The original fails with nvcc+msvc - there's a static order of initialization
issue leading to registered tests being cleared. The test then fails on
```
VERIFY(EigenTest::all().size()>0);
```
since `EigenTest` no longer contains any tests. The singleton pattern
fixes this.
Replace usage of `std::numeric_limits<...>::min/max_exponent` in
codebase where possible. Also replaced some other `numeric_limits`
usages in affected tests with the `NumTraits` equivalent.
The previous MR !443 failed for c++03 due to lack of `constexpr`.
Because of this, we need to keep around the `std::numeric_limits`
version in enum expressions until the switch to c++11.
Fixes#2148
Replace usage of `std::numeric_limits<...>::min/max_exponent` in
codebase. Also replaced some other `numeric_limits` usages in
affected tests with the `NumTraits` equivalent.
Fixes#2148
NVCC does not understand `__forceinline`, so we need to use `inline`
when compiling for GPU.
ICC specializes `std::complex` operators for `float` and `double`
by default, which cannot be used on device and conflict with Eigen's
workaround in CUDA/Complex.h. This can be prevented by defining
`_OVERRIDE_COMPLEX_SPECIALIZATION_` before including `<complex>`.
Added this define to the tests and to `Eigen/Core`, but this will
not work if the user includes `<complex>` before `<Eigen/Core>`.
ICC also seems to generate a duplicate `Map` symbol in
`PlainObjectBase`:
```
error: "Map" has already been declared in the current scope
static ConstMapType Map(const Scalar *data)
```
I tracked this down to `friend class Eigen::Map`. Putting the `friend`
statements at the bottom of the class seems to resolve this issue.
Fixes#2180
The original swap approach leads to potential undefined behavior (reading
uninitialized memory) and results in unnecessary copying of data for static
storage.
Here we pass down the move assignment to the underlying storage. Static
storage does a one-way copy, dynamic storage does a swap.
Modified the tests to no longer read from the moved-from matrix/tensor,
since that can lead to UB. Added a test to ensure we do not access
uninitialized memory in a move.
Fixes: #2119
The macro `__cplusplus` is not defined correctly in MSVC unless building
with the the `/Zc:__cplusplus` flag. Instead, it defines `_MSVC_LANG` to the
specified c++ standard version number.
Here we introduce `EIGEN_CPLUSPLUS` which will contain the c++ version
number both for MSVC and otherwise. This simplifies checks for supported
features.
Also replaced most instances of standard version checking via `__cplusplus`
with the existing `EIGEN_COMP_CXXVER` macro for better clarity.
Fixes: #2170
This is a new version of !423, which failed for MSVC.
Defined `EIGEN_OPTIMIZATION_BARRIER(X)` that uses inline assembly to
prevent operations involving `X` from crossing that barrier. Should
work on most `GNUC` compatible compilers (MSVC doesn't seem to need
this). This is a modified version adapted from what was used in
`psincos_float` and tested on more platforms
(see #1674, https://godbolt.org/z/73ezTG).
Modified `rint` to use the barrier to prevent the add/subtract rounding
trick from being optimized away.
Also fixed an edge case for large inputs that get bumped up a power of two
and ends up rounding away more than just the fractional part. If we are
over `2^digits` then just return the input. This edge case was missed in
the test since the test was comparing approximate equality, which was still
satisfied. Adding a strict equality option catches it.
It seems *sometimes* with aggressive optimizations the combination
`psub(padd(a, b), b)` trick to force rounding is compiled away. Here
we replace with inline assembly to prevent this (I tried `volatile`,
but that leads to additional loads from memory).
Also fixed an edge case for large inputs `a` where adding `b` bumps
the value up a power of two and ends up rounding away more than
just the fractional part. If we are over `2^digits` then just return
the input. This edge case was missed in the test since the test was
comparing approximate equality, which was still satisfied. Adding
a strict equality option catches it.
In SSE, by adding/subtracting 2^MantissaBits, we force rounding according to the
current rounding mode.
For NEON, we use the provided intrinsics for rint/floor/ceil if
available (armv8).
Related to #1969.
Since `numeric_limits<half>::max_exponent` is a static inline constant,
it cannot be directly passed by reference. This triggers a linker error
in recent versions of `g++-powerpc64le`.
Changing `half` to take inputs by value fixes this. Wrapping
`max_exponent` with `int(...)` to make an addressable integer also fixes this
and may help with other custom `Scalar` types down-the-road.
Also eliminated some compile warnings for powerpc.
With !406, we accidentally broke arm 32-bit NEON builds, since
`vsqrt_f32` is only available for 64-bit.
Here we add back the `rsqrt` implementation for 32-bit, relying
on a `prsqrt` implementation with better handling of edge cases.
Note that several of the 32-bit NEON packet tests are currently
failing - either due to denormal handling (NEON versions flush
to zero, but scalar paths don't) or due to accuracy (e.g. sin/cos).
The original will saturate if the input does not fit into an integer
type. Here we fix this, returning the input if it doesn't have
enough precision to have a fractional part.
Also added `pceil` for NEON.
Fixes#1969.
CMake complains that the package name does not match when the case
differs, e.g.:
```
CMake Warning (dev) at /usr/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:273 (message):
The package name passed to `find_package_handle_standard_args` (UMFPACK)
does not match the name of the calling package (Umfpack). This can lead to
problems in calling code that expects `find_package` result variables
(e.g., `_FOUND`) to follow a certain pattern.
Call Stack (most recent call first):
cmake/FindUmfpack.cmake:50 (find_package_handle_standard_args)
bench/spbench/CMakeLists.txt:24 (find_package)
This warning is for project developers. Use -Wno-dev to suppress it.
```
Here we rename the libraries to match their true cases.
The previous implementations produced garbage values if the exponent did
not fit within the exponent bits. See #2131 for a complete discussion,
and !375 for other possible implementations.
Here we implement the 4-factor version. See `pldexp_impl` in
`GenericPacketMathFunctions.h` for a full description.
The SSE `pcmp*` methods were moved down since `pcmp_le<Packet4i>`
requires `por`.
Left as a "TODO" is to delegate to a faster version if we know the
exponent does fit within the exponent bits.
Fixes#2131.
The workaround_9220 function was introduced a long time ago to
workaround a CMake issue with enable_language(OPTIONAL). Since then
CMake has clarified that the OPTIONAL keywords has not been
implemented[0].
A CheckLanguage module is now provided with CMake to check if a language
can be enabled. Use that instead.
[0] https://cmake.org/cmake/help/v3.18/command/enable_language.html
The new `generic_pow` implementation was failing for half/bfloat16 since
their construction from int/float is not `constexpr`. Modified
in `GenericPacketMathFunctions` to remove `constexpr`.
While adding tests for half/bfloat16, found other issues related to
implicit conversions.
Also needed to implement `numext::arg` for non-integer, non-complex,
non-float/double/long double types. These seem to be implicitly
converted to `std::complex<T>`, which then fails for half/bfloat16.
NVCC and older versions of clang do not fully support `std::complex` on device,
leading to either compile errors (Cannot call `__host__` function) or worse,
runtime errors (Illegal instruction). For most functions, we can
implement specialized `numext` versions. Here we specialize the standard
operators (with the exception of stream operators and member function operators
with a scalar that are already specialized in `<complex>`) so they can be used
in device code as well.
To import these operators into the current scope, use
`EIGEN_USING_STD_COMPLEX_OPERATORS`. By default, these are imported into
the `Eigen`, `Eigen:internal`, and `Eigen::numext` namespaces.
This allow us to remove specializations of the
sum/difference/product/quotient ops, and allow us to treat complex
numbers like most other scalars (e.g. in tests).
The recent addition of vectorized pow (!330) relies on `pfrexp` and
`pldexp`. This was missing for `Eigen::half` and `Eigen::bfloat16`.
Adding tests for these packet ops also exposed an issue with handling
negative values in `pfrexp`, returning an incorrect exponent.
Added the missing implementations, corrected the exponent in `pfrexp1`,
and added `packetmath` tests.
Test enters an infinite loop if size is 1x1 when choosing to select
unique indices for adding `inf` and `NaN` to the input. Here we
revert to non-unique indices, and split the `hypotNorm` check into
two cases: one where both `inf` and `NaN` are added, and one where
only `NaN` is added.
I ran some testing (comparing to `std::pow(double(x), double(y)))` for `x` in the set of all (positive) floats in the interval `[std::sqrt(std::numeric_limits<float>::min()), std::sqrt(std::numeric_limits<float>::max())]`, and `y` in `{2, sqrt(2), -sqrt(2)}` I get the following error statistics:
```
max_rel_error = 8.34405e-07
rms_rel_error = 2.76654e-07
```
If I widen the range to all normal float I see lower accuracy for arguments where the result is subnormal, e.g. for `y = sqrt(2)`:
```
max_rel_error = 0.666667
rms = 6.8727e-05
count = 1335165689
argmax = 2.56049e-32, 2.10195e-45 != 1.4013e-45
```
which seems reasonable, since these results are subnormals with only couple of significant bits left.
MSVC incorrectly handles `inf` cases for `std::sqrt<std::complex<T>>`.
Here we replace it with a custom version (currently used on GPU).
Also fixed the `packetmath` test, which previously skipped several
corner cases since `CHECK_CWISE1` only tests the first `PacketSize`
elements.
MSVC's uniform random number generator is not quite as uniform as
others, requiring a slightly wider threshold on the histogram test.
After inspecting histograms for several runs, there's no obvious
bias -- just some bins end up having slightly more less elements
(often > 2% but less than 2.5%).
The existing `Ref` class failed to consider cases where the Ref's
`Stride` setting *could* match the underlying referred object's stride,
but **didn't** at runtime. This led to trying to set invalid stride values,
causing runtime failures in some cases, and garbage due to mismatched
strides in others.
Here we add the missing runtime checks. This involves computing the
strides necessary to align with the referred object's storage, and
verifying we can actually set those strides at runtime.
In the `const` case, if it *may* be possible to refer to the original
storage at compile-time but fails at runtime, then we defer to the
`construct(...)` method that makes a copy.
Added more tests to check these cases.
Fixes#2093.
This is to support scalar `sqrt` of complex numbers `std::complex<T>` on
device, requested by Tensorflow folks.
Technically `std::complex` is not supported by NVCC on device
(though it is by clang), so the default `sqrt(std::complex<T>)` function only
works on the host. Here we create an overload to add back the
functionality.
Also modified the CMake file to add `--relaxed-constexpr` (or
equivalent) flag for NVCC to allow calling constexpr functions from
device functions, and added support for specifying compute architecture for
NVCC (was already available for clang).
For these to exist we would need to define `_USE_MATH_DEFINES` before
`cmath` or `math.h` is first included. However, we don't
control the include order for projects outside Eigen, so even defining
the macro in `Eigen/Core` does not fix the issue for projects that
end up including `<cmath>` before Eigen does (explicitly or transitively).
To fix this, we define `EIGEN_LOG2E` and `EIGEN_LN2` ourselves.
This allows the `packetmath` tests to pass for AVX512 on skylake.
Made `half` and `bfloat16` consistent in terms of ops they support.
Note the `log` tests are currently disabled for `bfloat16` since
they fail due to poor precision (they were previously disabled for
`Packet8bf` via test function specialization -- I just removed that
specialization and disabled it in the generic test).
Allows exclusion of doc and related targets to help when using eigen via add_subdirectory().
Requested by:
https://gitlab.com/libeigen/eigen/-/issues/1842
Also required making EIGEN_TEST_BUILD_DOCUMENTATION a dependent option on EIGEN_BUILD_DOC. This ensures documentation targets are properly defined when EIGEN_TEST_BUILD_DOCUMENTATION is ON.
The `half_float` test was failing with `-mcpu=cortex-a55` (native `__fp16`) due
to a bad NaN bit-pattern comparison (in the case of casting a float to `__fp16`,
the signaling `NaN` is quieted). There was also an inconsistency between
`numeric_limits<half>::quiet_NaN()` and `NumTraits::quiet_NaN()`. Here we
correct the inconsistency and compare NaNs according to the IEEE 754
definition.
Also modified the `bfloat16_float` test to match.
Tested with `cortex-a53` and `cortex-a55`.
Minimal implementation of AVX `Eigen::half` ops to bring in line
with `bfloat16`. Allows `packetmath_13` to pass.
Also adjusted `bfloat16` packet traits to match the supported set
of ops (e.g. Bessel is not actually implemented).
The `half_float` test was failing with `-mcpu=cortex-a55` (native `__fp16`) due
to a bad NaN bit-pattern comparison (in the case of casting a float to `__fp16`,
the signaling `NaN` is quieted). There was also an inconsistency between
`numeric_limits<half>::quiet_NaN()` and `NumTraits::quiet_NaN()`. Here we
correct the inconsistency and compare NaNs according to the IEEE 754
definition.
Also modified the `bfloat16_float` test to match.
Tested with `cortex-a53` and `cortex-a55`.
The AVX half implementation is incomplete, causing the `packetmath_13` test
to fail. This disables the test.
Also refactored the existing AVX implementation to use `bit_cast`
instead of direct access to `.x`.
Multiplication of column-major `DynamicSparseMatrix`es involves three
temporaries:
- two for transposing twice to sort the coefficients
(`ConservativeSparseSparseProduct.h`, L160-161)
- one for a final copy assignment (`SparseAssign.h`, L108)
The latter is avoided in an optimization for `SparseMatrix`.
Since `DynamicSparseMatrix` is deprecated in favor of `SparseMatrix`, it's not
worth the effort to optimize further, so I simply disabled counting
temporaries via a macro.
Note that due to the inclusion of `sparse_product.cpp`, the `sparse_extra`
tests actually re-run all the original `sparse_product` tests as well.
We may want to simply drop the `DynamicSparseMatrix` tests altogether, which
would eliminate the test duplication.
Related to #2048
The existing `TensorRandom.h` implementation makes the assumption that
`half` (`bfloat16`) has a `uint16_t` member `x` (`value`), which is not
always true. This currently fails on arm64, where `x` has type `__fp16`.
Added `bit_cast` specializations to allow casting to/from `uint16_t`
for both `half` and `bfloat16`. Also added tests in
`half_float`, `bfloat16_float`, and `cxx11_tensor_random` to catch
these errors in the future.
The `meta` test generates warnings with the latest version of clang due
to passing uninitialized variables as const reference arguments.
```
test/meta.cpp:102:45: error: variable 'f' is uninitialized when passed as a const reference argument here [-Werror,-Wuninitialized-const-reference]
VERIFY(( check_is_convertible(a.dot(b), f) ));
```
We don't actually use the variables, but initializing them eliminates the
new warning.
Fixes#2067.
When calling `internal::cast<S, std::complex<T>>(x)`, clang often
generates an implicit conversion warning due to an implicit cast
from type `S` to `T`. This currently affects the following tests:
- `basicstuff`
- `bfloat16_float`
- `cxx11_tensor_casts`
The implicit cast leads to widening/narrowing float conversions.
Widening warnings only seem to be generated by clang (`-Wdouble-promotion`).
To eliminate the warning, we explicitly cast the real-component first
from `S` to `T`. We also adjust tests to use `internal::cast` instead
of `static_cast` when a complex type may be involved.
Starting with ROCm 4.0, the `hipconfig --platform` command will return `amd` (prior return value was `hcc`). Updating the CMakeLists.txt files in the test dirs to account for this change.
Armv8.2-a provides a native half-precision floating point (__fp16 aka.
float16_t). This patch introduces
* __fp16 as underlying type of Eigen::half if this type is available
* the packet types Packet4hf and Packet8hf representing float16x4_t and
float16x8_t respectively
* packet-math for the above packets with corresponding scalar type Eigen::half
The packet-math functionality has been implemented by Ashutosh Sharma
<ashutosh.sharma@amperecomputing.com>.
This closes#1940.
The current `test/geo_alignedbox` tests fail on 32-bit arm due to small floating-point errors.
In particular, the following is not guaranteed to hold:
```
IsometryTransform identity = IsometryTransform::Identity();
BoxType transformedC;
transformedC.extend(c.transformed(identity));
VERIFY(transformedC.contains(c));
```
since `c.transformed(identity)` is ever-so-slightly different from `c`. Instead, we replace this test with one that checks an identity transform is within floating-point precision of `c`.
Also updated the condition on `AlignedBox::transform(...)` to only accept `Affine`, `AffineCompact`, and `Isometry` modes explicitly. Otherwise, invalid combinations of modes would also incorrectly pass the assertion.
PR 181 ( https://gitlab.com/libeigen/eigen/-/merge_requests/181 ) adds `__launch_bounds__(1024)` attribute to GPU kernels, that did not have that attribute explicitly specified.
That PR seems to cause regressions on the CUDA platform. This PR/commit makes the changes in PR 181, to be applicable for HIP only
- Introduce CMake option `EIGEN_SPLIT_TESTSUITE` that allows to divide the single test build target into several subtargets
- Add CI pipeline for merge request that can be run by GitLab's shared runners
- Add nightly CI pipeline
Some platforms define int64_t to be long long even for C++03. If this is
the case we miss the definition of internal::make_unsigned for this
type. If we just define the template we get duplicated definitions
errors for platforms defining int64_t as signed long for C++03.
We need to find a way to distinguish both cases at compile-time.
Starting with ROCm 3.5, the HIP compiler will change from HCC to hip-clang.
This compiler change introduce a change in the default value of the `__launch_bounds__` attribute associated with a GPU kernel. (default value means the value assumed by the compiler as the `__launch_bounds attribute__` value, when it is not explicitly specified by the user)
Currently (i.e. for HIP with ROCm 3.3 and older), the default value is 1024. That changes to 256 with ROCm 3.5 (i.e. hip-clang compiler). As a consequence of this change, if a GPU kernel with a `__luanch_bounds__` attribute of 256 is launched at runtime with a threads_per_block value > 256, it leads to a runtime error. This is leading to a couple of Eigen unit test failures with ROCm 3.5.
This commit adds an explicit `__launch_bounds(1024)__` attribute to every GPU kernel that currently does not have it explicitly specified (and hence will end up getting the default value of 256 with the change to hip-clang)
If we have explicit conversion operators available (C++11) we define
explicit casts from bfloat16 to other types. If not (C++03), we don't
define conversion operators but rely on implicit conversion chains from
bfloat16 over float to other types.
Specialized `bfloat16_impl::float_to_bfloat16_rtne(float)` for normal floating point numbers, infinity and zero, in order to improve the performance of `bfloat16::bfloat16(const T&)` for integer argument types.
A reduction of more than 20% of the runtime duration of conversion from int to bfloat16 was observed, using Visual C++ 2019 on Windows 10.
Conversion from `bfloat16` to `float` and `double` is lossless. It seems natural to allow the conversion to be implicit, as the C++ language also support implicit conversion from a smaller to a larger floating point type.
Intel's OneDLL bfloat16 implementation also has an implicit `operator float()`: https://github.com/oneapi-src/oneDNN/blob/v1.5/src/common/bfloat16.hpp
- Guard fundamental types that are not available pre C++11
- Separate subsequent angle brackets >> by spaces
- Allow casting of Eigen::half and Eigen::bfloat16 to complex types
The original tensor casts were only defined for
`SrcCoeffRatio`:`TgtCoeffRatio` 1:1, 1:2, 2:1, 4:1. Here we add the
missing 1:N and 8:1.
We also add casting `Eigen::half` to/from `std::complex<T>`, which
was missing to make it consistent with `Eigen:bfloat16`, and
generalize the overload to work for any complex type.
Tests were added to `basicstuff`, `packetmath`, and
`cxx11_tensor_casts` to test all cast configurations.
Added missing `pmadd<Packet2f>` for NEON. This leads to significant
improvement in precision than previous `pmul+padd`, which was causing
the `pcos` tests to fail. Also added an approx test with
`std::sin`/`std::cos` since otherwise returning any `a^2+b^2=1` would
pass.
Modified `log(denorm)` tests. Denorms are not always supported by all
systems (returns `::min`), are always flushed to zero on 32-bit arm,
and configurably flush to zero on sse/avx/aarch64. This leads to
inconsistent results across different systems (i.e. `-inf` vs `nan`).
Added a check for existence and exclude ARM.
Removed logistic exactness test, since scalar and vectorized versions
follow different code-paths due to differences in `pexp` and `pmadd`,
which result in slightly different values. For example, exactness always
fails on arm, aarch64, and altivec.
The NEON `pcast` operators are all implemented and tested for existing
packets. This requires adding a `pcast(a,b,c,d,e,f,g,h)` for casting
between `int64_t` and `int8_t` in `GenericPacketMath.h`.
Removed incorrect `HasHalfPacket` definition for NEON's
`Packet2l`/`Packet2ul`.
Adjustments were also made to the `packetmath` tests. These include
- minor bug fixes for cast tests (i.e. 4:1 casts, only casting for
packets that are vectorizable)
- added 8:1 cast tests
- random number generation
- original had uninteresting 0 to 0 casts for many casts between
floating-point and integers, and exhibited signed overflow
undefined behavior
Tested:
```
$ aarch64-linux-gnu-g++ -static -I./ '-DEIGEN_TEST_PART_ALL=1' test/packetmath.cpp -o packetmath
$ adb push packetmath /data/local/tmp/
$ adb shell "/data/local/tmp/packetmath"
```
The use of the `packet_traits<>::HasCast` field is currently inconsistent with
`type_casting_traits<>`, and is unused apart from within
`test/packetmath.cpp`. In addition, those packetmath cast tests do not
currently reflect how casts are performed in practice: they ignore the
`SrcCoeffRatio` and `TgtCoeffRatio` fields, assuming a 1:1 ratio.
Here we remove the unsed `HasCast`, and modify the packet cast tests to
better reflect their usage.
This change also contains a few minor cleanups:
1. Remove packet op pnot, which is not needed for anything other than pcmp_le_or_nan,
which can be done in other ways.
2. Remove the "HasInsert" enum, which is no longer needed since we removed the
corresponding packet ops.
3. Add faster pselect op for Packet4i when SSE4.1 is supported.
Among other things, this makes the fast transposeInPlace() method available for Matrix<bool>.
Run on ************** (72 X 2994 MHz CPUs); 2020-05-09T10:51:02.372347913-07:00
CPU: Intel Skylake Xeon with HyperThreading (36 cores) dL1:32KB dL2:1024KB dL3:24MB
Benchmark Time(ns) CPU(ns) Iterations
-----------------------------------------------------------------------
BM_TransposeInPlace<float>/4 9.77 9.77 71670320
BM_TransposeInPlace<float>/8 21.9 21.9 31929525
BM_TransposeInPlace<float>/16 66.6 66.6 10000000
BM_TransposeInPlace<float>/32 243 243 2879561
BM_TransposeInPlace<float>/59 844 844 829767
BM_TransposeInPlace<float>/64 933 933 750567
BM_TransposeInPlace<float>/128 3944 3945 177405
BM_TransposeInPlace<float>/256 16853 16853 41457
BM_TransposeInPlace<float>/512 204952 204968 3448
BM_TransposeInPlace<float>/1k 1053889 1053861 664
BM_TransposeInPlace<bool>/4 14.4 14.4 48637301
BM_TransposeInPlace<bool>/8 36.0 36.0 19370222
BM_TransposeInPlace<bool>/16 31.5 31.5 22178902
BM_TransposeInPlace<bool>/32 111 111 6272048
BM_TransposeInPlace<bool>/59 626 626 1000000
BM_TransposeInPlace<bool>/64 428 428 1632689
BM_TransposeInPlace<bool>/128 1677 1677 417377
BM_TransposeInPlace<bool>/256 7126 7126 96264
BM_TransposeInPlace<bool>/512 29021 29024 24165
BM_TransposeInPlace<bool>/1k 116321 116330 6068
commainitialier unit-test never actually called `test_block_recursion`, which also was not correctly implemented and would have caused too deep template recursion.
For random matrices with integer coefficients, many of the tests here lead to
integer overflows. When taking the norm() of a row/column, the squaredNorm()
often overflows to a negative value, leading to domain errors when taking the
sqrt(). This leads to a crash on some systems. By replacing the norm() call by
a squaredNorm(), the values still overflow, but at least there is no domain
error.
Addresses https://gitlab.com/libeigen/eigen/-/issues/1856
This provides a new op that matches std::rint and previous behavior of
pround. Also adds corresponding unsupported/../Tensor op.
Performance is the same as e. g. floor (tested SSE/AVX).
This fixes deprecated-copy warnings when compiling with GCC>=9
Also protect some additional Base-constructors from getting called by user code code (#1587)
This change re-instates the fast rational approximation of the logistic function for float32 in Eigen (removed in 66f07efeae), but uses the more accurate approximation 1/(1+exp(-1)) ~= exp(x) below -9. The exponential is only calculated on the vectorized path if at least one element in the SIMD input vector is less than -9.
This change also contains a few improvements to speed up the original float specialization of logistic:
- Introduce EIGEN_PREDICT_{FALSE,TRUE} for __builtin_predict and use it to predict that the logistic-only path is most likely (~2-3% speedup for the common case).
- Carefully set the upper clipping point to the smallest x where the approximation evaluates to exactly 1. This saves the explicit clamping of the output (~7% speedup).
The increased accuracy for tanh comes at a cost of 10-20% depending on instruction set.
The benchmarks below repeated calls
u = v.logistic() (u = v.tanh(), respectively)
where u and v are of type Eigen::ArrayXf, have length 8k, and v contains random numbers in [-1,1].
Benchmark numbers for logistic:
Before:
Benchmark Time(ns) CPU(ns) Iterations
-----------------------------------------------------------------
SSE
BM_eigen_logistic_float 4467 4468 155835 model_time: 4827
AVX
BM_eigen_logistic_float 2347 2347 299135 model_time: 2926
AVX+FMA
BM_eigen_logistic_float 1467 1467 476143 model_time: 2926
AVX512
BM_eigen_logistic_float 805 805 858696 model_time: 1463
After:
Benchmark Time(ns) CPU(ns) Iterations
-----------------------------------------------------------------
SSE
BM_eigen_logistic_float 2589 2590 270264 model_time: 4827
AVX
BM_eigen_logistic_float 1428 1428 489265 model_time: 2926
AVX+FMA
BM_eigen_logistic_float 1059 1059 662255 model_time: 2926
AVX512
BM_eigen_logistic_float 673 673 1000000 model_time: 1463
Benchmark numbers for tanh:
Before:
Benchmark Time(ns) CPU(ns) Iterations
-----------------------------------------------------------------
SSE
BM_eigen_tanh_float 2391 2391 292624 model_time: 4242
AVX
BM_eigen_tanh_float 1256 1256 554662 model_time: 2633
AVX+FMA
BM_eigen_tanh_float 823 823 866267 model_time: 1609
AVX512
BM_eigen_tanh_float 443 443 1578999 model_time: 805
After:
Benchmark Time(ns) CPU(ns) Iterations
-----------------------------------------------------------------
SSE
BM_eigen_tanh_float 2588 2588 273531 model_time: 4242
AVX
BM_eigen_tanh_float 1536 1536 452321 model_time: 2633
AVX+FMA
BM_eigen_tanh_float 1007 1007 694681 model_time: 1609
AVX512
BM_eigen_tanh_float 471 471 1472178 model_time: 805
This also adds pset1frombits helper to Packet[24]d.
Makes round ~45% slower for SSE: 1.65µs ± 1% before vs 2.45µs ± 2% after,
stil an order of magnitude faster than scalar version: 33.8µs ± 2%.
Ancient versions of CMake required else(), endif(), and similar block
termination commands to have arguments matching the command starting the block.
This is no longer the preferred style.
2. Simplify handling of special cases by taking advantage of the fact that the
builtin vrsqrt approximation handles negative, zero and +inf arguments correctly.
This speeds up the SSE and AVX implementations by ~20%.
3. Make the Newton-Raphson formula used for rsqrt more numerically robust:
Before: y = y * (1.5 - x/2 * y^2)
After: y = y * (1.5 - y * (x/2) * y)
Forming y^2 can overflow for very large or very small (denormalized) values of x, while x*y ~= 1. For AVX512, this makes it possible to compute accurate results for denormal inputs down to ~1e-42 in single precision.
4. Add a faster double precision implementation for Knights Landing using the vrsqrt28 instruction and a single Newton-Raphson iteration.
Benchmark results: https://bitbucket.org/snippets/rmlarsen/5LBq9o
- Split SpecialFunctions files in to a separate BesselFunctions file.
In particular add:
- Modified bessel functions of the second kind k0, k1, k0e, k1e
- Bessel functions of the first kind j0, j1
- Bessel functions of the second kind y0, y1
Depending on instruction set, significant speedups are observed for the vectorized path:
log1p wall time is reduced 60-93% (2.5x - 15x speedup)
expm1 wall time is reduced 0-85% (1x - 7x speedup)
The scalar path is slower by 20-30% due to the extra branch needed to handle +infinity correctly.
Full benchmarks measured on Intel(R) Xeon(R) Gold 6154 here: https://bitbucket.org/snippets/rmlarsen/MXBkpM
1. Fix buggy pcmp_eq and unit test for half types.
2. Add unit test for pselect and add specializations for SSE 4.1, AVX512, and half types.
3. Get rid of FIXME: Implement faster pnegate for half by XOR'ing with a sign bit mask.
* Allow specifying multiple GPU architectures. E.g.:
cmake -DEIGEN_CUDA_COMPUTE_ARCH="60;70"
* Pass CUDA SDK path to clang. Without it it will default to /usr/local/cuda
which may not be the right location, if cmake was invoked with
-DCUDA_TOOLKIT_ROOT_DIR=/some/other/CUDA/path
