CFD/eigen
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
10,541 Commits 2 Branches 88 Tags 112 MiB
C++ 84.6%
Fortran 8.9%
C 2.9%
CMake 1.9%
Cuda 1.2%
Other 0.5%
8c7a6feb8e
Go to file
Anuj Rawat 8c7a6feb8e Adding lowlevel APIs for optimized RHS packet load in TensorFlow 
SpatialConvolution

Low-level APIs are added in order to optimized packet load in gemm_pack_rhs
in TensorFlow SpatialConvolution. The optimization is for scenario when a
packet is split across 2 adjacent columns. In this case we read it as two
'partial' packets and then merge these into 1. Currently this only works for
Packet16f (AVX512) and Packet8f (AVX2). We plan to add this for other
packet types (such as Packet8d) also.

This optimization shows significant speedup in SpatialConvolution with
certain parameters. Some examples are below.

Benchmark parameters are specified as:
Batch size, Input dim, Depth, Num of filters, Filter dim

Speedup numbers are specified for number of threads 1, 2, 4, 8, 16.

AVX512:

Parameters                  | Speedup (Num of threads: 1, 2, 4, 8, 16)
----------------------------|------------------------------------------
128,   24x24,  3, 64,   5x5 |2.18X, 2.13X, 1.73X, 1.64X, 1.66X
128,   24x24,  1, 64,   8x8 |2.00X, 1.98X, 1.93X, 1.91X, 1.91X
 32,   24x24,  3, 64,   5x5 |2.26X, 2.14X, 2.17X, 2.22X, 2.33X
128,   24x24,  3, 64,   3x3 |1.51X, 1.45X, 1.45X, 1.67X, 1.57X
 32,   14x14, 24, 64,   5x5 |1.21X, 1.19X, 1.16X, 1.70X, 1.17X
128, 128x128,  3, 96, 11x11 |2.17X, 2.18X, 2.19X, 2.20X, 2.18X

AVX2:

Parameters                  | Speedup (Num of threads: 1, 2, 4, 8, 16)
----------------------------|------------------------------------------
128,   24x24,  3, 64,   5x5 | 1.66X, 1.65X, 1.61X, 1.56X, 1.49X
 32,   24x24,  3, 64,   5x5 | 1.71X, 1.63X, 1.77X, 1.58X, 1.68X
128,   24x24,  1, 64,   5x5 | 1.44X, 1.40X, 1.38X, 1.37X, 1.33X
128,   24x24,  3, 64,   3x3 | 1.68X, 1.63X, 1.58X, 1.56X, 1.62X
128, 128x128,  3, 96, 11x11 | 1.36X, 1.36X, 1.37X, 1.37X, 1.37X

In the higher level benchmark cifar10, we observe a runtime improvement
of around 6% for AVX512 on Intel Skylake server (8 cores).

On lower level PackRhs micro-benchmarks specified in TensorFlow
tensorflow/core/kernels/eigen_spatial_convolutions_test.cc, we observe
the following runtime numbers:

AVX512:

Parameters                                                     | Runtime without patch (ns) | Runtime with patch (ns) | Speedup
---------------------------------------------------------------|----------------------------|-------------------------|---------
BM_RHS_NAME(PackRhs, 128, 24, 24, 3, 64, 5, 5, 1, 1, 256, 56)  |  41350                     | 15073                   | 2.74X
BM_RHS_NAME(PackRhs, 32, 64, 64, 32, 64, 5, 5, 1, 1, 256, 56)  |   7277                     |  7341                   | 0.99X
BM_RHS_NAME(PackRhs, 32, 64, 64, 32, 64, 5, 5, 2, 2, 256, 56)  |   8675                     |  8681                   | 1.00X
BM_RHS_NAME(PackRhs, 32, 64, 64, 30, 64, 5, 5, 1, 1, 256, 56)  |  24155                     | 16079                   | 1.50X
BM_RHS_NAME(PackRhs, 32, 64, 64, 30, 64, 5, 5, 2, 2, 256, 56)  |  25052                     | 17152                   | 1.46X
BM_RHS_NAME(PackRhs, 32, 256, 256, 4, 16, 8, 8, 1, 1, 256, 56) |  18269                     | 18345                   | 1.00X
BM_RHS_NAME(PackRhs, 32, 256, 256, 4, 16, 8, 8, 2, 4, 256, 56) |  19468                     | 19872                   | 0.98X
BM_RHS_NAME(PackRhs, 32, 64, 64, 4, 16, 3, 3, 1, 1, 36, 432)   | 156060                     | 42432                   | 3.68X
BM_RHS_NAME(PackRhs, 32, 64, 64, 4, 16, 3, 3, 2, 2, 36, 432)   | 132701                     | 36944                   | 3.59X

AVX2:

Parameters                                                     | Runtime without patch (ns) | Runtime with patch (ns) | Speedup
---------------------------------------------------------------|----------------------------|-------------------------|---------
BM_RHS_NAME(PackRhs, 128, 24, 24, 3, 64, 5, 5, 1, 1, 256, 56)  | 26233                      | 12393                   | 2.12X
BM_RHS_NAME(PackRhs, 32, 64, 64, 32, 64, 5, 5, 1, 1, 256, 56)  |  6091                      |  6062                   | 1.00X
BM_RHS_NAME(PackRhs, 32, 64, 64, 32, 64, 5, 5, 2, 2, 256, 56)  |  7427                      |  7408                   | 1.00X
BM_RHS_NAME(PackRhs, 32, 64, 64, 30, 64, 5, 5, 1, 1, 256, 56)  | 23453                      | 20826                   | 1.13X
BM_RHS_NAME(PackRhs, 32, 64, 64, 30, 64, 5, 5, 2, 2, 256, 56)  | 23167                      | 22091                   | 1.09X
BM_RHS_NAME(PackRhs, 32, 256, 256, 4, 16, 8, 8, 1, 1, 256, 56) | 23422                      | 23682                   | 0.99X
BM_RHS_NAME(PackRhs, 32, 256, 256, 4, 16, 8, 8, 2, 4, 256, 56) | 23165                      | 23663                   | 0.98X
BM_RHS_NAME(PackRhs, 32, 64, 64, 4, 16, 3, 3, 1, 1, 36, 432)   | 72689                      | 44969                   | 1.62X
BM_RHS_NAME(PackRhs, 32, 64, 64, 4, 16, 3, 3, 2, 2, 36, 432)   | 61732                      | 39779                   | 1.55X

All benchmarks on Intel Skylake server with 8 cores.
2019-04-20 06:46:43 +00:00
bench update wrt recent changes 2019-02-21 17:19:36 +01:00
blas Split the implementation of i?amax/min into two. Based on PR-627 by Sameer Agarwal. 2019-04-15 17:18:03 +02:00
cmake Simplify handling of tests that must fail to compile. 2018-12-12 15:48:36 +01:00
debug MIsc. source and comment typos 2018-03-11 10:01:44 -04:00
demos Fixed compilation error due to obsolete internal::abs and internal::sqrt function calls 2014-03-26 22:02:48 -04:00
doc erm.. use proper id 2019-03-12 13:53:38 +01:00
Eigen Adding lowlevel APIs for optimized RHS packet load in TensorFlow 2019-04-20 06:46:43 +00:00
failtest PR 572: Add initializer list constructors to Matrix and Array (include unit tests and doc) 2019-01-21 16:25:57 +01:00
lapack Enable "old" CMP0026 policy (not perfect, but better than dozens of warning) 2018-12-08 18:59:51 +01:00
scripts Simplify handling and non-splitted tests and include split_test_helper.h instead of re-generating it. This also allows us to modify it without breaking existing build folder. 2018-07-16 18:55:40 +02:00
test Adding lowlevel APIs for optimized RHS packet load in TensorFlow 2019-04-20 06:46:43 +00:00
unsupported Adding lowlevel APIs for optimized RHS packet load in TensorFlow 2019-04-20 06:46:43 +00:00
.hgeol Added a pattern which forces LF line endings for *.sh files. 2013-07-31 18:20:58 +02:00
.hgignore ignore all *build* sub directories 2017-12-14 14:22:14 +01:00
CMakeLists.txt Bypass inline asm for non compatible compilers. 2019-01-23 23:43:13 +01:00
COPYING.BSD Intel(R) MKL support added. 2011-12-05 14:52:21 +07:00
COPYING.GPL there's no reason why we should follow the FSF's stupid recommendation for the naming of these files, right? This could give the wrong impression that Eigen is only GPL-licensed. 2009-11-14 23:26:07 -05:00
COPYING.LGPL Replace COPYING.LGPL by a copy of the LGPL 2.1 (instead of LGPL 3). 2012-09-10 13:27:44 -04:00
COPYING.MINPACK add COPYING.MINPACK 2012-07-15 11:46:22 -04:00
COPYING.MPL2 add COPYING.MPL2 2012-07-15 10:20:59 -04:00
COPYING.README Replace COPYING.LGPL by a copy of the LGPL 2.1 (instead of LGPL 3). 2012-09-10 13:27:44 -04:00
CTestConfig.cmake Optimize the product of a householder-sequence with the identity, and optimize the evaluation of a HouseholderSequence to a dense matrix using faster blocked product. 2018-07-11 17:16:50 +02:00
CTestCustom.cmake.in Allow to filter out build-error messages 2018-07-24 20:12:49 +02:00
eigen3.pc.in Further fixes for CMAKE_INSTALL_PREFIX correctness 2015-11-07 21:29:24 -05:00
INSTALL finally, the right fix: set CTEST_BUILD_TARGET. 2009-10-04 20:27:44 -04:00
README.md Add links where to make PRs and report bugs into README.md 2018-04-13 21:05:28 +00:00
signature_of_eigen3_matrix_library improve the scripts for building unit tests: 2009-11-25 21:26:37 -05:00

README.md

Eigen is a C++ template library for linear algebra: matrices, vectors, numerical solvers, and related algorithms.

For more information go to http://eigen.tuxfamily.org/.

For pull request please only use the official repository at https://bitbucket.org/eigen/eigen.

For bug reports and feature requests go to http://eigen.tuxfamily.org/bz.