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

Device Abstraction #610

Closed
wants to merge 62 commits into from
Closed

Device Abstraction #610

wants to merge 62 commits into from

Commits on Aug 12, 2014

  1. Wrap the CPU and GPU math functions in math backend classes

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    bc909c6 View commit details
    Browse the repository at this point in the history

  2. Add the math backend to the Layer base class

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    93bee80 View commit details
    Browse the repository at this point in the history

  3. Add device type independent getters to Blob

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    b9621c6 View commit details
    Browse the repository at this point in the history

  4. Remove tab from the code and reformat using google style

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    8697a72 View commit details
    Browse the repository at this point in the history

  5. Allow Layer::Forward and Backward to be overridden

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    6df6109 View commit details
    Browse the repository at this point in the history

  6. Use zero as the default return values of Blob data and diff methods

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    2883a62 View commit details
    Browse the repository at this point in the history

  7. Add and test device type ignorant Forward and Backward in ConcatLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    ed71d59 View commit details
    Browse the repository at this point in the history

  8. Add default implementations of Layer::Forward_cpu and Backward_cpu

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    7b1733c View commit details
    Browse the repository at this point in the history

  9. Directly implement device neutral Forward and Backward in ConcatLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    0f37050 View commit details
    Browse the repository at this point in the history

  10. Generalize the math backend classes into device wrapper classes

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    13f2f57 View commit details
    Browse the repository at this point in the history

  11. Add Device::copy_from_cpu for the data layers

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    79325f3 View commit details
    Browse the repository at this point in the history

  12. Unify the CPU and the GPU Forward of the DataLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    67d5621 View commit details
    Browse the repository at this point in the history

  13. Unify the CPU and the GPU Forward of the ImageDataLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    e57a703 View commit details
    Browse the repository at this point in the history

  14. Unify the CPU and the GPU Forward of the HDF5DataLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    aed27b0 View commit details
    Browse the repository at this point in the history

  15. Unify the CPU and the GPU Forward & Backward of the HDF5OutputDataLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    0703d7b View commit details
    Browse the repository at this point in the history

  16. Merge the CPU and the GPU Backward of the data layers

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    ed7582b View commit details
    Browse the repository at this point in the history

  17. Consolidate the CPU and GPU Forward of the WindowDataLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    9b01ff1 View commit details
    Browse the repository at this point in the history

  18. Deduplicate the CPU and the GPU Forward & Backward of the FlattenLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    063deba View commit details
    Browse the repository at this point in the history

  19. Use the newly implemented caffe_gpu_{add,sub} in the GPU device wrapper

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    b945f63 View commit details
    Browse the repository at this point in the history

  20. Unify the CPU/GPU Forward/Backward of the SigmoidCrossEntropyLossLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    b8c0ec1 View commit details
    Browse the repository at this point in the history

  21. Merge the CPU/GPU Forward/Backward of the SoftmaxWithLossLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    75a71e7 View commit details
    Browse the repository at this point in the history

  22. Use {const, mutable}_{data, diff} in the unified Forward/Backward

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    df4c2e5 View commit details
    Browse the repository at this point in the history

  23. Unify the CPU/GPU Forward/Backward of the InnerProductLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    716c29d View commit details
    Browse the repository at this point in the history

  24. Unify the CPU/GPU Forward/Backward of the SplitLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    f14911e View commit details
    Browse the repository at this point in the history

  25. Unify the CPU/GPU Forward/Backward of the EltwiseLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    faef386 View commit details
    Browse the repository at this point in the history

  26. Add im2col and col2im to wrap im2col_{cpu, gpu} and col2im_{cpu, gpu}

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    740820b View commit details
    Browse the repository at this point in the history

  27. Unify the CPU/GPU versions of Forward/Backward of the ConvolutionLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    0275bdf View commit details
    Browse the repository at this point in the history

  28. Unify the CPU/GPU versions of Forward/Backward of the Im2colLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    f70db65 View commit details
    Browse the repository at this point in the history

  29. Unify the CPU/GPU versions of Forward/Backward of the PowerLayer

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    ae17c17 View commit details
    Browse the repository at this point in the history

  30. Move im2col and col2im into the device wrapper classes

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    4890b55 View commit details
    Browse the repository at this point in the history

  31. Update the include guard of the util/device.hpp

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    31fb4b9 View commit details
    Browse the repository at this point in the history

  32. Device wrapper methods no longer pure virtual, default not implemented 

    Conflicts:
	include/caffe/util/device.hpp
    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    56e23fa View commit details
    Browse the repository at this point in the history

  33. Fix the rebase errors introduced when merge conflicts are resolved

    @kloudkl @jeffdonahue
    kloudkl authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    d7230b9 View commit details
    Browse the repository at this point in the history

  34. Add current-device data accessors to SyncedMem. Use to fix Convolutio… 

    …nLayer.

Conflicts:
	include/caffe/syncedmem.hpp
	src/caffe/syncedmem.cpp
    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    6618265 View commit details
    Browse the repository at this point in the history

  35. Fix using wrong device data. Fix tests. All tests pass on a CUDA box.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    d36ed3b View commit details
    Browse the repository at this point in the history

  36. Fix rebase errors.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    99380c9 View commit details
    Browse the repository at this point in the history

  37. Incorporate additional abstractions from 1605431.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    d45100f View commit details
    Browse the repository at this point in the history

  38. Move device.hpp up out of util/.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    ff9bff0 View commit details
    Browse the repository at this point in the history

  39. Make device factory usage more inline with layer factory.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    96ed3c2 View commit details
    Browse the repository at this point in the history

  40. Move devices into their own src dir.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    e12aa5b View commit details
    Browse the repository at this point in the history

  41. Incorporate all cpu math functions directly into CPUDevice.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    6db9c8c View commit details
    Browse the repository at this point in the history

  42. Fix compilation errors.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    1393aa4 View commit details
    Browse the repository at this point in the history

  43. Fix lint errors.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    72b0af1 View commit details
    Browse the repository at this point in the history

  44. Fix formatting.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    9420cea View commit details
    Browse the repository at this point in the history

  45. Try to obey google style guide.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    bddcae8 View commit details
    Browse the repository at this point in the history

  46. Replace all caffe_*() cpu functions with CPUDevice functions.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    f721ac6 View commit details
    Browse the repository at this point in the history

  47. Get rid of all caffe_* cpu functions.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    8fb944c View commit details
    Browse the repository at this point in the history

  48. Move caffe_rng_rand() into util/rng.hpp.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    af5e292 View commit details
    Browse the repository at this point in the history

  49. Implement all non-kernel gpu functions in GPUDevice and replace corre… 

    …sponding calls.

Get rid of now empty util/math_functions.cpp.
    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    e1c66dc View commit details
    Browse the repository at this point in the history

  50. Implement all kernel device functions under GPUDevice. Get rid of unu… 

    …sed files.
    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    c125a97 View commit details
    Browse the repository at this point in the history

  51. Implement sqr() and exp() for GPUDevice.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    8d111e4 View commit details
    Browse the repository at this point in the history

  52. Separate device.hpp into device-specific headers.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    39c8658 View commit details
    Browse the repository at this point in the history

  53. Use UVA for GPU copy. Get rid of copy_from_cpu().

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    e68a140 View commit details
    Browse the repository at this point in the history

  54. Explicitly instantiate/specialize int and unsigned int templates for … 

    …device class functions.
    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    e523b94 View commit details
    Browse the repository at this point in the history

  55. Fix sign kernel for unsigned int.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    8bff9ce View commit details
    Browse the repository at this point in the history

  56. Fix all rebase errors. All tests are passing.

    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    9a5b6d2 View commit details
    Browse the repository at this point in the history

  57. Remove GPU stubs for all layers with device-unified Forward() and Bac… 

    …kward().
    @jeffdonahue
    Rob Hess authored and jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    72e4c21 View commit details
    Browse the repository at this point in the history

  58. include math.h in filler and solver -- seems to be needed for pow and 

    sqrt in CPU_ONLY build.
    @jeffdonahue
    jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    95bf45f View commit details
    Browse the repository at this point in the history

  59. Add "#ifndef CPU_ONLY" guards around GPU device code so that CPU_ONLY 

    compilation works.
    @jeffdonahue
    jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    afd7bf5 View commit details
    Browse the repository at this point in the history

  60. Save warnings when building device files.

    @jeffdonahue
    jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    2e23318 View commit details
    Browse the repository at this point in the history

  61. Fix post-rebase errors.

    @jeffdonahue
    jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    e30ba4d View commit details
    Browse the repository at this point in the history

  62. Fix post-rebase errors, including: 

    -device-abstracted version of MVN
-new memset/memcpy wrappers (set_void and copy_void)
-fixing MKL switching logic
    @jeffdonahue
    jeffdonahue committed Aug 12, 2014
    Configuration menu
    Copy the full SHA
    d52d669 View commit details
    Browse the repository at this point in the history