adds unary functions for var<Matrix> #2527

SteveBronder · 2021-07-01T03:38:18Z

Summary

Adds several unary functions for var matrix as well as overloads for division.

Tests

Tests added for each new function

./runTests.py -j4 test/unit/math/mix/fun/divide_test.cpp \
test/unit/math/mix/fun/inv_cloglog_test.cpp \
test/unit/math/mix/fun/inv_sqrt_test.cpp \
test/unit/math/mix/fun/inv_square_test.cpp \
test/unit/math/mix/fun/inv_test.cpp \
test/unit/math/mix/fun/lambertw_test.cpp \
test/unit/math/mix/fun/lgamma_test.cpp \
test/unit/math/mix/fun/log10_test.cpp \
test/unit/math/mix/fun/log1m_exp_test.cpp \
test/unit/math/mix/fun/log1m_inv_logit_test.cpp \
test/unit/math/mix/fun/log2_test.cpp \
test/unit/math/mix/fun/logit_test.cpp

Side Effects

Nopes

Release notes

Adds several unary functions for var as well as division.

Checklist

Math issue How to add static matrix? #1805
Copyright holder: Steve Bronder

The copyright holder is typically you or your assignee, such as a university or company. By submitting this pull request, the copyright holder is agreeing to the license the submitted work under the following licenses:
- Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
- Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
the basic tests are passing
- unit tests pass (to run, use: ./runTests.py test/unit)
- header checks pass, (make test-headers)
- dependencies checks pass, (make test-math-dependencies)
- docs build, (make doxygen)
- code passes the built in C++ standards checks (make cpplint)
the code is written in idiomatic C++ and changes are documented in the doxygen
the new changes are tested

…4.1 (tags/RELEASE_600/final)

andrjohns · 2021-07-02T02:53:45Z

I can review this one, just lmk when it's ready

stan-buildbot · 2021-07-02T07:22:02Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.07	3.13	0.98	-1.88% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.98	-1.54% slower
eight_schools/eight_schools.stan	0.11	0.12	0.98	-2.51% slower
gp_regr/gp_regr.stan	0.16	0.16	1.02	2.01% faster
irt_2pl/irt_2pl.stan	5.87	5.91	0.99	-0.69% slower
performance.compilation	89.27	86.85	1.03	2.72% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.59	8.65	0.99	-0.6% slower
pkpd/one_comp_mm_elim_abs.stan	29.59	29.63	1.0	-0.13% slower
sir/sir.stan	127.19	128.98	0.99	-1.41% slower
gp_regr/gen_gp_data.stan	0.03	0.04	1.0	-0.47% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.0	2.99	1.01	0.56% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.39	0.39	0.99	-0.97% slower
arK/arK.stan	2.56	2.55	1.0	0.14% faster
arma/arma.stan	0.65	0.65	1.0	-0.33% slower
garch/garch.stan	0.64	0.64	1.0	-0.36% slower
Mean result: 0.996549733023

Jenkins Console Log
Blue Ocean
Commit hash: 87a66ec

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

…4.1 (tags/RELEASE_600/final)

hsbadr · 2021-07-03T02:23:21Z

@SteveBronder JFYI; After e80abe9:

stan/math/prim/fun/square.hpp:56:58: error: incomplete type ‘stan::math::apply_scalar_unary<stan::math::square_fun, stan::math::var_value<double>, void>’ used in nested name specifier
   56 |   return apply_scalar_unary<square_fun, Container>::apply(x);
          detected during:
            instantiation of "auto stan::math::square(const Container &) [with Container=stan::math::var, <unnamed>=(void *)nullptr, <unnamed>=(void *)nullptr, <unnamed>=(void *)nullptr]" at line 207 of "stan/math/prim/core/complex_base.hpp"
            instantiation of "stan::math::complex_base<ValueType>::complex_type &stan::math::complex_base<ValueType>::operator/=(const std::complex<U> &) [with ValueType=stan::math::var, U=stan::math::var]" at line 24 of "stan/math/prim/core/operator_division.hpp"
            instantiation of "stan::complex_return_t<U, V> stan::math::internal::complex_divide(const U &, const V &) [with U=std::complex<stan::math::var>, V=std::complex<stan::math::var>]" at line 122 of "stan/math/rev/core/operator_division.hpp"

SteveBronder · 2021-07-03T07:00:11Z

Oh my! Yes much appreciated! Where did you get that error?

hsbadr · 2021-07-03T11:01:53Z

Oh my! Yes much appreciated! Where did you get that error?

Testing with the experimental version of RStan.

This reverts commit 4b434b0.

hsbadr · 2021-07-06T16:48:16Z

@SteveBronder

stan/math/prim/core/complex_base.hpp:207:34: error: no matching function for call to ‘square(stan::math::complex_base<stan::math::var_value<double> >::value_type)’
  207 |     value_type sum_sq_im = square(other.real()) + square(other.imag());

stan/math/rev/core/operator_division.hpp:122:34:   required from here
/usr/include/c++/11.1.0/type_traits:2514:11: error: no type named ‘type’ in ‘struct std::enable_if<false, void>’
 2514 |     using enable_if_t = typename enable_if<_Cond, _Tp>::type;

Log:

In file included from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/std_complex.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:11,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_divide_equal.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core.hpp:29,
                 from /usr/lib/R/library/StanHeaders/include/src/stan/model/model_base.hpp:5,
                 from Module.cpp:2:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp: In instantiation of ‘stan::math::complex_base<ValueType>::complex_type& stan::math::complex_base<ValueType>::operator/=(const std::
complex<_Up>&) [with U = stan::math::var_value<double>; ValueType = stan::math::var_value<double>; stan::math::complex_base<ValueType>::complex_type = std::complex<stan::math::var_value<double> >]’:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/operator_division.hpp:24:5:   required from ‘stan::complex_return_t<U, V> stan::math::internal::complex_divide(const U&, const V&) [with U = std::com
plex<stan::math::var_value<double> >; V = std::complex<stan::math::var_value<double> >; stan::complex_return_t<U, V> = std::complex<stan::math::var_value<double> >]’
/usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:122:34:   required from here
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp:207:34: error: no matching function for call to ‘square(stan::math::complex_base<stan::math::var_value<double> >::value_type)’
  207 |     value_type sum_sq_im = square(other.real()) + square(other.imag());
      |                            ~~~~~~^~~~~~~~~~~~~~
In file included from /usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp:4,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/std_complex.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:11,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_divide_equal.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core.hpp:29,
                 from /usr/lib/R/library/StanHeaders/include/src/stan/model/model_base.hpp:5,
                 from Module.cpp:2:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:68:13: note: candidate: ‘template<class Container, stan::require_container_st<std::is_arithmetic, Container>* <anonymous> > auto stan::math
::square(const Container&)’
   68 | inline auto square(const Container& x) {
      |             ^~~~~~
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:68:13: note:   template argument deduction/substitution failed:
In file included from /usr/include/c++/11.1.0/unordered_map:38,
                 from /usr/lib/R/library/Rcpp/include/Rcpp/platform/compiler.h:153,
                 from /usr/lib/R/library/Rcpp/include/Rcpp/r/headers.h:66,
                 from /usr/lib/R/library/Rcpp/include/RcppCommon.h:30,
                 from /usr/lib/R/library/Rcpp/include/Rcpp.h:27,
                 from Module.cpp:1:
/usr/include/c++/11.1.0/type_traits: In substitution of ‘template<bool _Cond, class _Tp> using enable_if_t = typename std::enable_if::type [with bool _Cond = false; _Tp = void]’:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/meta/require_helpers.hpp:19:7:   required by substitution of ‘template<class Check> using require_t = std::enable_if_t<Check::value> [with Check = std::in
tegral_constant<bool, false>]’
/usr/lib/R/library/StanHeaders/include/stan/math/prim/meta/is_container.hpp:26:1:   required by substitution of ‘template<template<class ...> class TypeCheck, class ... Check> using require_container_st = sta
n::require_t<std::integral_constant<bool, stan::math::conjunction<std::integral_constant<bool, stan::math::disjunction<stan::is_eigen<typename std::decay<Check>::type>, stan::is_std_vector<typename std::decay
<Check>::type, void> >::value>..., TypeCheck<typename stan::scalar_type<Check, void>::type>...>::value> > [with TypeCheck = std::is_arithmetic; Check = {stan::math::var_value<double, void>}]’
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:67:66:   required from ‘stan::math::complex_base<ValueType>::complex_type& stan::math::complex_base<ValueType>::operator/=(const std::compl
ex<_Up>&) [with U = stan::math::var_value<double>; ValueType = stan::math::var_value<double>; stan::math::complex_base<ValueType>::complex_type = std::complex<stan::math::var_value<double> >]’
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/operator_division.hpp:24:5:   required from ‘stan::complex_return_t<U, V> stan::math::internal::complex_divide(const U&, const V&) [with U = std::com
plex<stan::math::var_value<double> >; V = std::complex<stan::math::var_value<double> >; stan::complex_return_t<U, V> = std::complex<stan::math::var_value<double> >]’
/usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:122:34:   required from here
/usr/include/c++/11.1.0/type_traits:2514:11: error: no type named ‘type’ in ‘struct std::enable_if<false, void>’
 2514 |     using enable_if_t = typename enable_if<_Cond, _Tp>::type;
      |           ^~~~~~~~~~~
In file included from /usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp:4,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/std_complex.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:11,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_divide_equal.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core.hpp:29,
                 from /usr/lib/R/library/StanHeaders/include/src/stan/model/model_base.hpp:5,
                 from Module.cpp:2:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp: In instantiation of ‘stan::math::complex_base<ValueType>::complex_type& stan::math::complex_base<ValueType>::operator/=(const std::
complex<_Up>&) [with U = stan::math::var_value<double>; ValueType = stan::math::var_value<double>; stan::math::complex_base<ValueType>::complex_type = std::complex<stan::math::var_value<double> >]’:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/operator_division.hpp:24:5:   required from ‘stan::complex_return_t<U, V> stan::math::internal::complex_divide(const U&, const V&) [with U = std::com
plex<stan::math::var_value<double> >; V = std::complex<stan::math::var_value<double> >; stan::complex_return_t<U, V> = std::complex<stan::math::var_value<double> >]’
/usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:122:34:   required from here
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:54:13: note: candidate: ‘template<class Container, stan::require_not_stan_scalar_t<Container>* <anonymous>, stan::require_not_var_matrix_t<
Container>* <anonymous>, stan::require_not_nonscalar_prim_or_rev_kernel_expression_t<Container>* <anonymous> > auto stan::math::square(const Container&)’
   54 | inline auto square(const Container& x) {
      |             ^~~~~~
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:54:13: note:   template argument deduction/substitution failed:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:26:15: note: candidate: ‘double stan::math::square(double)’
   26 | inline double square(double x) { return std::pow(x, 2); }
      |               ^~~~~~
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:26:29: note:   no known conversion for argument 1 from ‘stan::math::complex_base<stan::math::var_value<double> >::value_type’ {aka ‘stan::m
ath::var_value<double>’} to ‘double’
   26 | inline double square(double x) { return std::pow(x, 2); }
      |                      ~~~~~~~^
In file included from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/std_complex.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:11,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_divide_equal.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core.hpp:29,
                 from /usr/lib/R/library/StanHeaders/include/src/stan/model/model_base.hpp:5,
                 from Module.cpp:2:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp:207:57: error: no matching function for call to ‘square(stan::math::complex_base<stan::math::var_value<double> >::value_type)’
  207 |     value_type sum_sq_im = square(other.real()) + square(other.imag());
      |                                                   ~~~~~~^~~~~~~~~~~~~~
In file included from /usr/lib/R/library/StanHeaders/include/stan/math/prim/core/complex_base.hpp:4,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/std_complex.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_division.hpp:11,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core/operator_divide_equal.hpp:5,
                 from /usr/lib/R/library/StanHeaders/include/stan/math/rev/core.hpp:29,
                 from /usr/lib/R/library/StanHeaders/include/src/stan/model/model_base.hpp:5,
                 from Module.cpp:2:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:68:13: note: candidate: ‘template<class Container, stan::require_container_st<std::is_arithmetic, Container>* <anonymous> > auto stan::math
::square(const Container&)’
   68 | inline auto square(const Container& x) {
      |             ^~~~~~
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:68:13: note:   template argument deduction/substitution failed:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:54:13: note: candidate: ‘template<class Container, stan::require_not_stan_scalar_t<Container>* <anonymous>, stan::require_not_var_matrix_t<
Container>* <anonymous>, stan::require_not_nonscalar_prim_or_rev_kernel_expression_t<Container>* <anonymous> > auto stan::math::square(const Container&)’
   54 | inline auto square(const Container& x) {
      |             ^~~~~~
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:54:13: note:   template argument deduction/substitution failed:
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:26:15: note: candidate: ‘double stan::math::square(double)’
   26 | inline double square(double x) { return std::pow(x, 2); }
      |               ^~~~~~
/usr/lib/R/library/StanHeaders/include/stan/math/prim/fun/square.hpp:26:29: note:   no known conversion for argument 1 from ‘stan::math::complex_base<stan::math::var_value<double> >::value_type’ {aka ‘stan::m
ath::var_value<double>’} to ‘double’
   26 | inline double square(double x) { return std::pow(x, 2); }
      |                      ~~~~~~~^
make: *** [/usr/lib64/R/etc/Makeconf:177: Module.o] Error 1

SteveBronder · 2021-07-06T16:54:54Z

@hsbadr ty! Yes I'm seeing that locally as well. The commit I just made should handle that. The problem was that the complex number stuff was using square() but was importing stuff in a weird way. I just changed the complex number stuff so it used val() * val() instead of directly using square() so should be better

hsbadr · 2021-07-06T17:02:40Z

@hsbadr ty! Yes I'm seeing that locally as well. The commit I just made should handle that. The problem was that the complex number stuff was using square() but was importing stuff in a weird way. I just changed the complex number stuff so it used val() * val() instead of directly using square() so should be better

Yes, it compiles successfully now. Thanks!

wds15 · 2021-07-07T07:18:00Z

This PR is also stalled because of a Jenkins hiccup as it looks.

@hsbadr sorry for off-topic... but is there any hope for rstan to hit CRAN any time soon? I am asking as our next major production system is gearing up for an update and I would like to get rstan in a more recent version. If you have any suggestions let me know... should I ping you on discourse (what was your username there again?).

…4.1 (tags/RELEASE_600/final)

hsbadr · 2021-07-07T13:57:41Z

is there any hope for rstan to hit CRAN any time soon?

Not until @bgoodri gives it sometime or agrees on creating a new package.

I am asking as our next major production system is gearing up for an update and I would like to get rstan in a more recent version. If you have any suggestions let me know

You may add our r-packages to your repos:

install.packages("StanHeaders", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))
install.packages("rstan", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))

or install the experimental version from GitHub.

should I ping you on discourse (what was your username there again?).

hsbadr

…4.1 (tags/RELEASE_600/final)

SteveBronder · 2021-07-07T21:59:12Z

Aight @andrjohns I think this is ready! Couple notes to point out

This adds support for scalar ./ Matrix division
For one of the divide() functions I had to add a is_fvar_or_arithmetic meta type trait
I think the thing that needs most scrutizined are some of the requires I used
Like when I did add() I ended up moving divide() for vars to core/operator_divide
There was some weird goings ons with includes because std::complex<var>'s divide used the square() stan math function, but square() for var was not defined by the time it was used. So in lieu of that I made them into a * a. Not great but also not the worst

Besides that I think it's a pretty normal PR!

stan-buildbot · 2021-07-08T00:22:45Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.15	3.07	1.03	2.56% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	1.0	-0.37% slower
eight_schools/eight_schools.stan	0.11	0.12	0.98	-2.16% slower
gp_regr/gp_regr.stan	0.16	0.16	1.0	0.46% faster
irt_2pl/irt_2pl.stan	5.96	5.88	1.01	1.28% faster
performance.compilation	89.56	86.82	1.03	3.06% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.99	8.58	1.05	4.49% faster
pkpd/one_comp_mm_elim_abs.stan	29.37	30.08	0.98	-2.44% slower
sir/sir.stan	129.74	132.8	0.98	-2.36% slower
gp_regr/gen_gp_data.stan	0.03	0.03	1.0	-0.27% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.0	2.99	1.0	0.47% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.38	0.42	0.91	-9.96% slower
arK/arK.stan	2.53	2.54	1.0	-0.06% slower
arma/arma.stan	0.64	0.65	0.99	-1.32% slower
garch/garch.stan	0.65	0.64	1.0	0.42% faster
Mean result: 0.996836816714

Jenkins Console Log
Blue Ocean
Commit hash: f0d6a83

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

andrjohns

Looking good! My main feedback is that I think most of the additional rev overloads wouldn't be necessary, as long as the existing specifications were modified to use Math functions that are scalar/Matrix agnostic

andrjohns · 2021-07-09T02:45:27Z

stan/math/prim/fun/divide.hpp

+ */
+template <typename Scalar, typename Mat, require_eigen_t<Mat>* = nullptr,
+          require_t<bool_constant<std::is_arithmetic<Scalar>::value
+                                  || is_fvar<Scalar>::value>>* = nullptr,


I wonder if you could replace all of these specialisations with:

template <typename Ta, typename Tb> inline auto divide(const Ta& a, const Tb& b) { return (as_array_or_scalar(a) / as_array_or_scalar(b)).matrix(); }

With the appropriate requires to disambiguate from the varmat overloads?

Oh absolutely! Much cleaner!

stan/math/rev/core/operator_division.hpp

stan/math/rev/fun/divide.hpp

stan/math/rev/fun/exp.hpp

andrjohns · 2021-07-09T03:13:51Z

stan/math/rev/fun/inv.hpp

+template <typename T, require_eigen_t<T>* = nullptr>
+inline auto inv(const var_value<T>& a) {
+  auto denom = to_arena(a.val().array().square());
+  return make_callback_var(inv(a.val()), [a, denom](auto& vi) mutable {
+    a.adj().array() -= vi.adj().array() / denom;
+  });
+}
+


If you use square(), you can combine both overloads:

template <typename T> inline auto inv(const var_value<T>& a) { auto denom = to_arena(square(a.val())); return make_callback_var(inv(a.val()), [a, denom](auto& vi) mutable { a.adj() -= vi.adj() / denom; }); }

andrjohns · 2021-07-09T03:15:25Z

stan/math/rev/fun/inv_cloglog.hpp

@@ -31,6 +31,15 @@ inline var inv_cloglog(const var& a) {
                           });
 }

+template <typename T, require_eigen_t<T>* = nullptr>
+inline auto inv_cloglog(const var_value<T>& a) {
+  auto precomp_exp = (a.val().array() - a.val().array().exp()).exp();


Same here. if you use the Stan Math exp() then you can collapse the overloads

andrjohns · 2021-07-09T03:17:17Z

stan/math/rev/fun/inv_sqrt.hpp

@@ -35,6 +36,14 @@ inline var inv_sqrt(const var& a) {
  });
 }

+template <typename T, require_eigen_t<T>* = nullptr>
+inline auto inv_sqrt(const var_value<T>& a) {
+  auto denom = to_arena(a.val().array() * a.val().array().sqrt());


I'll stop adding comments for this, but same here

SteveBronder · 2021-07-13T02:19:57Z

Ty on the catch for reducing the number of overloads, I had to make a little change to as_array_or_scalar() and was able to remove almost all the extra overloads!

…4.1 (tags/RELEASE_600/final)

SteveBronder · 2021-07-14T20:08:48Z

@andrjohns I think this is ready for review!

stan-buildbot · 2021-07-14T20:50:45Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.05	3.08	0.99	-0.88% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.97	-2.62% slower
eight_schools/eight_schools.stan	0.11	0.11	0.99	-1.3% slower
gp_regr/gp_regr.stan	0.16	0.16	1.01	1.21% faster
irt_2pl/irt_2pl.stan	5.93	5.91	1.0	0.33% faster
performance.compilation	89.03	86.57	1.03	2.76% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.61	8.56	1.01	0.55% faster
pkpd/one_comp_mm_elim_abs.stan	29.3	29.01	1.01	0.96% faster
sir/sir.stan	126.82	121.95	1.04	3.84% faster
gp_regr/gen_gp_data.stan	0.04	0.04	1.0	-0.05% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.07	3.0	1.02	2.23% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.4	0.4	1.0	-0.44% slower
arK/arK.stan	2.53	2.53	1.0	0.15% faster
arma/arma.stan	0.64	0.64	1.0	0.05% faster
garch/garch.stan	0.64	0.64	1.0	-0.43% slower
Mean result: 1.00450323365

Jenkins Console Log
Blue Ocean
Commit hash: 4ebad3b

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

andrjohns

Just some super minor comments, almost done!

andrjohns · 2021-07-16T00:34:28Z

stan/math/prim/core/complex_base.hpp

@@ -204,7 +204,8 @@ class complex_base {
  template <typename U>
  complex_type& operator/=(const std::complex<U>& other) {
    using stan::math::square;
-    value_type sum_sq_im = square(other.real()) + square(other.imag());
+    value_type sum_sq_im
+        = (other.real() * other.real()) + (other.imag() * other.imag());


Might be worth opening a new issue about the square oddities so it gets looked at

Yeah I'm going to open a bigger issue about our include orders.

andrjohns · 2021-07-16T00:36:12Z

stan/math/prim/meta/is_stan_scalar_or_eigen.hpp

+ * Extends std::true_type if all the provided types are either fvar or
+ * an arithmetic type, extends std::false_type otherwise.


Need to update the doc on this

stan/math/rev/core/operator_division.hpp

andrjohns · 2021-07-16T00:39:28Z

stan/math/rev/core/var.hpp

@@ -295,6 +295,13 @@ class var_value<T, require_floating_point_t<T>> {
  }
 };

+namespace internal {
+template <typename T>
+using require_matrix_var_value = require_t<bool_constant<


I'm not a huge fan of defining additional require_s within individual headers, but if you don't see this being used anywhere else then it's fine to leave

So this one is very particular. I'd like to leave it here till unless we need it in other places in the future. This was just in the vari<> template at first and hasn't been used anywhere else. I just pulled it out because it's a bit cleaner looking.

stan-buildbot · 2021-07-17T01:44:45Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.03	3.02	1.0	0.3% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.94	-6.54% slower
eight_schools/eight_schools.stan	0.12	0.11	1.02	1.5% faster
gp_regr/gp_regr.stan	0.16	0.16	0.97	-3.26% slower
irt_2pl/irt_2pl.stan	5.9	5.99	0.98	-1.54% slower
performance.compilation	89.06	86.39	1.03	2.99% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.64	8.56	1.01	0.98% faster
pkpd/one_comp_mm_elim_abs.stan	30.75	29.22	1.05	4.98% faster
sir/sir.stan	131.76	123.94	1.06	5.94% faster
gp_regr/gen_gp_data.stan	0.04	0.03	1.01	1.39% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.07	2.99	1.03	2.54% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.39	0.41	0.94	-6.72% slower
arK/arK.stan	2.53	2.54	1.0	-0.44% slower
arma/arma.stan	0.64	0.65	0.99	-1.22% slower
garch/garch.stan	0.64	0.64	1.0	0.18% faster
Mean result: 1.0019343151

Jenkins Console Log
Blue Ocean
Commit hash: c9c6f0d

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

andrjohns

LGTM!

SteveBronder and others added 3 commits June 30, 2021 23:32

adds unary functions for var<Matrix>

4e263ff

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

711bf32

…4.1 (tags/RELEASE_600/final)

missing append to make/local in jenkins for CPU opencl

87a66ec

SteveBronder and others added 2 commits July 2, 2021 17:55

fixup operator/ to work for var<Matrix> types

e80abe9

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

26b5a78

…4.1 (tags/RELEASE_600/final)

SteveBronder added 2 commits July 6, 2021 11:58

fix headers

4b434b0

Revert "fix headers"

c534e81

This reverts commit 4b434b0.

fix headers

77c0b2b

SteveBronder added 3 commits July 6, 2021 13:30

fix template for square

329f76a

update cos

1b4b135

trying fvar and arithmetic specialization for divide()

5b84c99

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

a50a6c0

…4.1 (tags/RELEASE_600/final)

SteveBronder and others added 2 commits July 7, 2021 11:38

moves is_fvar_or_arithmetic to it's own header

2895732

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

f0d6a83

…4.1 (tags/RELEASE_600/final)

andrjohns requested changes Jul 9, 2021

View reviewed changes

update to reduce number of overloads for var<matrix> functions

c8bfbf1

merge to develop

8de7f58

stan-buildbot and others added 8 commits July 13, 2021 02:21

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

1a8a635

…4.1 (tags/RELEASE_600/final)

update to make additions as double double

631b505

fix array err when adding scalar and eigen type

90373bf

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

0d50121

…4.1 (tags/RELEASE_600/final)

fix log

a1f89e8

update to cover templates for OpenCL types

b58c301

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

35f8930

…4.1 (tags/RELEASE_600/final)

delete is_fvar_or_arithmetic

4ebad3b

andrjohns requested changes Jul 16, 2021

View reviewed changes

update docs for is_stan_scalar_or_eigen

c9c6f0d

andrjohns approved these changes Jul 18, 2021

View reviewed changes

SteveBronder merged commit 8cd7dd0 into develop Jul 18, 2021

rok-cesnovar deleted the feature/varmat-funs-misc1 branch July 19, 2021 06:25

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

adds unary functions for var<Matrix> #2527

adds unary functions for var<Matrix> #2527

SteveBronder commented Jul 1, 2021

andrjohns commented Jul 2, 2021

stan-buildbot commented Jul 2, 2021

hsbadr commented Jul 3, 2021 •

edited

Loading

SteveBronder commented Jul 3, 2021

hsbadr commented Jul 3, 2021

hsbadr commented Jul 6, 2021

SteveBronder commented Jul 6, 2021

hsbadr commented Jul 6, 2021

wds15 commented Jul 7, 2021

hsbadr commented Jul 7, 2021

SteveBronder commented Jul 7, 2021

stan-buildbot commented Jul 8, 2021

andrjohns left a comment

andrjohns Jul 9, 2021

SteveBronder Jul 13, 2021

andrjohns Jul 9, 2021

andrjohns Jul 9, 2021

andrjohns Jul 9, 2021

SteveBronder commented Jul 13, 2021

SteveBronder commented Jul 14, 2021

stan-buildbot commented Jul 14, 2021

andrjohns left a comment •

edited

Loading

andrjohns Jul 16, 2021

SteveBronder Jul 16, 2021

andrjohns Jul 16, 2021

andrjohns Jul 16, 2021

SteveBronder Jul 16, 2021

stan-buildbot commented Jul 17, 2021

andrjohns left a comment

		* Extends std::true_type if all the provided types are either fvar or
		* an arithmetic type, extends std::false_type otherwise.

adds unary functions for var<Matrix> #2527

adds unary functions for var<Matrix> #2527

Conversation

SteveBronder commented Jul 1, 2021

Summary

Tests

Side Effects

Release notes

Checklist

andrjohns commented Jul 2, 2021

stan-buildbot commented Jul 2, 2021

hsbadr commented Jul 3, 2021 • edited Loading

SteveBronder commented Jul 3, 2021

hsbadr commented Jul 3, 2021

hsbadr commented Jul 6, 2021

SteveBronder commented Jul 6, 2021

hsbadr commented Jul 6, 2021

wds15 commented Jul 7, 2021

hsbadr commented Jul 7, 2021

SteveBronder commented Jul 7, 2021

stan-buildbot commented Jul 8, 2021

andrjohns left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SteveBronder commented Jul 13, 2021

SteveBronder commented Jul 14, 2021

stan-buildbot commented Jul 14, 2021

andrjohns left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

stan-buildbot commented Jul 17, 2021

andrjohns left a comment

Choose a reason for hiding this comment

hsbadr commented Jul 3, 2021 •

edited

Loading

andrjohns left a comment •

edited

Loading