The original `fill` implementation introduced a 5x regression on my
nvidia Quadro K1200. @rohitsan reported up to 100x regression for
HIP. This restores performance.
For custom scalars, zero is not necessarily represented by
a zeroed-out memory block (e.g. gnu MPFR). We therefore
cannot rely on `memset` if we want to fill a matrix or tensor
with zeroes. Instead, we should rely on `fill`, which for trivial
types does end up getting converted to a `memset` under-the-hood
(at least with gcc/clang).
Requires adding a `fill(begin, end, v)` to `TensorDevice`.
Replaced all potentially bad instances of memset with fill.
Fixes#2245.
Allows absolute and relative paths for
- `INCLUDE_INSTALL_DIR`
- `CMAKEPACKAGE_INSTALL_DIR`
- `PKGCONFIG_INSTALL_DIR`
Type should be `PATH` not `STRING`. Contrary to !211, these don't
seem to be made absolute if user-defined - according to the doc any
directories should use `PATH` type, which allows a file dialog
to be used via the GUI. It also better handles file separators.
If user provides an absolute path, it will be made relative to
`CMAKE_INSTALL_PREFIX` so that the `configure_packet_config_file` will
work.
Fixes#2155 and #2269.
The extra [TOC] tag is generating a huge floating duplicated
table-of-contents, which obscures the majority of the page
(see bottom of https://eigen.tuxfamily.org/dox/unsupported/eigen_tensors.html).
Remove it.
Also, headers do not support markup (see
[doxygen bug](https://github.com/doxygen/doxygen/issues/7467)), so
backticks like
```
```
end up generating titles that looks like
```
Constructor <tt>Tensor<double,2></tt>
```
Removing backticks for now. To generate proper formatted headers, we
must directly use html instead of markdown, i.e.
```
<h2>Constructor <code>Tensor<double,2></code></h2>
```
which is ugly.
Fixes#2254.
- Move constructors can only be defaulted as NOEXCEPT if all members
have NOEXCEPT move constructors.
- gcc 4.8 has some funny parsing bug in `a < b->c`, thinking `b-` is a template parameter.
For empty or single-column matrices, the current `PartialPivLU`
currently dereferences a `nullptr` or accesses memory out-of-bounds.
Here we adjust the checks to avoid this.
As written, depending on multithreading/gpu, the returned index from
`argmin`/`argmax` is not currently stable. Here we modify the functors
to always keep the first occurence (i.e. if the value is equal to the
current min/max, then keep the one with the smallest index).
This is otherwise causing unpredictable results in some TF tests.
We can't make guarantees on alignment for existing calls to `pset`,
so we should default to loading unaligned. But in that case, we should
just use `ploadu` directly. For loading constants, this load should hopefully
get optimized away.
This is causing segfaults in Google Maps.
MinGW spits out version strings like: `x86_64-w64-mingw32-g++ (GCC)
10-win32 20210110`, which causes the version extraction to fail.
Added support for this with tests.
Also added `make_unsigned` for `long long`, since mingw seems to
use that for `uint64_t`.
Related to #2268. CMake and build passes for me after this.
Currently TF lite needs to hack around with the Tensor headers in order
to customize the contraction dispatch method. Here we add simple `#ifndef`
guards to allow them to provide their own dispatch prior to inclusion.
This used to work for non-class types (e.g. raw function pointers) in
Eigen 3.3. This was changed in commit 11f55b29 to optimize the
evaluator:
> `sizeof((A-B).cwiseAbs2())` with A,B Vector4f is now 16 bytes, instead of 48 before this optimization.
though I cannot reproduce the 16 byte result. Both before the change
and after, with multiple compilers/versions, I always get a result of 40 bytes.
https://godbolt.org/z/MsjTc1PGe
This change modifies the code slightly to allow non-class types. The
final generated code is identical, and the expression remains 40 bytes
for the `abs2` sample case.
Fixes#2251
