devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity
9,828 Commits 15 Branches 93 Tags
ea8fa5e86f377cde0bd29705ed8a7e4c58a0f83a
Commit Graph

217 Commits

Author SHA1 Message Date
Deven Desai
b6cc0961b1 updates based on PR feedback 
There are two major changes (and a few minor ones which are not listed here...see PR discussion for details)

1. Eigen::half implementations for HIP and CUDA have been merged.
This means that
- `CUDA/Half.h` and `HIP/hcc/Half.h` got merged to a new file `GPU/Half.h`
- `CUDA/PacketMathHalf.h` and `HIP/hcc/PacketMathHalf.h` got merged to a new file `GPU/PacketMathHalf.h`
- `CUDA/TypeCasting.h` and `HIP/hcc/TypeCasting.h` got merged to a new file `GPU/TypeCasting.h`

After this change the `HIP/hcc` directory only contains one file `math_constants.h`. That will go away too once that file becomes a part of the HIP install.

2. new macros EIGEN_GPUCC, EIGEN_GPU_COMPILE_PHASE and EIGEN_HAS_GPU_FP16 have been added and the code has been updated to use them where appropriate.
- `EIGEN_GPUCC` is the same as `(EIGEN_CUDACC || EIGEN_HIPCC)`
- `EIGEN_GPU_DEVICE_COMPILE` is the same as `(EIGEN_CUDA_ARCH || EIGEN_HIP_DEVICE_COMPILE)`
- `EIGEN_HAS_GPU_FP16` is the same as `(EIGEN_HAS_CUDA_FP16 or EIGEN_HAS_HIP_FP16)`
 2018-06-14 10:21:54 -04:00
Deven Desai
8fbd47052b Adding support for using Eigen in HIP kernels. 
This commit enables the use of Eigen on HIP kernels / AMD GPUs. Support has been added along the same lines as what already exists for using Eigen in CUDA kernels / NVidia GPUs.

Application code needs to explicitly define EIGEN_USE_HIP when using Eigen in HIP kernels. This is because some of the CUDA headers get picked up by default during Eigen compile (irrespective of whether or not the underlying compiler is CUDACC/NVCC, for e.g. Eigen/src/Core/arch/CUDA/Half.h). In order to maintain this behavior, the EIGEN_USE_HIP macro is used to switch to using the HIP version of those header files (see Eigen/Core and unsupported/Eigen/CXX11/Tensor)


Use the "-DEIGEN_TEST_HIP" cmake option to enable the HIP specific unit tests.
 2018-06-06 10:12:58 -04:00
Christoph Hertzberg
42715533f1 bug #1493: Make representation of HouseholderSequence consistent and working for complex numbers. Made corresponding unit test actually test that. Also simplify implementation of QR decompositions 2018-04-15 10:15:28 +02:00
luz.paz
e3912f5e63 MIsc. source and comment typos 
Found using `codespell` and `grep` from downstream FreeCAD
 2018-03-11 10:01:44 -04:00
Gael Guennebaud
09a16ba42f bug #1412: fix compilation with nvcc+MSVC 2018-01-17 23:13:16 +01:00
Eugene Chereshnev
f558ad2955 Fix incorrect ldvt in LAPACKE call from JacobiSVD 2018-01-03 12:55:52 -08:00
Gael Guennebaud
4c03b3511e Fix issue with boost::multiprec in previous commit 2017-11-08 23:28:01 +01:00
Gael Guennebaud
e9d2888e74 Improve debugging tests and output in BDCSVD 2017-11-08 10:26:03 +01:00
Gael Guennebaud
e8468ea91b Fix overflow issues in BDCSVD 2017-11-08 10:24:28 +01:00
Gael Guennebaud
bbd97b4095 Add a EIGEN_NO_CUDA option, and introduce EIGEN_CUDACC and EIGEN_CUDA_ARCH aliases 2017-07-17 01:02:51 +02:00
Gael Guennebaud
731c8c704d bug #1403: more scalar conversions fixes in BDCSVD 2017-06-09 15:45:49 +02:00
Benoit Steiner
09ae0e6586 Adjusted the EIGEN_DEVICE_FUNC qualifiers to make sure that: 
* they're used consistently between the declaration and the definition of a function
  * we avoid calling host only methods from host device methods.
 2017-03-01 11:47:47 -08:00
Gael Guennebaud
6572825703 bug #1395: fix the use of compile-time vectors as inputs of JacobiSVD. 2017-02-20 13:44:37 +01:00
Gael Guennebaud
645a8e32a5 Fix compilation of JacobiSVD for vectors type 2017-01-31 16:22:54 +01:00
Niels Ole Salscheider
51fef87408 Make sure not to call numext::maxi on expression templates 2016-11-12 12:20:57 +01:00
Gael Guennebaud
fa9049a544 Let be consistent and consider any denormal number as zero. 2016-09-15 11:24:03 +02:00
Gael Guennebaud
3cb914f332 bug #1266: remove CUDA guards on MatrixBase::<decomposition> definitions. (those used to break old nvcc versions that we propably don't care anymore) 2016-09-06 09:55:50 +02:00
Gael Guennebaud
35a8e94577 bug #1167: simplify installation of header files using cmake's install(DIRECTORY ...) command. 2016-08-29 10:59:37 +02:00
Gael Guennebaud
9c663e4ee8 Clean references to MKL in LAPACKe support. 2016-07-25 18:20:08 +02:00
Gael Guennebaud
0c06077efa Rename MKL files 2016-07-25 18:00:47 +02:00
Gael Guennebaud
4d54e3dd33 bug #173: remove dependency to MKL for LAPACKe backend. 2016-07-25 17:55:07 +02:00
Gael Guennebaud
1b2049fbda Enforce scalar types in calls to max/min (helps with expression template scalar types) 2016-07-25 12:35:10 +02:00
Gael Guennebaud
5ea0864c81 Fix regression in a previous commit: some diagonal entry might not be treated by the 2x2 real preconditioner. 2016-07-13 18:37:54 +02:00
Gael Guennebaud
b4343aa67e Avoid division by very small entries when extracting singularvalues, and explicitly handle the 1x1 complex case. 2016-07-12 17:22:03 +02:00
Gael Guennebaud
c1f9ca9254 Update RealQZ to reduce 2x2 diagonal block of T corresponding to non reduced diagonal block of S to positive diagonal form. 
This step involve a real 2x2 SVD problem. The respective routine is thus in src/misc/ to be shared by both EVD and AVD modules.
 2016-06-09 17:11:03 +02:00
Christoph Hertzberg
dacb469bc9 Enable and fix -Wdouble-conversion warnings 2016-05-05 13:35:45 +02:00
Gael Guennebaud
75a94b9662 Improve documentation of BDCSVD 2016-05-04 12:53:14 +02:00
Gael Guennebaud
e2ca478485 bug #1214: consider denormals as zero in D&C SVD. This also workaround infinite binary search when compiling with ICC's unsafe optimizations. 2016-05-03 23:15:29 +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
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
Gael Guennebaud
feef39e2d1 Fix underflow in JacoviSVD's complex to real preconditioner 2016-04-13 22:49:51 +02:00
Benoit Steiner
81d340984a Removed executable bit from header files 2016-03-23 16:15:02 -07:00
Gael Guennebaud
8b0d1eb0f7 Fix numerous doxygen shortcomings, and workaround some clang -Wdocumentation warnings 2016-01-01 21:45:06 +01:00
Gael Guennebaud
29bb599e03 Fix numerous doxygen issues in auto-link generation 2015-12-30 16:04:24 +01:00
Gael Guennebaud
774dba87c8 merge 2015-12-11 23:28:44 +01:00
Gael Guennebaud
b60a8967f5 bug #1134: fix JacobiSVD pre-allocation 
(grafted from f22036f5f8
)
 2015-12-11 11:59:11 +01:00
Gael Guennebaud
30b5c4cd14 Remove useless "explicit", and fix inline/static order. 2015-12-11 10:59:39 +01:00
Benoit Steiner
06a22ca5bd Added support for sigmoid function to the tensor module 2015-07-17 09:29:00 -07:00
Gael Guennebaud
06036d8bb1 Fix compilation of BDCSVD with DEFAULT_TO_ROWMAJOR 2015-06-19 10:37:25 +02:00
Gael Guennebaud
4aba24a1b2 Clean argument names of some functions 2015-06-09 13:32:12 +02:00
Gael Guennebaud
9a2447b0c9 Fix shadow warnings triggered by clang 2015-06-09 09:11:12 +02:00
Gael Guennebaud
3c38589984 Remove most of the dynamic memory allocations that occured in D&C SVD. Still remains the calls to JacobiSVD and UpperBidiagonalization. 2015-03-31 22:54:47 +02:00
Gael Guennebaud
35d3053d55 Fix regression introduced in 3b169d792d 2015-03-31 09:23:53 +02:00
Christoph Hertzberg
3b169d792d Suppress unused variable warning 2015-03-31 00:49:08 +02:00
Christoph Hertzberg
58af8bf90c bug #982: Make sure numext::maxi and numext::mini are called correctly, in case Scalar expressions return expression templates. 2015-03-30 16:47:22 +02:00
Gael Guennebaud
1b8cc9af43 Slight numerical stability improvement in 2x2 svd 2015-03-27 10:55:00 +01:00
Gael Guennebaud
d27968eb7e D&C SVD: directly falls back to JacobiSVD for very small problems (by-pass upper-bidiagonalization) 2015-03-24 13:38:07 +01:00
Gael Guennebaud
4472f3e578 Avoid SVD: consider denormalized small numbers as zero when computing the rank of the matrix 2015-03-23 09:40:21 +01:00
Gael Guennebaud
8580eb6808 bug #949: add static assertion for incompatible scalar types in dense end-user decompositions. 2015-03-13 21:06:20 +01:00
Gael Guennebaud
1c0e8bcf09 Fix unused variable warning. 2015-02-16 17:21:30 +01:00