This implements `Eigen::internal::assume_aligned` to match the API for C++20 standard as best as possible using either `std::assume_aligned` or `__builtin_assume_aligned` if available. If neither is available, the function is a no-op.
The override macro `EIGEN_ASSUME_ALIGNED` was changed to a `EIGEN_DONT_ASSUME_ALIGNED`, which now forces the function to be a no-op.
See merge request libeigen/eigen!2052
(cherry picked from commit 8716f109e4)
f8191848 Fix pcmp_* for HVX to to comply with the new definition of true = Scalar(1).
7cc169d9 Revert "Fix pcmp_* for HVX to to comply with the new definition of true = Scalar(1)."
06999845 Merge branch eigen:master into master
b5fde61f Merge branch eigen:master into master
10d10d60 Merge branch eigen:master into master
9398d6ad Merge branch eigen:master into master
7804d5a4 Merge branch eigen:master into master
0068623c Merge branch eigen:master into master
b0ffc9cf Merge branch eigen:master into master
f3791c80 Merge branch eigen:master into master
74275f0c Merge branch eigen:master into master
b095614e Merge branch eigen:master into master
1312a696 Merge branch eigen:master into master
e6dd44d2 Merge branch eigen:master into master
8ac67769 Implement assume_aligned using the standard API if available.
97b299fa Format.
b31798be Fix typos.
04b3d312 Unformat.
Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
Activates vectorization of the Eigen::half versions of the tanh and
logistic functions when they run on Neon. Both functions convert their
inputs to float before computing the output, and as a result of this
commit, the conversions and the computation in float are vectorized.
GCC 4.8 doesn't seem to like the `g` register constraint, failing to
compile with "error: 'asm' operand requires impossible reload".
Tested `r` instead, and that seems to work, even with latest compilers.
Also fixed some minor macro issues to eliminate warnings on armv7.
Fixes#2315.
