Benoit Steiner
e3a15a03a4
Don't explicitely evaluate the subexpression from TensorForcedEval::evalSubExprIfNeeded, as it will be done when executing the EvalTo subexpression
2016-01-24 23:04:50 -08:00
Benoit Steiner
bd207ce11e
Added missing EIGEN_DEVICE_FUNC qualifier
2016-01-24 20:36:05 -08:00
Benoit Steiner
cb4e53ff7f
Merged in ville-k/eigen/tensorflow_fix (pull request PR-153)
...
Add ctor for long
2016-01-22 19:11:31 -08:00
Benoit Steiner
3aeeca32af
Leverage the new blocking code in the tensor contraction code.
2016-01-22 16:36:30 -08:00
Benoit Steiner
4beb447e27
Created a mechanism to enable contraction mappers to determine the best blocking strategy.
2016-01-22 14:37:26 -08:00
Gael Guennebaud
6a44ccb58b
Backout changeset 690bc950f7
2016-01-22 15:03:53 +01:00
Ville Kallioniemi
9b6c72958a
Update to latest default branch
2016-01-21 23:08:54 -07:00
Benoit Steiner
c33479324c
Fixed a constness bug
2016-01-21 17:08:11 -08:00
Jan Prach
690bc950f7
fix clang warnings
...
"braces around scalar initializer"
2016-01-20 19:35:59 -08:00
Benoit Steiner
7ce932edd3
Small cleanup and small fix to the contraction of row major tensors
2016-01-20 18:12:08 -08:00
Benoit Steiner
47076bf00e
Reduce the register pressure exerted by the tensor mappers whenever possible. This improves the performance of the contraction of a matrix with a vector by about 35%.
2016-01-20 14:51:48 -08:00
Ville Kallioniemi
2832175a68
Use explicitly 32 bit integer types in constructors.
2016-01-19 20:12:17 -07:00
Benoit Steiner
df79c00901
Improved the formatting of the code
2016-01-19 17:24:08 -08:00
Benoit Steiner
6d472d8375
Moved the contraction mapping code to its own file to make the code more manageable.
2016-01-19 17:22:05 -08:00
Benoit Steiner
b3b722905f
Improved code indentation
2016-01-19 17:09:47 -08:00
Benoit Steiner
5b7713dd33
Record whether the underlying tensor storage can be accessed directly during the evaluation of an expression.
2016-01-19 17:05:10 -08:00
Ville Kallioniemi
63fb66f53a
Add ctor for long
2016-01-17 21:25:36 -07:00
Benoit Steiner
34057cff23
Fixed a race condition that could affect some reductions on CUDA devices.
2016-01-15 15:11:56 -08:00
Benoit Steiner
0461f0153e
Made it possible to compare tensor dimensions inside a CUDA kernel.
2016-01-15 11:22:16 -08:00
Benoit Steiner
aed4cb1269
Use warp shuffles instead of shared memory access to speedup the inner reduction kernel.
2016-01-14 21:45:14 -08:00
Benoit Steiner
8fe2532e70
Fixed a boundary condition bug in the outer reduction kernel
2016-01-14 09:29:48 -08:00
Benoit Steiner
9f013a9d86
Properly record the rank of reduced tensors in the tensor traits.
2016-01-13 14:24:37 -08:00
Benoit Steiner
79b69b7444
Trigger the optimized matrix vector path more conservatively.
2016-01-12 15:21:09 -08:00
Benoit Steiner
d920d57f38
Improved the performance of the contraction of a 2d tensor with a 1d tensor by a factor of 3 or more. This helps speedup LSTM neural networks.
2016-01-12 11:32:27 -08:00
Benoit Steiner
bd7d901da9
Reverted a previous change that tripped nvcc when compiling in debug mode.
2016-01-11 17:49:44 -08:00
Benoit Steiner
c5e6900400
Silenced a few compilation warnings.
2016-01-11 17:06:39 -08:00
Benoit Steiner
f894736d61
Updated the tensor traits: the alignment is not part of the Flags enum anymore
2016-01-11 16:42:18 -08:00
Benoit Steiner
4f7714d72c
Enabled the use of fixed dimensions from within a cuda kernel.
2016-01-11 16:01:00 -08:00
Benoit Steiner
01c55d37e6
Deleted unused variable.
2016-01-11 15:53:19 -08:00
Benoit Steiner
0504c56ea7
Silenced a nvcc compilation warning
2016-01-11 15:49:21 -08:00
Benoit Steiner
b523771a24
Silenced several compilation warnings triggered by nvcc.
2016-01-11 14:25:43 -08:00
Benoit Steiner
2c3b13eded
Merged in jeremy_barnes/eigen/shader-model-3.0 (pull request PR-152)
...
Alternative way of forcing instantiation of device kernels without causing warnings or requiring device to device kernel invocations.
2016-01-11 11:43:37 -08:00
Benoit Steiner
2ccb1c8634
Fixed a bug in the dispatch of optimized reduction kernels.
2016-01-11 10:36:37 -08:00
Benoit Steiner
780623261e
Re-enabled the optimized reduction CUDA code.
2016-01-11 09:07:14 -08:00
Jeremy Barnes
91678f489a
Cleaned up double-defined macro from last commit
2016-01-10 22:44:45 -05:00
Jeremy Barnes
403a7cb6c3
Alternative way of forcing instantiation of device kernels without
...
causing warnings or requiring device to device kernel invocations.
This allows Tensorflow to work on SM 3.0 (ie, Amazon EC2) machines.
2016-01-10 22:39:13 -05:00
Benoit Steiner
e76904af1b
Simplified the dispatch code.
2016-01-08 16:50:57 -08:00
Benoit Steiner
d726e864ac
Made it possible to use array of size 0 on CUDA devices
2016-01-08 16:38:14 -08:00
Benoit Steiner
3358dfd5dd
Reworked the dispatch of optimized cuda reduction kernels to workaround a nvcc bug that prevented the code from compiling in optimized mode in some cases
2016-01-08 16:28:53 -08:00
Benoit Steiner
53749ff415
Prevent nvcc from miscompiling the cuda metakernel. Unfortunately this reintroduces some compulation warnings but it's much better than having to deal with random assertion failures.
2016-01-08 13:53:40 -08:00
Benoit Steiner
6639b7d6e8
Removed a couple of partial specialization that confuse nvcc and result in errors such as this:
...
error: more than one partial specialization matches the template argument list of class "Eigen::internal::get<3, Eigen::internal::numeric_list<std::size_t, 1UL, 1UL, 1UL, 1UL>>"
"Eigen::internal::get<n, Eigen::internal::numeric_list<T, a, as...>>"
"Eigen::internal::get<n, Eigen::internal::numeric_list<T, as...>>"
2016-01-07 18:45:19 -08:00
Benoit Steiner
0cb2ca5de2
Fixed a typo.
2016-01-06 18:50:28 -08:00
Benoit Steiner
213459d818
Optimized the performance of broadcasting of scalars.
2016-01-06 18:47:45 -08:00
Benoit Steiner
cfff40b1d4
Improved the performance of reductions on CUDA devices
2016-01-04 17:25:00 -08:00
Benoit Steiner
515dee0baf
Added a 'divup' util to compute the floor of the quotient of two integers
2016-01-04 16:29:26 -08:00
Gael Guennebaud
978c379ed7
Add missing ctor from uint
2015-12-30 12:52:38 +01:00
Benoit Steiner
bdcbc66a5c
Don't attempt to vectorize mean reductions of integers since we can't use
...
SSE or AVX instructions to divide 2 integers.
2015-12-22 17:51:55 -08:00
Benoit Steiner
a1e08fb2a5
Optimized the configuration of the outer reduction cuda kernel
2015-12-22 16:30:10 -08:00
Benoit Steiner
9c7d96697b
Added missing define
2015-12-22 16:11:07 -08:00
Benoit Steiner
e7e6d01810
Made sure the optimized gpu reduction code is actually compiled.
2015-12-22 15:07:33 -08:00
