tree 14213030ab815330430b78c30ae7de57f9aa5f2c
parent 82d61375825a507ddcf4ff157ba790556dc99376
author Rasmus Munk Larsen <rmlarsen@google.com> 1673049419 -0800
committer Antonio Sanchez <cantonios@google.com> 1674452670 -0800

Update Eigen to commit:3564668908afc66351c1c3cc47dca2fcdb91dc12

CHANGELOG
=========
356466890 - Fix overalign check.
f3929ac7e - Fix EIGEN_HAS_CXX17_OVERALIGN for icc
1b33a6374 - Fixes git add . doesn'\''t include scripts/buildtests.in
a8bab0d8a - Patch SparseLU
910f6f65d - Adjust thresholds for bfloat16 product tests that are currently failing.
311cc0f9c - Enable NEON pcmp, plset, and complex psqrt
dbf7ae6f9 - Fix up C++ version detection macros and cmake tests.
bb6675caf - Fix incorrect NEON native fp16 multiplication.
dd85d2694 - Revert "Avoid mixing types in CompressedStorage.h"
c4fb6af24 - Enable NEON pabs for unsigned int types
400bc5cd5 - Add sparse_basic_1 to smoke tests.
04e4f0bb2 - Add missing colon in SparseMatrix.h.
3d8a8def8 - Avoid mixing types in CompressedStorage.h
4bb244679 - Add operators to CompressedStorageIterator
e1aee4ab3 - Update test of numext::signbit.
3717854a2 - Use numext::signbit instead of std::signbit, which is not defined for bfloat16.
37de43290 - Avoid using std::raise() for divide by zero
62de593c4 - Allow std::initializer_list constructors in constexpr expressions
6d3e3678b - optimize equalspace packetop
200483194 - add EqualSpaced / setEqualSpaced
273f80384 - Add BDCSVD_LAPACKE binding
03c9b4738 - Enable direct access for NestByValue.
b59f18b4f - Increase L2 and L3 cache size for Power10.
c614b2bbd - Fix index type for sparse index sorting.
44fe53915 - add sparse sort inner vectors function
d19416714 - Fix the bug using neon instruction fmla for data type half
31ab62d34 - Add support for Power10 (AltiVec) MMA instructions for bfloat16.
dcb042a87 - Fix serialization for non-compressed matrices.
2260e11eb - Fix reshape strides when input has non-zero inner stride.
23524ab6f - Changing BiCGSTAB parameters initialization so that it works with custom types
ab2b26fbc - Fix sparseLU solver when destination has a non-unit stride.
551eebc8c - Add synchronize method to all devices.
b7551bff9 - Fix a bunch of annoying compiler warnings in tests
e7b1ad031 - Add serialization for sparse matrix and sparse vector.
044f3f623 - Fix bug in handmade_aligned_realloc
672868393 - Small cleanup of IDRS.h
02805bd56 - Fix AVX2 psignbit
399ce1ed6 - Fix duplicate execution code for Power 8 Altivec in pstore_partial.
6431dfdb5 - Cross product for vectors of size 2. Fixes #1037
8588d8c74 - Correct pnegate for floating-point zero.
5eacb9e11 - Put brackets around unsigned type names.
37e40dca8 - Fix ambiguity in PPC for vec_splats call.
7dc6db75d - Fix typo in CholmodSupport
9b6d624ea - fix neon
7e398e943 - Add missing return keyword in psignbit for NEON.
82b152dbe - Add signbit function
8f8e36458 - Remove recently added sparse assert in SparseMapBase.
01a31b81b - Remove unused parameter name.
c5b896c5a - Allow empty matrices to be resized.
886aad136 - Disable patan for double on PPC.
ab407b2b6 - Fix handmade_aligned_malloc offset computation.
adb30efb2 - Add assert for invalid outerIndexPtr array in SparseMapBase.
c27d1abe4 - Fix pragma check for disabling fastmath.
a22637137 - Change handmade_aligned_malloc/realloc/free to store a 1 byte offset instead of absolute address
bf48d4633 - Explicitly state that indices must be sorted.

PiperOrigin-RevId: 500278312
Change-Id: I5ba2533555adb8e254879b45d370bb8fccb74c8f