d
4a8bec0
fixed ddr batch and renamed queues
3dae9e1
added additional variants to application selection
a614680
expressive emulation and report generation
2ec2aa7
cleaned up application files
7c008f8
fixed emulation dependency in test
81fe6c1
draft userguide
6a1ddd4
fixed gitlab ci
0ae48a0
added power scripts
e49803c
minor change
5d343b2
updated changelog
08e97e1
ci: test fft1d
6152fd0
updated ci with 2d, 3d variants
9b11859
removed fft3d bram due to lack of cpu resources
8e5c316
cmake build type selection
58b7279
removed caching cmake variables for sdk and bsp
b8a7795
modified readme and updated userguide
8cebccd
@arjunramaswami arjunramaswami merged commit b1616a6 into master Oct 18, 2021
@arjunramaswami arjunramaswami deleted the simplebatch_svm branch October 18, 2021 12:02
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

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

@arjunramaswami