llvm-project/polly/lib/Support/ISLTools.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

892 lines
29 KiB
C++
Raw Normal View History

//===------ ISLTools.cpp ----------------------------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// Tools, utilities, helpers and extensions useful in conjunction with the
// Integer Set Library (isl).
//
//===----------------------------------------------------------------------===//
#include "polly/Support/ISLTools.h"
#include "llvm/Support/raw_ostream.h"
#include <cassert>
#include <vector>
using namespace polly;
namespace {
/// Create a map that shifts one dimension by an offset.
///
/// Example:
/// makeShiftDimAff({ [i0, i1] -> [o0, o1] }, 1, -2)
/// = { [i0, i1] -> [i0, i1 - 1] }
///
/// @param Space The map space of the result. Must have equal number of in- and
/// out-dimensions.
/// @param Pos Position to shift.
/// @param Amount Value added to the shifted dimension.
///
/// @return An isl_multi_aff for the map with this shifted dimension.
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::multi_aff makeShiftDimAff(isl::space Space, int Pos, int Amount) {
auto Identity = isl::multi_aff::identity(Space);
if (Amount == 0)
return Identity;
auto ShiftAff = Identity.get_aff(Pos);
ShiftAff = ShiftAff.set_constant_si(Amount);
return Identity.set_aff(Pos, ShiftAff);
}
/// Construct a map that swaps two nested tuples.
///
/// @param FromSpace1 { Space1[] }
/// @param FromSpace2 { Space2[] }
///
/// @return { [Space1[] -> Space2[]] -> [Space2[] -> Space1[]] }
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::basic_map makeTupleSwapBasicMap(isl::space FromSpace1,
isl::space FromSpace2) {
// Fast-path on out-of-quota.
if (!FromSpace1 || !FromSpace2)
return {};
assert(FromSpace1.is_set());
assert(FromSpace2.is_set());
unsigned Dims1 = FromSpace1.dim(isl::dim::set);
unsigned Dims2 = FromSpace2.dim(isl::dim::set);
isl::space FromSpace =
FromSpace1.map_from_domain_and_range(FromSpace2).wrap();
isl::space ToSpace = FromSpace2.map_from_domain_and_range(FromSpace1).wrap();
isl::space MapSpace = FromSpace.map_from_domain_and_range(ToSpace);
isl::basic_map Result = isl::basic_map::universe(MapSpace);
for (unsigned i = 0u; i < Dims1; i += 1)
Result = Result.equate(isl::dim::in, i, isl::dim::out, Dims2 + i);
for (unsigned i = 0u; i < Dims2; i += 1) {
Result = Result.equate(isl::dim::in, Dims1 + i, isl::dim::out, i);
}
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
/// Like makeTupleSwapBasicMap(isl::space,isl::space), but returns
/// an isl_map.
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map makeTupleSwapMap(isl::space FromSpace1, isl::space FromSpace2) {
isl::basic_map BMapResult = makeTupleSwapBasicMap(FromSpace1, FromSpace2);
return isl::map(BMapResult);
}
} // anonymous namespace
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map polly::beforeScatter(isl::map Map, bool Strict) {
isl::space RangeSpace = Map.get_space().range();
isl::map ScatterRel =
Strict ? isl::map::lex_gt(RangeSpace) : isl::map::lex_ge(RangeSpace);
return Map.apply_range(ScatterRel);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map polly::beforeScatter(isl::union_map UMap, bool Strict) {
isl::union_map Result = isl::union_map::empty(UMap.get_space());
for (isl::map Map : UMap.get_map_list()) {
isl::map After = beforeScatter(Map, Strict);
Result = Result.add_map(After);
}
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map polly::afterScatter(isl::map Map, bool Strict) {
isl::space RangeSpace = Map.get_space().range();
isl::map ScatterRel =
Strict ? isl::map::lex_lt(RangeSpace) : isl::map::lex_le(RangeSpace);
return Map.apply_range(ScatterRel);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map polly::afterScatter(const isl::union_map &UMap, bool Strict) {
isl::union_map Result = isl::union_map::empty(UMap.get_space());
for (isl::map Map : UMap.get_map_list()) {
isl::map After = afterScatter(Map, Strict);
Result = Result.add_map(After);
}
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map polly::betweenScatter(isl::map From, isl::map To, bool InclFrom,
bool InclTo) {
isl::map AfterFrom = afterScatter(From, !InclFrom);
isl::map BeforeTo = beforeScatter(To, !InclTo);
return AfterFrom.intersect(BeforeTo);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map polly::betweenScatter(isl::union_map From, isl::union_map To,
bool InclFrom, bool InclTo) {
isl::union_map AfterFrom = afterScatter(From, !InclFrom);
isl::union_map BeforeTo = beforeScatter(To, !InclTo);
return AfterFrom.intersect(BeforeTo);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map polly::singleton(isl::union_map UMap, isl::space ExpectedSpace) {
if (!UMap)
return nullptr;
if (isl_union_map_n_map(UMap.get()) == 0)
return isl::map::empty(ExpectedSpace);
isl::map Result = isl::map::from_union_map(UMap);
assert(!Result || Result.get_space().has_equal_tuples(ExpectedSpace));
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::set polly::singleton(isl::union_set USet, isl::space ExpectedSpace) {
if (!USet)
return nullptr;
if (isl_union_set_n_set(USet.get()) == 0)
return isl::set::empty(ExpectedSpace);
isl::set Result(USet);
assert(!Result || Result.get_space().has_equal_tuples(ExpectedSpace));
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
unsigned polly::getNumScatterDims(const isl::union_map &Schedule) {
unsigned Dims = 0;
for (isl::map Map : Schedule.get_map_list()) {
// Map.dim would return UINT_MAX.
if (!Map)
continue;
Dims = std::max(Dims, Map.dim(isl::dim::out));
}
return Dims;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::space polly::getScatterSpace(const isl::union_map &Schedule) {
if (!Schedule)
return nullptr;
unsigned Dims = getNumScatterDims(Schedule);
isl::space ScatterSpace = Schedule.get_space().set_from_params();
return ScatterSpace.add_dims(isl::dim::set, Dims);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map polly::makeIdentityMap(const isl::union_set &USet,
bool RestrictDomain) {
isl::union_map Result = isl::union_map::empty(USet.get_space());
for (isl::set Set : USet.get_set_list()) {
isl::map IdentityMap = isl::map::identity(Set.get_space().map_from_set());
if (RestrictDomain)
IdentityMap = IdentityMap.intersect_domain(Set);
Result = Result.add_map(IdentityMap);
}
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map polly::reverseDomain(isl::map Map) {
isl::space DomSpace = Map.get_space().domain().unwrap();
isl::space Space1 = DomSpace.domain();
isl::space Space2 = DomSpace.range();
isl::map Swap = makeTupleSwapMap(Space1, Space2);
return Map.apply_domain(Swap);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map polly::reverseDomain(const isl::union_map &UMap) {
isl::union_map Result = isl::union_map::empty(UMap.get_space());
for (isl::map Map : UMap.get_map_list()) {
auto Reversed = reverseDomain(std::move(Map));
Result = Result.add_map(Reversed);
}
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::set polly::shiftDim(isl::set Set, int Pos, int Amount) {
int NumDims = Set.dim(isl::dim::set);
if (Pos < 0)
Pos = NumDims + Pos;
assert(Pos < NumDims && "Dimension index must be in range");
isl::space Space = Set.get_space();
Space = Space.map_from_domain_and_range(Space);
isl::multi_aff Translator = makeShiftDimAff(Space, Pos, Amount);
isl::map TranslatorMap = isl::map::from_multi_aff(Translator);
return Set.apply(TranslatorMap);
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_set polly::shiftDim(isl::union_set USet, int Pos, int Amount) {
isl::union_set Result = isl::union_set::empty(USet.get_space());
for (isl::set Set : USet.get_set_list()) {
isl::set Shifted = shiftDim(Set, Pos, Amount);
Result = Result.add_set(Shifted);
}
return Result;
}
isl::map polly::shiftDim(isl::map Map, isl::dim Dim, int Pos, int Amount) {
int NumDims = Map.dim(Dim);
if (Pos < 0)
Pos = NumDims + Pos;
assert(Pos < NumDims && "Dimension index must be in range");
isl::space Space = Map.get_space();
switch (Dim) {
case isl::dim::in:
Space = Space.domain();
break;
case isl::dim::out:
Space = Space.range();
break;
default:
llvm_unreachable("Unsupported value for 'dim'");
}
Space = Space.map_from_domain_and_range(Space);
isl::multi_aff Translator = makeShiftDimAff(Space, Pos, Amount);
isl::map TranslatorMap = isl::map::from_multi_aff(Translator);
switch (Dim) {
case isl::dim::in:
return Map.apply_domain(TranslatorMap);
case isl::dim::out:
return Map.apply_range(TranslatorMap);
default:
llvm_unreachable("Unsupported value for 'dim'");
}
}
isl::union_map polly::shiftDim(isl::union_map UMap, isl::dim Dim, int Pos,
int Amount) {
isl::union_map Result = isl::union_map::empty(UMap.get_space());
for (isl::map Map : UMap.get_map_list()) {
isl::map Shifted = shiftDim(Map, Dim, Pos, Amount);
Result = Result.add_map(Shifted);
}
return Result;
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
void polly::simplify(isl::set &Set) {
Set = isl::manage(isl_set_compute_divs(Set.copy()));
Set = Set.detect_equalities();
Set = Set.coalesce();
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
void polly::simplify(isl::union_set &USet) {
USet = isl::manage(isl_union_set_compute_divs(USet.copy()));
USet = USet.detect_equalities();
USet = USet.coalesce();
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
void polly::simplify(isl::map &Map) {
Map = isl::manage(isl_map_compute_divs(Map.copy()));
Map = Map.detect_equalities();
Map = Map.coalesce();
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
void polly::simplify(isl::union_map &UMap) {
UMap = isl::manage(isl_union_map_compute_divs(UMap.copy()));
UMap = UMap.detect_equalities();
UMap = UMap.coalesce();
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map polly::computeReachingWrite(isl::union_map Schedule,
isl::union_map Writes, bool Reverse,
bool InclPrevDef, bool InclNextDef) {
// { Scatter[] }
isl::space ScatterSpace = getScatterSpace(Schedule);
// { ScatterRead[] -> ScatterWrite[] }
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::map Relation;
if (Reverse)
Relation = InclPrevDef ? isl::map::lex_lt(ScatterSpace)
: isl::map::lex_le(ScatterSpace);
else
Relation = InclNextDef ? isl::map::lex_gt(ScatterSpace)
: isl::map::lex_ge(ScatterSpace);
// { ScatterWrite[] -> [ScatterRead[] -> ScatterWrite[]] }
isl::map RelationMap = Relation.range_map().reverse();
// { Element[] -> ScatterWrite[] }
isl::union_map WriteAction = Schedule.apply_domain(Writes);
// { ScatterWrite[] -> Element[] }
isl::union_map WriteActionRev = WriteAction.reverse();
// { Element[] -> [ScatterUse[] -> ScatterWrite[]] }
isl::union_map DefSchedRelation =
isl::union_map(RelationMap).apply_domain(WriteActionRev);
// For each element, at every point in time, map to the times of previous
// definitions. { [Element[] -> ScatterRead[]] -> ScatterWrite[] }
isl::union_map ReachableWrites = DefSchedRelation.uncurry();
if (Reverse)
ReachableWrites = ReachableWrites.lexmin();
else
ReachableWrites = ReachableWrites.lexmax();
// { [Element[] -> ScatterWrite[]] -> ScatterWrite[] }
isl::union_map SelfUse = WriteAction.range_map();
if (InclPrevDef && InclNextDef) {
// Add the Def itself to the solution.
ReachableWrites = ReachableWrites.unite(SelfUse).coalesce();
} else if (!InclPrevDef && !InclNextDef) {
// Remove Def itself from the solution.
ReachableWrites = ReachableWrites.subtract(SelfUse);
}
// { [Element[] -> ScatterRead[]] -> Domain[] }
return ReachableWrites.apply_range(Schedule.reverse());
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_map
polly::computeArrayUnused(isl::union_map Schedule, isl::union_map Writes,
isl::union_map Reads, bool ReadEltInSameInst,
bool IncludeLastRead, bool IncludeWrite) {
// { Element[] -> Scatter[] }
isl::union_map ReadActions = Schedule.apply_domain(Reads);
isl::union_map WriteActions = Schedule.apply_domain(Writes);
// { [Element[] -> DomainWrite[]] -> Scatter[] }
isl::union_map EltDomWrites =
Writes.reverse().range_map().apply_range(Schedule);
// { [Element[] -> Scatter[]] -> DomainWrite[] }
isl::union_map ReachingOverwrite = computeReachingWrite(
Schedule, Writes, true, ReadEltInSameInst, !ReadEltInSameInst);
// { [Element[] -> Scatter[]] -> DomainWrite[] }
isl::union_map ReadsOverwritten =
ReachingOverwrite.intersect_domain(ReadActions.wrap());
// { [Element[] -> DomainWrite[]] -> Scatter[] }
isl::union_map ReadsOverwrittenRotated =
reverseDomain(ReadsOverwritten).curry().reverse();
isl::union_map LastOverwrittenRead = ReadsOverwrittenRotated.lexmax();
// { [Element[] -> DomainWrite[]] -> Scatter[] }
isl::union_map BetweenLastReadOverwrite = betweenScatter(
LastOverwrittenRead, EltDomWrites, IncludeLastRead, IncludeWrite);
// { [Element[] -> Scatter[]] -> DomainWrite[] }
isl::union_map ReachingOverwriteZone = computeReachingWrite(
Schedule, Writes, true, IncludeLastRead, IncludeWrite);
// { [Element[] -> DomainWrite[]] -> Scatter[] }
isl::union_map ReachingOverwriteRotated =
reverseDomain(ReachingOverwriteZone).curry().reverse();
// { [Element[] -> DomainWrite[]] -> Scatter[] }
isl::union_map WritesWithoutReads = ReachingOverwriteRotated.subtract_domain(
ReadsOverwrittenRotated.domain());
return BetweenLastReadOverwrite.unite(WritesWithoutReads)
.domain_factor_domain();
}
Introduce isl C++ bindings, Part 1: value_ptr style interface Over the last couple of months several authors of independent isl C++ bindings worked together to jointly design an official set of isl C++ bindings which combines their experience in developing isl C++ bindings. The new bindings have been designed around a value pointer style interface and remove the need for explicit pointer managenent and instead use C++ language features to manage isl objects. This commit introduces the smart-pointer part of the isl C++ bindings and replaces the current IslPtr<T> classes, which served the very same purpose, but had to be manually maintained. Instead, we now rely on automatically generated classes for each isl object, which provide value_ptr semantics. An isl object has the following smart pointer interface: inline set manage(__isl_take isl_set *ptr); class set { friend inline set manage(__isl_take isl_set *ptr); isl_set *ptr = nullptr; inline explicit set(__isl_take isl_set *ptr); public: inline set(); inline set(const set &obj); inline set &operator=(set obj); inline ~set(); inline __isl_give isl_set *copy() const &; inline __isl_give isl_set *copy() && = delete; inline __isl_keep isl_set *get() const; inline __isl_give isl_set *release(); inline bool is_null() const; } The interface and behavior of the new value pointer style classes is inspired by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3339.pdf, which proposes a std::value_ptr, a smart pointer that applies value semantics to its pointee. We currently only provide a limited set of public constructors and instead require provide a global overloaded type constructor method "isl::obj isl::manage(isl_obj *)", which allows to convert an isl_set* to an isl::set by calling 'S = isl::manage(s)'. This pattern models the make_unique() constructor for unique pointers. The next two functions isl::obj::get() and isl::obj::release() are taken directly from the std::value_ptr proposal: S.get() extracts the raw pointer of the object managed by S. S.release() extracts the raw pointer of the object managed by S and sets the object in S to null. We additionally add std::obj::copy(). S.copy() returns a raw pointer refering to a copy of S, which is a shortcut for "isl::obj(oldobj).release()", a functionality commonly needed when interacting directly with the isl C interface where all methods marked with __isl_take require consumable raw pointers. S.is_null() checks if S manages a pointer or if the managed object is currently null. We add this function to provide a more explicit way to check if the pointer is empty compared to a direct conversion to bool. This commit also introduces a couple of polly-specific extensions that cover features currently not handled by the official isl C++ bindings draft, but which have been provided by IslPtr<T> and are consequently added to avoid code churn. These extensions include: - operator bool() : Conversion from objects to bool - construction from nullptr_t - get_ctx() method - take/keep/give methods, which match the currently used naming convention of IslPtr<T> in Polly. They just forward to (release/get/manage). - raw_ostream printers We expect that these extensions are over time either removed or upstreamed to the official isl bindings. We also export a couple of classes that have not yet been exported in isl (e.g., isl::space) As part of the code review, the following two questions were asked: - Why do we not use a standard smart pointer? std::value_ptr was a proposal that has not been accepted. It is consequently not available in the standard library. Even if it would be available, we want to expand this interface with a complete method interface that is conveniently available from each managed pointer. The most direct way to achieve this is to generate a specialiced value style pointer class for each isl object type and add any additional methods to this class. The relevant changes follow in subsequent commits. - Why do we not use templates or macros to avoid code duplication? It is certainly possible to use templates or macros, but as this code is auto-generated there is no need to make writing this code more efficient. Also, most of these classes will be specialized with individual member functions in subsequent commits, such that there will be little code reuse to exploit. Hence, we decided to do so at the moment. These bindings are not yet officially part of isl, but the draft is already very stable. The smart pointer interface itself did not change since serveral months. Adding this code to Polly is against our normal policy of only importing official isl code. In this case however, we make an exception to showcase a non-trivial use case of these bindings which should increase confidence in these bindings and will help upstreaming them to isl. Tags: #polly Reviewed By: Meinersbur Differential Revision: https://reviews.llvm.org/D30325 llvm-svn: 297452
2017-03-10 19:41:03 +08:00
isl::union_set polly::convertZoneToTimepoints(isl::union_set Zone,
bool InclStart, bool InclEnd) {
if (!InclStart && InclEnd)
return Zone;
auto ShiftedZone = shiftDim(Zone, -1, -1);
if (InclStart && !InclEnd)
return ShiftedZone;
else if (!InclStart && !InclEnd)
return Zone.intersect(ShiftedZone);
assert(InclStart && InclEnd);
return Zone.unite(ShiftedZone);
}
isl::union_map polly::convertZoneToTimepoints(isl::union_map Zone, isl::dim Dim,
bool InclStart, bool InclEnd) {
if (!InclStart && InclEnd)
return Zone;
auto ShiftedZone = shiftDim(Zone, Dim, -1, -1);
if (InclStart && !InclEnd)
return ShiftedZone;
else if (!InclStart && !InclEnd)
return Zone.intersect(ShiftedZone);
assert(InclStart && InclEnd);
return Zone.unite(ShiftedZone);
}
isl::map polly::convertZoneToTimepoints(isl::map Zone, isl::dim Dim,
bool InclStart, bool InclEnd) {
if (!InclStart && InclEnd)
return Zone;
auto ShiftedZone = shiftDim(Zone, Dim, -1, -1);
if (InclStart && !InclEnd)
return ShiftedZone;
else if (!InclStart && !InclEnd)
return Zone.intersect(ShiftedZone);
assert(InclStart && InclEnd);
return Zone.unite(ShiftedZone);
}
isl::map polly::distributeDomain(isl::map Map) {
// Note that we cannot take Map apart into { Domain[] -> Range1[] } and {
// Domain[] -> Range2[] } and combine again. We would loose any relation
// between Range1[] and Range2[] that is not also a constraint to Domain[].
isl::space Space = Map.get_space();
isl::space DomainSpace = Space.domain();
if (!DomainSpace)
return {};
unsigned DomainDims = DomainSpace.dim(isl::dim::set);
isl::space RangeSpace = Space.range().unwrap();
isl::space Range1Space = RangeSpace.domain();
if (!Range1Space)
return {};
unsigned Range1Dims = Range1Space.dim(isl::dim::set);
isl::space Range2Space = RangeSpace.range();
if (!Range2Space)
return {};
unsigned Range2Dims = Range2Space.dim(isl::dim::set);
isl::space OutputSpace =
DomainSpace.map_from_domain_and_range(Range1Space)
.wrap()
.map_from_domain_and_range(
DomainSpace.map_from_domain_and_range(Range2Space).wrap());
isl::basic_map Translator = isl::basic_map::universe(
Space.wrap().map_from_domain_and_range(OutputSpace.wrap()));
for (unsigned i = 0; i < DomainDims; i += 1) {
Translator = Translator.equate(isl::dim::in, i, isl::dim::out, i);
Translator = Translator.equate(isl::dim::in, i, isl::dim::out,
DomainDims + Range1Dims + i);
}
for (unsigned i = 0; i < Range1Dims; i += 1)
Translator = Translator.equate(isl::dim::in, DomainDims + i, isl::dim::out,
DomainDims + i);
for (unsigned i = 0; i < Range2Dims; i += 1)
Translator = Translator.equate(isl::dim::in, DomainDims + Range1Dims + i,
isl::dim::out,
DomainDims + Range1Dims + DomainDims + i);
return Map.wrap().apply(Translator).unwrap();
}
isl::union_map polly::distributeDomain(isl::union_map UMap) {
isl::union_map Result = isl::union_map::empty(UMap.get_space());
for (isl::map Map : UMap.get_map_list()) {
auto Distributed = distributeDomain(Map);
Result = Result.add_map(Distributed);
}
return Result;
}
isl::union_map polly::liftDomains(isl::union_map UMap, isl::union_set Factor) {
// { Factor[] -> Factor[] }
isl::union_map Factors = makeIdentityMap(Factor, true);
return Factors.product(UMap);
}
isl::union_map polly::applyDomainRange(isl::union_map UMap,
isl::union_map Func) {
// This implementation creates unnecessary cross products of the
// DomainDomain[] and Func. An alternative implementation could reverse
// domain+uncurry,apply Func to what now is the domain, then undo the
// preparing transformation. Another alternative implementation could create a
// translator map for each piece.
// { DomainDomain[] }
isl::union_set DomainDomain = UMap.domain().unwrap().domain();
// { [DomainDomain[] -> DomainRange[]] -> [DomainDomain[] -> NewDomainRange[]]
// }
isl::union_map LifetedFunc = liftDomains(std::move(Func), DomainDomain);
return UMap.apply_domain(LifetedFunc);
}
isl::map polly::intersectRange(isl::map Map, isl::union_set Range) {
isl::set RangeSet = Range.extract_set(Map.get_space().range());
return Map.intersect_range(RangeSet);
}
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
isl::map polly::subtractParams(isl::map Map, isl::set Params) {
auto MapSpace = Map.get_space();
auto ParamsMap = isl::map::universe(MapSpace).intersect_params(Params);
return Map.subtract(ParamsMap);
}
isl::set polly::subtractParams(isl::set Set, isl::set Params) {
isl::space SetSpace = Set.get_space();
isl::set ParamsSet = isl::set::universe(SetSpace).intersect_params(Params);
return Set.subtract(ParamsSet);
}
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
isl::val polly::getConstant(isl::pw_aff PwAff, bool Max, bool Min) {
assert(!Max || !Min); // Cannot return min and max at the same time.
isl::val Result;
isl::stat Stat = PwAff.foreach_piece(
[=, &Result](isl::set Set, isl::aff Aff) -> isl::stat {
if (Result && Result.is_nan())
return isl::stat::ok();
// TODO: If Min/Max, we can also determine a minimum/maximum value if
// Set is constant-bounded.
if (!Aff.is_cst()) {
Result = isl::val::nan(Aff.get_ctx());
return isl::stat::error();
}
isl::val ThisVal = Aff.get_constant_val();
if (!Result) {
Result = ThisVal;
return isl::stat::ok();
}
if (Result.eq(ThisVal))
return isl::stat::ok();
if (Max && ThisVal.gt(Result)) {
Result = ThisVal;
return isl::stat::ok();
}
if (Min && ThisVal.lt(Result)) {
Result = ThisVal;
return isl::stat::ok();
}
// Not compatible
Result = isl::val::nan(Aff.get_ctx());
return isl::stat::error();
});
if (Stat.is_error())
return {};
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
return Result;
}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
static void foreachPoint(const isl::set &Set,
const std::function<void(isl::point P)> &F) {
Set.foreach_point([&](isl::point P) -> isl::stat {
F(P);
return isl::stat::ok();
});
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
static void foreachPoint(isl::basic_set BSet,
const std::function<void(isl::point P)> &F) {
foreachPoint(isl::set(BSet), F);
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
/// Determine the sorting order of the sets @p A and @p B without considering
/// the space structure.
///
/// Ordering is based on the lower bounds of the set's dimensions. First
/// dimensions are considered first.
static int flatCompare(const isl::basic_set &A, const isl::basic_set &B) {
// Quick bail-out on out-of-quota.
if (!A || !B)
return 0;
unsigned ALen = A.dim(isl::dim::set);
unsigned BLen = B.dim(isl::dim::set);
unsigned Len = std::min(ALen, BLen);
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
for (unsigned i = 0; i < Len; i += 1) {
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
isl::basic_set ADim =
A.project_out(isl::dim::param, 0, A.dim(isl::dim::param))
.project_out(isl::dim::set, i + 1, ALen - i - 1)
.project_out(isl::dim::set, 0, i);
isl::basic_set BDim =
B.project_out(isl::dim::param, 0, B.dim(isl::dim::param))
.project_out(isl::dim::set, i + 1, BLen - i - 1)
.project_out(isl::dim::set, 0, i);
isl::basic_set AHull = isl::set(ADim).convex_hull();
isl::basic_set BHull = isl::set(BDim).convex_hull();
bool ALowerBounded =
bool(isl::set(AHull).dim_has_any_lower_bound(isl::dim::set, 0));
bool BLowerBounded =
bool(isl::set(BHull).dim_has_any_lower_bound(isl::dim::set, 0));
int BoundedCompare = BLowerBounded - ALowerBounded;
if (BoundedCompare != 0)
return BoundedCompare;
if (!ALowerBounded || !BLowerBounded)
continue;
isl::pw_aff AMin = isl::set(ADim).dim_min(0);
isl::pw_aff BMin = isl::set(BDim).dim_min(0);
isl::val AMinVal = polly::getConstant(AMin, false, true);
isl::val BMinVal = polly::getConstant(BMin, false, true);
int MinCompare = AMinVal.sub(BMinVal).sgn();
if (MinCompare != 0)
return MinCompare;
}
// If all the dimensions' lower bounds are equal or incomparable, sort based
// on the number of dimensions.
return ALen - BLen;
}
/// Compare the sets @p A and @p B according to their nested space structure.
/// Returns 0 if the structure is considered equal.
/// If @p ConsiderTupleLen is false, the number of dimensions in a tuple are
/// ignored, i.e. a tuple with the same name but different number of dimensions
/// are considered equal.
static int structureCompare(const isl::space &ASpace, const isl::space &BSpace,
bool ConsiderTupleLen) {
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
int WrappingCompare = bool(ASpace.is_wrapping()) - bool(BSpace.is_wrapping());
if (WrappingCompare != 0)
return WrappingCompare;
if (ASpace.is_wrapping() && BSpace.is_wrapping()) {
isl::space AMap = ASpace.unwrap();
isl::space BMap = BSpace.unwrap();
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
int FirstResult =
structureCompare(AMap.domain(), BMap.domain(), ConsiderTupleLen);
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
if (FirstResult != 0)
return FirstResult;
return structureCompare(AMap.range(), BMap.range(), ConsiderTupleLen);
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
std::string AName;
if (ASpace.has_tuple_name(isl::dim::set))
AName = ASpace.get_tuple_name(isl::dim::set);
std::string BName;
if (BSpace.has_tuple_name(isl::dim::set))
BName = BSpace.get_tuple_name(isl::dim::set);
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
int NameCompare = AName.compare(BName);
if (NameCompare != 0)
return NameCompare;
if (ConsiderTupleLen) {
int LenCompare = BSpace.dim(isl::dim::set) - ASpace.dim(isl::dim::set);
if (LenCompare != 0)
return LenCompare;
}
return 0;
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
/// Compare the sets @p A and @p B according to their nested space structure. If
/// the structure is the same, sort using the dimension lower bounds.
/// Returns an std::sort compatible bool.
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
static bool orderComparer(const isl::basic_set &A, const isl::basic_set &B) {
isl::space ASpace = A.get_space();
isl::space BSpace = B.get_space();
// Ignoring number of dimensions first ensures that structures with same tuple
// names, but different number of dimensions are still sorted close together.
int TupleNestingCompare = structureCompare(ASpace, BSpace, false);
if (TupleNestingCompare != 0)
return TupleNestingCompare < 0;
int TupleCompare = structureCompare(ASpace, BSpace, true);
if (TupleCompare != 0)
return TupleCompare < 0;
return flatCompare(A, B) < 0;
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
/// Print a string representation of @p USet to @p OS.
///
/// The pieces of @p USet are printed in a sorted order. Spaces with equal or
/// similar nesting structure are printed together. Compared to isl's own
/// printing function the uses the structure itself as base of the sorting, not
/// a hash of it. It ensures that e.g. maps spaces with same domain structure
/// are printed together. Set pieces with same structure are printed in order of
/// their lower bounds.
///
/// @param USet Polyhedra to print.
/// @param OS Target stream.
/// @param Simplify Whether to simplify the polyhedron before printing.
/// @param IsMap Whether @p USet is a wrapped map. If true, sets are
/// unwrapped before printing to again appear as a map.
static void printSortedPolyhedra(isl::union_set USet, llvm::raw_ostream &OS,
bool Simplify, bool IsMap) {
if (!USet) {
OS << "<null>\n";
return;
}
if (Simplify)
simplify(USet);
// Get all the polyhedra.
std::vector<isl::basic_set> BSets;
for (isl::set Set : USet.get_set_list()) {
for (isl::basic_set BSet : Set.get_basic_set_list()) {
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
BSets.push_back(BSet);
}
}
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
if (BSets.empty()) {
OS << "{\n}\n";
return;
}
// Sort the polyhedra.
llvm::sort(BSets, orderComparer);
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
// Print the polyhedra.
bool First = true;
for (const isl::basic_set &BSet : BSets) {
std::string Str;
if (IsMap)
Str = isl::map(BSet.unwrap()).to_str();
else
Str = isl::set(BSet).to_str();
size_t OpenPos = Str.find_first_of('{');
assert(OpenPos != std::string::npos);
size_t ClosePos = Str.find_last_of('}');
assert(ClosePos != std::string::npos);
if (First)
OS << llvm::StringRef(Str).substr(0, OpenPos + 1) << "\n ";
else
OS << ";\n ";
OS << llvm::StringRef(Str).substr(OpenPos + 1, ClosePos - OpenPos - 2);
First = false;
}
assert(!First);
OS << "\n}\n";
}
static void recursiveExpand(isl::basic_set BSet, int Dim, isl::set &Expanded) {
int Dims = BSet.dim(isl::dim::set);
if (Dim >= Dims) {
Expanded = Expanded.unite(BSet);
return;
}
isl::basic_set DimOnly =
BSet.project_out(isl::dim::param, 0, BSet.dim(isl::dim::param))
.project_out(isl::dim::set, Dim + 1, Dims - Dim - 1)
.project_out(isl::dim::set, 0, Dim);
if (!DimOnly.is_bounded()) {
recursiveExpand(BSet, Dim + 1, Expanded);
return;
}
foreachPoint(DimOnly, [&, Dim](isl::point P) {
isl::val Val = P.get_coordinate_val(isl::dim::set, 0);
isl::basic_set FixBSet = BSet.fix_val(isl::dim::set, Dim, Val);
recursiveExpand(FixBSet, Dim + 1, Expanded);
});
}
/// Make each point of a set explicit.
///
/// "Expanding" makes each point a set contains explicit. That is, the result is
/// a set of singleton polyhedra. Unbounded dimensions are not expanded.
///
/// Example:
/// { [i] : 0 <= i < 2 }
/// is expanded to:
/// { [0]; [1] }
static isl::set expand(const isl::set &Set) {
isl::set Expanded = isl::set::empty(Set.get_space());
for (isl::basic_set BSet : Set.get_basic_set_list())
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
recursiveExpand(BSet, 0, Expanded);
return Expanded;
}
/// Expand all points of a union set explicit.
///
/// @see expand(const isl::set)
static isl::union_set expand(const isl::union_set &USet) {
isl::union_set Expanded = isl::union_set::empty(USet.get_space());
for (isl::set Set : USet.get_set_list()) {
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
isl::set SetExpanded = expand(Set);
Expanded = Expanded.add_set(SetExpanded);
}
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
return Expanded;
}
LLVM_DUMP_METHOD void polly::dumpPw(const isl::set &Set) {
printSortedPolyhedra(Set, llvm::errs(), true, false);
}
LLVM_DUMP_METHOD void polly::dumpPw(const isl::map &Map) {
printSortedPolyhedra(Map.wrap(), llvm::errs(), true, true);
}
LLVM_DUMP_METHOD void polly::dumpPw(const isl::union_set &USet) {
printSortedPolyhedra(USet, llvm::errs(), true, false);
}
LLVM_DUMP_METHOD void polly::dumpPw(const isl::union_map &UMap) {
printSortedPolyhedra(UMap.wrap(), llvm::errs(), true, true);
}
LLVM_DUMP_METHOD void polly::dumpPw(__isl_keep isl_set *Set) {
dumpPw(isl::manage_copy(Set));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpPw(__isl_keep isl_map *Map) {
dumpPw(isl::manage_copy(Map));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpPw(__isl_keep isl_union_set *USet) {
dumpPw(isl::manage_copy(USet));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpPw(__isl_keep isl_union_map *UMap) {
dumpPw(isl::manage_copy(UMap));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpExpanded(const isl::set &Set) {
printSortedPolyhedra(expand(Set), llvm::errs(), false, false);
}
LLVM_DUMP_METHOD void polly::dumpExpanded(const isl::map &Map) {
printSortedPolyhedra(expand(Map.wrap()), llvm::errs(), false, true);
}
LLVM_DUMP_METHOD void polly::dumpExpanded(const isl::union_set &USet) {
printSortedPolyhedra(expand(USet), llvm::errs(), false, false);
}
LLVM_DUMP_METHOD void polly::dumpExpanded(const isl::union_map &UMap) {
printSortedPolyhedra(expand(UMap.wrap()), llvm::errs(), false, true);
}
LLVM_DUMP_METHOD void polly::dumpExpanded(__isl_keep isl_set *Set) {
dumpExpanded(isl::manage_copy(Set));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpExpanded(__isl_keep isl_map *Map) {
dumpExpanded(isl::manage_copy(Map));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpExpanded(__isl_keep isl_union_set *USet) {
dumpExpanded(isl::manage_copy(USet));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
LLVM_DUMP_METHOD void polly::dumpExpanded(__isl_keep isl_union_map *UMap) {
dumpExpanded(isl::manage_copy(UMap));
[Polly] Add dumpPw() and dumpExpanded() functions. NFC. These functions print a multi-line and sorted representation of unions of polyhedra. Each polyhedron (basic_{ast/map}) has its own line. First sort key is the polyhedron's hierachical space structure. Secondary sort key is the lower bound of the polyhedron, which should ensure that the polyhedral are printed in approximately ascending order. Example output of dumpPw(): [p_0, p_1, p_2] -> { Stmt0[0] -> [0, 0]; Stmt0[i0] -> [i0, 0] : 0 < i0 <= 5 - p_2; Stmt1[0] -> [0, 2] : p_1 = 1 and p_0 = -1; Stmt2[0] -> [0, 1] : p_1 >= 3 + p_0; Stmt3[0] -> [0, 3]; } In contrast dumpExpanded() prints each point in the sets, unless there is an unbounded dimension that cannot be expandend. This is useful for reduced test cases where the loop counts are set to some constant to understand a bug. Example output of dumpExpanded( { [MemRef_A[i0] -> [i1]] : (exists (e0 = floor((1 + i1)/3): i0 = 1 and 3e0 <= i1 and 3e0 >= -1 + i1 and i1 >= 15 and i1 <= 25)) or (exists (e0 = floor((i1)/3): i0 = 0 and 3e0 < i1 and 3e0 >= -2 + i1 and i1 > 0 and i1 <= 11)) }): { [MemRef_A[0] ->[1]]; [MemRef_A[0] ->[2]]; [MemRef_A[0] ->[4]]; [MemRef_A[0] ->[5]]; [MemRef_A[0] ->[7]]; [MemRef_A[0] ->[8]]; [MemRef_A[0] ->[10]]; [MemRef_A[0] ->[11]]; [MemRef_A[1] ->[15]]; [MemRef_A[1] ->[16]]; [MemRef_A[1] ->[18]]; [MemRef_A[1] ->[19]]; [MemRef_A[1] ->[21]]; [MemRef_A[1] ->[22]]; [MemRef_A[1] ->[24]]; [MemRef_A[1] ->[25]] } Differential Revision: https://reviews.llvm.org/D38349 llvm-svn: 314525
2017-09-29 23:45:40 +08:00
}
#endif