Permalink
Commits on Jan 13, 2017
  1. Merge pull request #911 from gchanan/convWarning

    Avoid strict aliasing warning in float/half conversions.
    soumith committed on GitHub Jan 13, 2017
  2. Avoid strict aliasing warning in float/half conversions.

    Verified that at least for GCC 4.47 this generates identical code.
    Gregory Chanan committed Jan 13, 2017
Commits on Jan 11, 2017
  1. Merge pull request #908 from CDLuminate/really-fix-arm64-build

    simd.h: really fix the arm64 (i.e. Aarch64) build
    soumith committed on GitHub Jan 11, 2017
Commits on Jan 10, 2017
  1. Merge pull request #906 from torch/installpaths

    provide install paths through require 'torch.paths'
    andresy committed on GitHub Jan 10, 2017
  2. provide install paths through require 'torch.paths'

    andresy committed Jan 10, 2017
  3. fixing arm64 build

    soumith committed on GitHub Jan 10, 2017
Commits on Jan 6, 2017
  1. Merge pull request #899 from gchanan/TensorCopyHalf

    Fix THHalf issues with MSVC.
    soumith committed on GitHub Jan 6, 2017
Commits on Jan 5, 2017
  1. Fix THHalf issues with MSVC.

    Gregory Chanan committed Jan 5, 2017
  2. Merge pull request #895 from colesbury/threadArgErrorHandler

    Fix condition for threadArgErrorHandler
    soumith committed on GitHub Jan 5, 2017
  3. Fix condition for threadArgErrorHandler

    Some error handlers may not have any data associated with them
    colesbury committed Jan 5, 2017
Commits on Jan 3, 2017
  1. Merge pull request #892 from temerick/patch-3

    Update FindARM.cmake
    soumith committed on GitHub Jan 3, 2017
  2. Update FindARM.cmake

    Fix typos
    temerick committed on GitHub Jan 3, 2017
Commits on Jan 2, 2017
  1. small cmake fix

    soumith committed Jan 2, 2017
Commits on Jan 1, 2017
  1. Merge pull request #873 from nkoumchatzky/contiguous-cat-1d

    Different code path for contiguous tensors along dimension 1 in cat + fix cat bug + fix ambiguity
    soumith committed on GitHub Jan 1, 2017
  2. Merge pull request #890 from torch/halfcleanup

    Removing Half conditional macros, and enabling half compilation by default
    soumith committed on GitHub Jan 1, 2017
  3. Removing TH_GENERIC_USE_HALF, TH_NATIVE_HALF, TH_GENERIC_NO_MATH (rep…

    …laced where appropriate with TH_REAL_IS_HALF), removed half from THGenerateAllTypes, added an explicit THGenerateHalfType.h
    soumith committed Jan 1, 2017
  4. Merge pull request #889 from torch/revert-888-halfgeneric

    Revert "TH_GENERIC_USE_HALF=1 by default, half enabled by default"
    soumith committed on GitHub Jan 1, 2017
  5. Revert "TH_GENERIC_USE_HALF=1 by default, half enabled by default"

    soumith committed on GitHub Jan 1, 2017
  6. Merge pull request #888 from torch/halfgeneric

    TH_GENERIC_USE_HALF=1 by default, half enabled by default
    soumith committed on GitHub Jan 1, 2017
  7. TH_GENERIC_USE_HALF=1 by default, half enabled by default

    soumith committed Jan 1, 2017
Commits on Dec 31, 2016
  1. cmake C flags fix

    soumith committed Dec 31, 2016
Commits on Dec 30, 2016
  1. adding TH_SO_VERSION option

    soumith committed Dec 30, 2016
  2. Merge pull request #886 from torch/simdcpuidfix

    use __get_cpuid when available
    soumith committed on GitHub Dec 30, 2016
  3. use __get_cpuid when available

    soumith committed Dec 30, 2016
  4. Merge pull request #875 from atkayu/add_histc2

    Add a new function bhistc to calculate histogram of batch of images only once
    soumith committed on GitHub Dec 30, 2016
Commits on Dec 29, 2016
  1. Merge pull request #882 from amrobbins/ppcvectorinstxns

    Add support for VSX vector instructions on PPC
    soumith committed on GitHub Dec 29, 2016
  2. Add support for torch.HalfTensor (#874)

    * Add support for torch.HalfTensor.
    
    * Improvements/Simplifications for torch.HalfTensor.
    
    Improvements/Simplifications:
    1) Defines half type as TH_Half, so as to not conflict with cutorch
    version.  Previously, these were defined as the same "half" type and
    required proper ordering of includes to ensure type was only defined
    once, which would have affected all downstream projects.
    2) No longer generates math functions that are not actually defined
    on torch.HalfTensor, e.g. maskedFill, map, etc.
    3) Adds tests for all available torch.HalfTensor functions
    4) Allows compiling without TH_GENERIC_USE_HALF (so if there's a
    problem can just unset that in CMakeLists rather than backing out)
    5) Some simplifications: removes a new copy optimization and
    some TH_HALF literal definitions
    
    Limitations:
    Because match functions are not defined, some "non-math" operators
    on torch.HalfTensor give an error message, e.g. __index__/__newindex__
    with a ByteTensor apply a mask, but masks aren't implemented.  These
    limitations aren't always obvious, (e.g. for documentation purposes),
    but they should always give an error message.
    
    * Rename TH_HALF to THHalf.
    gchanan committed with soumith Dec 29, 2016
Commits on Dec 28, 2016
  1. Add support for VSX vector instructions on PPC

    Added support for the fill, diff, scale, mul and add functions using
    PPC CPU vector instructions. These are used in place of the versions
    of these functions written for x86, when compiled on PPC.
    
    This fixes a compile failure on PPC
    amrobbins committed Dec 28, 2016
  2. rename histc2 to bhistc

    atkayu committed Dec 28, 2016
Commits on Dec 26, 2016
  1. Add a different code path for catting contiguous tensors along the fi…

    …rst dimension, for speed reasons.
    
    Fix a bug in cat when catting with an empty tensor along first dim (it added an extra dim).
    Fix the ambiguous 'catting along last dimension' sentence in the doc and change the behavior to pick the maximum last dimension over all input tensors.
    Now empty tensors are allowed.
    nkoumchatzky committed Dec 21, 2016
Commits on Dec 22, 2016
  1. add a new function histc2

    atkayu committed Dec 22, 2016
Commits on Dec 20, 2016
  1. Merge pull request #872 from pavanky/modfix

    correctness fixes for mod and remainder for integer type tensors.
    soumith committed on GitHub Dec 20, 2016