Commit Graph

7896 Commits

Author SHA1 Message Date
Rasmus Munk Larsen
235e83aba6 Eigen cost model part 1. This implements a basic recursive framework to estimate the cost of evaluating tensor expressions. 2016-04-14 13:57:35 -07:00
Gael Guennebaud
68897c52f3 Add extreme values to the imaginary part for SVD unit tests. 2016-04-14 22:47:30 +02:00
Gael Guennebaud
20f387fafa Improve numerical robustness of JacoviSVD:
- avoid noise amplification in complex to real conversion
 - compare off-diagonal entries to the current biggest diagonal entry: no need to bother about a 2x2 block containing ridiculously small entries compared to the rest of the matrix.
2016-04-14 22:46:55 +02:00
Benoit Steiner
7718749fee Force the inlining of the << operator on half floats 2016-04-14 11:51:54 -07:00
Benoit Steiner
5379d2b594 Inline the << operator on half floats 2016-04-14 11:40:48 -07:00
Benoit Steiner
5912ad877c Silenced a compilation warning 2016-04-14 11:40:14 -07:00
Benoit Steiner
2b6e3de02f Added tests to validate flooring and ceiling of fp16 2016-04-14 11:39:18 -07:00
Benoit Steiner
6f23e945f6 Added simple test for numext::sqrt and numext::pow on fp16 2016-04-14 10:32:52 -07:00
Benoit Steiner
72510c80e1 Added basic test for trigonometric functions on fp16 2016-04-14 10:27:24 -07:00
Benoit Steiner
7b3d7acebe Added support for fp16 to test_isApprox, test_isMuchSmallerThan, and test_isApproxOrLessThan 2016-04-14 10:25:50 -07:00
Benoit Steiner
5c13765ee3 Added ability to printf fp16 2016-04-14 10:24:52 -07:00
Benoit Steiner
c7167fee0e Added support for fp16 to the sigmoid function 2016-04-14 10:08:33 -07:00
Benoit Steiner
f6003f0873 Made the test msvc friendly 2016-04-14 09:47:26 -07:00
Gael Guennebaud
3551dea887 Cleaning pass on rcond estimator. 2016-04-14 16:45:41 +02:00
Gael Guennebaud
d8a3bdaa24 remove useless include 2016-04-14 15:18:56 +02:00
Gael Guennebaud
d402adc3d7 Better use .data() than &coeffRef(0) 2016-04-14 15:18:08 +02:00
Gael Guennebaud
ea7087ef31 Merged in rmlarsen/eigen (pull request PR-174)
Add matrix condition number estimation module.
2016-04-14 15:11:33 +02:00
Benoit Steiner
36f5a10198 Properly gate the definition of the error and gamma functions for fp16 2016-04-13 18:44:48 -07:00
Benoit Steiner
10b69810d1 Improved support for trigonometric functions on GPU 2016-04-13 16:00:51 -07:00
Benoit Steiner
d6105b53b8 Added basic implementation of the lgamma, digamma, igamma, igammac, polygamma, and zeta function for fp16 2016-04-13 15:26:02 -07:00
Gael Guennebaud
703251f10f merge 2016-04-13 23:45:10 +02:00
Gael Guennebaud
39211ba46b Fix JacobiSVD for complex when the complex-to-real update already gives a diagonal 2x2 block. 2016-04-13 23:43:26 +02:00
Benoit Steiner
2986253259 Cleaned up the implementation of digamma 2016-04-13 14:24:06 -07:00
Benoit Steiner
d5de1a8220 Pulled latest updates from trunk 2016-04-13 14:17:11 -07:00
Benoit Steiner
87ca15c4e8 Added support for sin, cos, tan, and tanh on fp16 2016-04-13 14:12:38 -07:00
Gael Guennebaud
2c9e4fa417 Add debug output for random unit test 2016-04-13 22:56:12 +02:00
Gael Guennebaud
7d1391d049 Turn a converge check to a warning 2016-04-13 22:50:54 +02:00
Gael Guennebaud
feef39e2d1 Fix underflow in JacoviSVD's complex to real preconditioner 2016-04-13 22:49:51 +02:00
Gael Guennebaud
f4e12272f1 Fix corner case in unit test. 2016-04-13 22:18:02 +02:00
Gael Guennebaud
a95e1a273e Fix warning in unit tests 2016-04-13 22:00:38 +02:00
Benoit Steiner
bf3f6688f0 Added support for computing cos, sin, tan, and tanh on GPU. 2016-04-13 11:55:08 -07:00
Benoit Steiner
473c8380ea Added constructors to convert unsigned integers into fp16 2016-04-13 11:03:37 -07:00
Gael Guennebaud
42a3352a3b Workaround a division by zero when outerstride==0 2016-04-13 19:02:02 +02:00
Gael Guennebaud
6f960b83ff Make use of is_same_dense helper instead of extract_data to detect input/outputs are the same. 2016-04-13 18:47:12 +02:00
Gael Guennebaud
b7716c0328 Fix incomplete previous patch on matrix comparision. 2016-04-13 18:32:56 +02:00
Gael Guennebaud
2630d97c62 Fix detection of same matrices when both matrices are not handled by extract_data. 2016-04-13 18:26:08 +02:00
Gael Guennebaud
512ba0ac76 Add regression unit tests for half-packet vectorization 2016-04-13 18:16:35 +02:00
Gael Guennebaud
06447e0a39 Improve half-packet vectorization logic to distinguish linear versus inner traversal modes. 2016-04-13 18:15:49 +02:00
Gael Guennebaud
bbb8854bf7 Enable half-packet in reduxions. 2016-04-13 13:02:34 +02:00
Benoit Steiner
e9b12cc1f7 Fixed compilation warnings generated by clang 2016-04-12 20:53:18 -07:00
Benoit Steiner
eaeb6ca93a Enable the benchmarks for algebraic and transcendental fnctions on fp16. 2016-04-12 16:29:00 -07:00
Benoit Steiner
aa1ba8bbd2 Don't put a command at the end of an enumerator list 2016-04-12 16:28:11 -07:00
Benoit Steiner
e49945ced4 Pulled latest update from trunk 2016-04-12 14:13:41 -07:00
Benoit Steiner
25d05c4b8f Fixed the vectorization logic test 2016-04-12 14:13:25 -07:00
Benoit Steiner
53121c0119 Turned on the contraction benchmarks for fp16 2016-04-12 14:11:52 -07:00
Gael Guennebaud
b67c983291 Enable the use of half-packet in coeff-based product.
For instance, Matrix4f*Vector4f is now vectorized again when using AVX.
2016-04-12 23:03:03 +02:00
Benoit Steiner
e3a184785c Fixed the zeta test 2016-04-12 11:12:36 -07:00
Benoit Steiner
3b76df64fc Defer the decision to vectorize tensor CUDA code to the meta kernel. This makes it possible to decide to vectorize or not depending on the capability of the target cuda architecture. In particular, this enables us to vectorize the processing of fp16 when running on device of capability >= 5.3 2016-04-12 10:58:51 -07:00
Rasmus Larsen
6498dadc2f Merged eigen/eigen into default 2016-04-11 17:42:05 -07:00
Benoit Steiner
748c4c4599 More accurate cost estimates for exp, log, tanh, and sqrt. 2016-04-11 13:11:04 -07:00