archive/blender-archive
Archived
1
1
Fork 0
Code Pull Requests Packages Activity
This repository has been archived on 2023-10-09. You can view files and clone it, but cannot push or open issues or pull requests.
Files
ae94e36cfb2f3bc9a99b638782092d9c71d4b3c7
blender-archive/source/blender/functions/FN_multi_function_params.hh

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

347 lines
12 KiB
C++
Raw Normal View History

File headers: SPDX License migration Use a shorter/simpler license convention, stops the header taking so much space. Follow the SPDX license specification: https://spdx.org/licenses - C/C++/objc/objc++ - Python - Shell Scripts - CMake, GNUmakefile While most of the source tree has been included - `./extern/` was left out. - `./intern/cycles` & `./intern/atomic` are also excluded because they use different header conventions. doc/license/SPDX-license-identifiers.txt has been added to list SPDX all used identifiers. See P2788 for the script that automated these edits. Reviewed By: brecht, mont29, sergey Ref D14069
2022-02-11 09:07:11 +11:00
/* SPDX-License-Identifier: GPL-2.0-or-later */
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
#pragma once
/** \file
* \ingroup fn
*
* This file provides an MFParams and MFParamsBuilder structure.
*
* `MFParamsBuilder` is used by a function caller to be prepare all parameters that are passed into
* the function. `MFParams` is then used inside the called function to access the parameters.
*/
Geometry Nodes: refactor multi-threading in field evaluation Previously, there was a fixed grain size for all multi-functions. That was not sufficient because some functions could benefit a lot from smaller grain sizes. This refactors adds a new `MultiFunction::call_auto` method which has the same effect as just calling `MultiFunction::call` but additionally figures out how to execute the specific multi-function efficiently. It determines a good grain size and decides whether the mask indices should be shifted or not. Most multi-function evaluations benefit from this, but medium sized work loads (1000 - 50000 elements) benefit from it the most. Especially when expensive multi-functions (e.g. noise) is involved. This is because for smaller work loads, threading is rarely used and for larger work loads threading worked fine before already. With this patch, multi-functions can specify execution hints, that allow the caller to execute it most efficiently. These execution hints still have to be added to more functions. Some performance measurements of a field evaluation involving noise and math nodes, ordered by the number of elements being evaluated: ``` 1,000,000: 133 ms -> 120 ms 100,000: 30 ms -> 18 ms 10,000: 20 ms -> 2.7 ms 1,000: 4 ms -> 0.5 ms 100: 0.5 ms -> 0.4 ms ```
2021-11-26 11:05:47 +01:00
#include <mutex>
BLI: move generic data structures to blenlib This is a follow up to rB2252bc6a5527cd7360d1ccfe7a2d1bc640a8dfa6.
2022-03-19 08:26:29 +01:00
#include "BLI_generic_pointer.hh"
#include "BLI_generic_vector_array.hh"
#include "BLI_generic_virtual_vector_array.hh"
BLI: rename resource collector to resource scope Differential Revision: https://developer.blender.org/D10857
2021-04-01 15:55:08 +02:00
#include "BLI_resource_scope.hh"
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
#include "FN_multi_function_signature.hh"
Cleanup: use nested namespaces
2020-07-03 14:25:20 +02:00
namespace blender::fn {
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
class MFParamsBuilder {
private:
BLI: rename resource collector to resource scope Differential Revision: https://developer.blender.org/D10857
2021-04-01 15:55:08 +02:00
ResourceScope scope_;
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
const MFSignature *signature_;
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
IndexMask mask_;
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int64_t min_array_size_;
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
Vector<GVArray> virtual_arrays_;
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
Vector<GMutableSpan> mutable_spans_;
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
Vector<const GVVectorArray *> virtual_vector_arrays_;
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
Vector<GVectorArray *> vector_arrays_;
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
Geometry Nodes: refactor multi-threading in field evaluation Previously, there was a fixed grain size for all multi-functions. That was not sufficient because some functions could benefit a lot from smaller grain sizes. This refactors adds a new `MultiFunction::call_auto` method which has the same effect as just calling `MultiFunction::call` but additionally figures out how to execute the specific multi-function efficiently. It determines a good grain size and decides whether the mask indices should be shifted or not. Most multi-function evaluations benefit from this, but medium sized work loads (1000 - 50000 elements) benefit from it the most. Especially when expensive multi-functions (e.g. noise) is involved. This is because for smaller work loads, threading is rarely used and for larger work loads threading worked fine before already. With this patch, multi-functions can specify execution hints, that allow the caller to execute it most efficiently. These execution hints still have to be added to more functions. Some performance measurements of a field evaluation involving noise and math nodes, ordered by the number of elements being evaluated: ``` 1,000,000: 133 ms -> 120 ms 100,000: 30 ms -> 18 ms 10,000: 20 ms -> 2.7 ms 1,000: 4 ms -> 0.5 ms 100: 0.5 ms -> 0.4 ms ```
2021-11-26 11:05:47 +01:00
std::mutex mutex_;
Vector<std::pair<int, GMutableSpan>> dummy_output_spans_;
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
friend class MFParams;
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
MFParamsBuilder(const MFSignature &signature, const IndexMask mask)
: signature_(&signature), mask_(mask), min_array_size_(mask.min_array_size())
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
}
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
public:
MFParamsBuilder(const class MultiFunction &fn, int64_t size);
/**
* The indices referenced by the #mask has to live longer than the params builder. This is
* because the it might have to destruct elements for all masked indices in the end.
*/
MFParamsBuilder(const class MultiFunction &fn, const IndexMask *mask);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
Functions: add utility methods to parameter builder
2021-08-20 11:43:54 +02:00
template<typename T> void add_readonly_single_input_value(T value, StringRef expected_name = "")
{
Geometry Nodes: refactor multi-threading in field evaluation Previously, there was a fixed grain size for all multi-functions. That was not sufficient because some functions could benefit a lot from smaller grain sizes. This refactors adds a new `MultiFunction::call_auto` method which has the same effect as just calling `MultiFunction::call` but additionally figures out how to execute the specific multi-function efficiently. It determines a good grain size and decides whether the mask indices should be shifted or not. Most multi-function evaluations benefit from this, but medium sized work loads (1000 - 50000 elements) benefit from it the most. Especially when expensive multi-functions (e.g. noise) is involved. This is because for smaller work loads, threading is rarely used and for larger work loads threading worked fine before already. With this patch, multi-functions can specify execution hints, that allow the caller to execute it most efficiently. These execution hints still have to be added to more functions. Some performance measurements of a field evaluation involving noise and math nodes, ordered by the number of elements being evaluated: ``` 1,000,000: 133 ms -> 120 ms 100,000: 30 ms -> 18 ms 10,000: 20 ms -> 2.7 ms 1,000: 4 ms -> 0.5 ms 100: 0.5 ms -> 0.4 ms ```
2021-11-26 11:05:47 +01:00
this->add_readonly_single_input(VArray<T>::ForSingle(std::move(value), min_array_size_),
expected_name);
Functions: add utility methods to parameter builder
2021-08-20 11:43:54 +02:00
}
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
template<typename T> void add_readonly_single_input(const T *value, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Cleanup: simplify resource scope methods Previously, a debug name had to be passed to all methods that added a resource to the `ResourceScope`. The idea was that this would make it easier to find certain bugs. In reality I never found this to be useful, and it was mostly annoying. The thing is, something that is in a resource scope never leaks (unless the resource scope is not destructed of course). Removing the name parameter makes the structure easier to use.
2021-09-14 16:08:09 +02:00
this->add_readonly_single_input(
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray::ForSingleRef(CPPType::get<T>(), min_array_size_, value), expected_name);
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
}
void add_readonly_single_input(const GSpan span, StringRef expected_name = "")
{
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
this->add_readonly_single_input(GVArray::ForSpan(span), expected_name);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: simplify adding a single input to a multi-function This is used by the upcoming new geometry nodes evaluator.
2021-05-13 13:23:53 +02:00
void add_readonly_single_input(GPointer value, StringRef expected_name = "")
{
Cleanup: simplify resource scope methods Previously, a debug name had to be passed to all methods that added a resource to the `ResourceScope`. The idea was that this would make it easier to find certain bugs. In reality I never found this to be useful, and it was mostly annoying. The thing is, something that is in a resource scope never leaks (unless the resource scope is not destructed of course). Removing the name parameter makes the structure easier to use.
2021-09-14 16:08:09 +02:00
this->add_readonly_single_input(
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray::ForSingleRef(*value.type(), min_array_size_, value.get()), expected_name);
Functions: simplify adding a single input to a multi-function This is used by the upcoming new geometry nodes evaluator.
2021-05-13 13:23:53 +02:00
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
void add_readonly_single_input(GVArray varray, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
this->assert_current_param_type(MFParamType::ForSingleInput(varray.type()), expected_name);
BLI_assert(varray.size() >= min_array_size_);
virtual_arrays_.append(varray);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
void add_readonly_vector_input(const GVectorArray &vector_array, StringRef expected_name = "")
{
Cleanup: simplify resource scope methods Previously, a debug name had to be passed to all methods that added a resource to the `ResourceScope`. The idea was that this would make it easier to find certain bugs. In reality I never found this to be useful, and it was mostly annoying. The thing is, something that is in a resource scope never leaks (unless the resource scope is not destructed of course). Removing the name parameter makes the structure easier to use.
2021-09-14 16:08:09 +02:00
this->add_readonly_vector_input(scope_.construct<GVVectorArray_For_GVectorArray>(vector_array),
expected_name);
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
}
Functions: add utility methods to parameter builder
2021-08-20 11:43:54 +02:00
void add_readonly_vector_input(const GSpan single_vector, StringRef expected_name = "")
{
this->add_readonly_vector_input(
Cleanup: simplify resource scope methods Previously, a debug name had to be passed to all methods that added a resource to the `ResourceScope`. The idea was that this would make it easier to find certain bugs. In reality I never found this to be useful, and it was mostly annoying. The thing is, something that is in a resource scope never leaks (unless the resource scope is not destructed of course). Removing the name parameter makes the structure easier to use.
2021-09-14 16:08:09 +02:00
scope_.construct<GVVectorArray_For_SingleGSpan>(single_vector, min_array_size_),
Functions: add utility methods to parameter builder
2021-08-20 11:43:54 +02:00
expected_name);
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
void add_readonly_vector_input(const GVVectorArray &ref, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
this->assert_current_param_type(MFParamType::ForVectorInput(ref.type()), expected_name);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(ref.size() >= min_array_size_);
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
virtual_vector_arrays_.append(&ref);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
template<typename T> void add_uninitialized_single_output(T *value, StringRef expected_name = "")
Particles: initial object socket and emitter node support Object sockets work now, but only the new Object Transforms and the Particle Mesh Emitter node use it. The emitter does not actually use the mesh surface yet. Instead, new particles are just emitted around the origin of the object. Internally, handles to object data blocks are passed around in the network, instead of raw object pointers. Using handles has a couple of benefits: * The caller of the function has control over which handles can be resolved and therefore limit access to specific data. The set of data blocks that is accessed by a node tree should be known statically. This is necessary for a proper integration with the dependency graph. * When the pointer to an object changes (e.g. after restarting Blender), all handles are still valid. * When an object is deleted, the handle is invalidated without causing crashes. * The handle is just an integer that can be stored per particle and can be cached easily. The mapping between handles and their corresponding data blocks is stored in the Simulation data block.
2020-07-21 17:20:05 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
this->add_uninitialized_single_output(GMutableSpan(CPPType::get<T>(), value, 1),
expected_name);
Particles: initial object socket and emitter node support Object sockets work now, but only the new Object Transforms and the Particle Mesh Emitter node use it. The emitter does not actually use the mesh surface yet. Instead, new particles are just emitted around the origin of the object. Internally, handles to object data blocks are passed around in the network, instead of raw object pointers. Using handles has a couple of benefits: * The caller of the function has control over which handles can be resolved and therefore limit access to specific data. The set of data blocks that is accessed by a node tree should be known statically. This is necessary for a proper integration with the dependency graph. * When the pointer to an object changes (e.g. after restarting Blender), all handles are still valid. * When an object is deleted, the handle is invalidated without causing crashes. * The handle is just an integer that can be stored per particle and can be cached easily. The mapping between handles and their corresponding data blocks is stored in the Simulation data block.
2020-07-21 17:20:05 +02:00
}
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
void add_uninitialized_single_output(GMutableSpan ref, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
this->assert_current_param_type(MFParamType::ForSingleOutput(ref.type()), expected_name);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(ref.size() >= min_array_size_);
mutable_spans_.append(ref);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
void add_ignored_single_output(StringRef expected_name = "")
{
this->assert_current_param_name(expected_name);
const int param_index = this->current_param_index();
const MFParamType &param_type = signature_->param_types[param_index];
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
BLI_assert(param_type.category() == MFParamCategory::SingleOutput);
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
const CPPType &type = param_type.data_type().single_type();
/* An empty span indicates that this is ignored. */
const GMutableSpan dummy_span{type};
mutable_spans_.append(dummy_span);
}
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
void add_vector_output(GVectorArray &vector_array, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
this->assert_current_param_type(MFParamType::ForVectorOutput(vector_array.type()),
expected_name);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(vector_array.size() >= min_array_size_);
vector_arrays_.append(&vector_array);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
void add_single_mutable(GMutableSpan ref, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
this->assert_current_param_type(MFParamType::ForMutableSingle(ref.type()), expected_name);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(ref.size() >= min_array_size_);
mutable_spans_.append(ref);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
void add_vector_mutable(GVectorArray &vector_array, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
this->assert_current_param_type(MFParamType::ForMutableVector(vector_array.type()),
expected_name);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(vector_array.size() >= min_array_size_);
vector_arrays_.append(&vector_array);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
GMutableSpan computed_array(int param_index)
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(ELEM(signature_->param_types[param_index].category(),
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
MFParamCategory::SingleOutput,
MFParamCategory::SingleMutable));
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = signature_->data_index(param_index);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
return mutable_spans_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
GVectorArray &computed_vector_array(int param_index)
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(ELEM(signature_->param_types[param_index].category(),
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
MFParamCategory::VectorOutput,
MFParamCategory::VectorMutable));
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = signature_->data_index(param_index);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
return *vector_arrays_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
BLI: rename resource collector to resource scope Differential Revision: https://developer.blender.org/D10857
2021-04-01 15:55:08 +02:00
ResourceScope &resource_scope()
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
{
BLI: rename resource collector to resource scope Differential Revision: https://developer.blender.org/D10857
2021-04-01 15:55:08 +02:00
return scope_;
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
}
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
private:
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
void assert_current_param_type(MFParamType param_type, StringRef expected_name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
UNUSED_VARS_NDEBUG(param_type, expected_name);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
#ifdef DEBUG
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int param_index = this->current_param_index();
Particles: improve mesh emitter Particles are now emitted from vertices of the mesh.
2020-07-23 17:57:11 +02:00
if (expected_name != "") {
StringRef actual_name = signature_->param_names[param_index];
BLI_assert(actual_name == expected_name);
}
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
MFParamType expected_type = signature_->param_types[param_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
BLI_assert(expected_type == param_type);
#endif
}
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
void assert_current_param_name(StringRef expected_name)
{
UNUSED_VARS_NDEBUG(expected_name);
#ifdef DEBUG
if (expected_name.is_empty()) {
return;
}
const int param_index = this->current_param_index();
StringRef actual_name = signature_->param_names[param_index];
BLI_assert(actual_name == expected_name);
#endif
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int current_param_index() const
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
return virtual_arrays_.size() + mutable_spans_.size() + virtual_vector_arrays_.size() +
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
vector_arrays_.size();
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
};
class MFParams {
private:
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
MFParamsBuilder *builder_;
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
public:
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
MFParams(MFParamsBuilder &builder) : builder_(&builder)
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
template<typename T> VArray<T> readonly_single_input(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
const GVArray &varray = this->readonly_single_input(param_index, name);
return varray.typed<T>();
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
const GVArray &readonly_single_input(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::SingleInput);
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = builder_->signature_->data_index(param_index);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
return builder_->virtual_arrays_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
/**
* \return True when the caller provided a buffer for this output parameter. This allows the
* called multi-function to skip some computation. It is still valid to call
* #uninitialized_single_output when this returns false. In this case a new temporary buffer is
* allocated.
*/
bool single_output_is_required(int param_index, StringRef name = "")
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::SingleOutput);
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
int data_index = builder_->signature_->data_index(param_index);
return !builder_->mutable_spans_[data_index].is_empty();
}
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
template<typename T>
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
MutableSpan<T> uninitialized_single_output(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
return this->uninitialized_single_output(param_index, name).typed<T>();
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
GMutableSpan uninitialized_single_output(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::SingleOutput);
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = builder_->signature_->data_index(param_index);
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
GMutableSpan span = builder_->mutable_spans_[data_index];
Geometry Nodes: refactor multi-threading in field evaluation Previously, there was a fixed grain size for all multi-functions. That was not sufficient because some functions could benefit a lot from smaller grain sizes. This refactors adds a new `MultiFunction::call_auto` method which has the same effect as just calling `MultiFunction::call` but additionally figures out how to execute the specific multi-function efficiently. It determines a good grain size and decides whether the mask indices should be shifted or not. Most multi-function evaluations benefit from this, but medium sized work loads (1000 - 50000 elements) benefit from it the most. Especially when expensive multi-functions (e.g. noise) is involved. This is because for smaller work loads, threading is rarely used and for larger work loads threading worked fine before already. With this patch, multi-functions can specify execution hints, that allow the caller to execute it most efficiently. These execution hints still have to be added to more functions. Some performance measurements of a field evaluation involving noise and math nodes, ordered by the number of elements being evaluated: ``` 1,000,000: 133 ms -> 120 ms 100,000: 30 ms -> 18 ms 10,000: 20 ms -> 2.7 ms 1,000: 4 ms -> 0.5 ms 100: 0.5 ms -> 0.4 ms ```
2021-11-26 11:05:47 +01:00
if (!span.is_empty()) {
return span;
Functions: support optional outputs in multi-function Sometimes not all outputs of a multi-function are required by the caller. In those cases it would be a waste of compute resources to calculate the unused values anyway. Now, the caller of a multi-function can specify when a specific output is not used. The called function can check if an output is unused and may ignore it. Multi-functions can still computed unused outputs as before if they don't want to check if a specific output is unused. The multi-function procedure system has been updated to support ignored outputs in call instructions. An ignored output just has no variable assigned to it. The field system has been updated to generate a multi-function procedure where unused outputs are ignored.
2021-09-14 14:52:44 +02:00
}
Geometry Nodes: refactor multi-threading in field evaluation Previously, there was a fixed grain size for all multi-functions. That was not sufficient because some functions could benefit a lot from smaller grain sizes. This refactors adds a new `MultiFunction::call_auto` method which has the same effect as just calling `MultiFunction::call` but additionally figures out how to execute the specific multi-function efficiently. It determines a good grain size and decides whether the mask indices should be shifted or not. Most multi-function evaluations benefit from this, but medium sized work loads (1000 - 50000 elements) benefit from it the most. Especially when expensive multi-functions (e.g. noise) is involved. This is because for smaller work loads, threading is rarely used and for larger work loads threading worked fine before already. With this patch, multi-functions can specify execution hints, that allow the caller to execute it most efficiently. These execution hints still have to be added to more functions. Some performance measurements of a field evaluation involving noise and math nodes, ordered by the number of elements being evaluated: ``` 1,000,000: 133 ms -> 120 ms 100,000: 30 ms -> 18 ms 10,000: 20 ms -> 2.7 ms 1,000: 4 ms -> 0.5 ms 100: 0.5 ms -> 0.4 ms ```
2021-11-26 11:05:47 +01:00
/* The output is ignored by the caller, but the multi-function does not handle this case. So
* create a temporary buffer that the multi-function can write to. */
return this->ensure_dummy_single_output(data_index);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Geometry Nodes: support Noise Texture node This makes the Noise Texture node available in geometry nodes. It should behave the same as in shader node, with the exception that it does not have an implicit position input yet. That will be added separately. Differential Revision: https://developer.blender.org/D12467
2021-09-20 13:12:25 +02:00
/**
* Same as #uninitialized_single_output, but returns an empty span when the output is not
* required.
*/
template<typename T>
MutableSpan<T> uninitialized_single_output_if_required(int param_index, StringRef name = "")
{
return this->uninitialized_single_output_if_required(param_index, name).typed<T>();
}
GMutableSpan uninitialized_single_output_if_required(int param_index, StringRef name = "")
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::SingleOutput);
Geometry Nodes: support Noise Texture node This makes the Noise Texture node available in geometry nodes. It should behave the same as in shader node, with the exception that it does not have an implicit position input yet. That will be added separately. Differential Revision: https://developer.blender.org/D12467
2021-09-20 13:12:25 +02:00
int data_index = builder_->signature_->data_index(param_index);
return builder_->mutable_spans_[data_index];
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
template<typename T>
const VVectorArray<T> &readonly_vector_input(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
const GVVectorArray &vector_array = this->readonly_vector_input(param_index, name);
Cleanup: simplify resource scope methods Previously, a debug name had to be passed to all methods that added a resource to the `ResourceScope`. The idea was that this would make it easier to find certain bugs. In reality I never found this to be useful, and it was mostly annoying. The thing is, something that is in a resource scope never leaks (unless the resource scope is not destructed of course). Removing the name parameter makes the structure easier to use.
2021-09-14 16:08:09 +02:00
return builder_->scope_.construct<VVectorArray_For_GVVectorArray<T>>(vector_array);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
const GVVectorArray &readonly_vector_input(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::VectorInput);
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = builder_->signature_->data_index(param_index);
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
return *builder_->virtual_vector_arrays_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
template<typename T>
GVectorArray_TypedMutableRef<T> vector_output(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
return {this->vector_output(param_index, name)};
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
GVectorArray &vector_output(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::VectorOutput);
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = builder_->signature_->data_index(param_index);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
return *builder_->vector_arrays_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
template<typename T> MutableSpan<T> single_mutable(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
return this->single_mutable(param_index, name).typed<T>();
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
GMutableSpan single_mutable(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::SingleMutable);
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = builder_->signature_->data_index(param_index);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
return builder_->mutable_spans_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
template<typename T>
GVectorArray_TypedMutableRef<T> vector_mutable(int param_index, StringRef name = "")
Functions: Various improvements to the spans and generic data structures Most of this code is covered by unit tests.
2020-06-22 15:48:08 +02:00
{
Functions: refactor virtual array data structures When a function is executed for many elements (e.g. per point) it is often the case that some parameters are different for every element and other parameters are the same (there are some more less common cases). To simplify writing such functions one can use a "virtual array". This is a data structure that has a value for every index, but might not be stored as an actual array internally. Instead, it might be just a single value or is computed on the fly. There are various tradeoffs involved when using this data structure which are mentioned in `BLI_virtual_array.hh`. It is called "virtual", because it uses inheritance and virtual methods. Furthermore, there is a new virtual vector array data structure, which is an array of vectors. Both these types have corresponding generic variants, which can be used when the data type is not known at compile time. This is typically the case when building a somewhat generic execution system. The function system used these virtual data structures before, but now they are more versatile. I've done this refactor in preparation for the attribute processor and other features of geometry nodes. I moved the typed virtual arrays to blenlib, so that they can be used independent of the function system. One open question for me is whether all the generic data structures (and `CPPType`) should be moved to blenlib as well. They are well isolated and don't really contain any business logic. That can be done later if necessary.
2021-03-21 19:31:24 +01:00
return {this->vector_mutable(param_index, name)};
Functions: Various improvements to the spans and generic data structures Most of this code is covered by unit tests.
2020-06-22 15:48:08 +02:00
}
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
GVectorArray &vector_mutable(int param_index, StringRef name = "")
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
this->assert_correct_param(param_index, name, MFParamCategory::VectorMutable);
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
int data_index = builder_->signature_->data_index(param_index);
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
return *builder_->vector_arrays_[data_index];
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
private:
Refactor: Update integer type usage This updates the usage of integer types in code I wrote according to our new style guides. Major changes: * Use signed instead of unsigned integers in many places. * C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`). * Hash values for C++ containers are 64 bit wide now (instead of 32 bit). I do hope that I broke no builds, but it is quite likely that some compiler reports slightly different errors. Please let me know when there are any errors. If the fix is small, feel free to commit it yourself. I compiled successfully on linux with gcc and on windows.
2020-07-20 12:16:20 +02:00
void assert_correct_param(int param_index, StringRef name, MFParamType param_type)
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
UNUSED_VARS_NDEBUG(param_index, name, param_type);
#ifdef DEBUG
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(builder_->signature_->param_types[param_index] == param_type);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
if (name.size() > 0) {
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(builder_->signature_->param_names[param_index] == name);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
#endif
}
Geometry Nodes: refactor array devirtualization Goals: * Better high level control over where devirtualization occurs. There is always a trade-off between performance and compile-time/binary-size. * Simplify using array devirtualization. * Better performance for cases where devirtualization wasn't used before. Many geometry nodes accept fields as inputs. Internally, that means that the execution functions have to accept so called "virtual arrays" as inputs. Those can be e.g. actual arrays, just single values, or lazily computed arrays. Due to these different possible virtual arrays implementations, access to individual elements is slower than it would be if everything was just a normal array (access does through a virtual function call). For more complex execution functions, this overhead does not matter, but for small functions (like a simple addition) it very much does. The virtual function call also prevents the compiler from doing some optimizations (e.g. loop unrolling and inserting simd instructions). The solution is to "devirtualize" the virtual arrays for small functions where the overhead is measurable. Essentially, the function is generated many times with different array types as input. Then there is a run-time dispatch that calls the best implementation. We have been doing devirtualization in e.g. math nodes for a long time already. This patch just generalizes the concept and makes it easier to control. It also makes it easier to investigate the different trade-offs when it comes to devirtualization. Nodes that we've optimized using devirtualization before didn't get a speedup. However, a couple of nodes are using devirtualization now, that didn't before. Those got a 2-4x speedup in common cases. * Map Range * Random Value * Switch * Combine XYZ Differential Revision: https://developer.blender.org/D14628
2022-04-26 17:12:34 +02:00
void assert_correct_param(int param_index, StringRef name, MFParamCategory category)
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
{
UNUSED_VARS_NDEBUG(param_index, name, category);
#ifdef DEBUG
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(builder_->signature_->param_types[param_index].category() == category);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
if (name.size() > 0) {
Cleanup: use trailing underscore for non-public data members
2020-07-03 14:20:42 +02:00
BLI_assert(builder_->signature_->param_names[param_index] == name);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
}
#endif
}
Geometry Nodes: refactor multi-threading in field evaluation Previously, there was a fixed grain size for all multi-functions. That was not sufficient because some functions could benefit a lot from smaller grain sizes. This refactors adds a new `MultiFunction::call_auto` method which has the same effect as just calling `MultiFunction::call` but additionally figures out how to execute the specific multi-function efficiently. It determines a good grain size and decides whether the mask indices should be shifted or not. Most multi-function evaluations benefit from this, but medium sized work loads (1000 - 50000 elements) benefit from it the most. Especially when expensive multi-functions (e.g. noise) is involved. This is because for smaller work loads, threading is rarely used and for larger work loads threading worked fine before already. With this patch, multi-functions can specify execution hints, that allow the caller to execute it most efficiently. These execution hints still have to be added to more functions. Some performance measurements of a field evaluation involving noise and math nodes, ordered by the number of elements being evaluated: ``` 1,000,000: 133 ms -> 120 ms 100,000: 30 ms -> 18 ms 10,000: 20 ms -> 2.7 ms 1,000: 4 ms -> 0.5 ms 100: 0.5 ms -> 0.4 ms ```
2021-11-26 11:05:47 +01:00
GMutableSpan ensure_dummy_single_output(int data_index);
Functions: Multi Function This adds the `MultiFunction` type and some smallish utility types that it uses. A `MultiFunction` encapsulates a function that is optimized for throughput by always processing many elements at once. This is an important part of the new particle system, because it allows us to execute user generated node trees for many particles efficiently. Reviewers: brecht Differential Revision: https://developer.blender.org/D8030
2020-06-16 16:35:57 +02:00
};
Cleanup: use nested namespaces
2020-07-03 14:25:20 +02:00
} // namespace blender::fn
Copy Permalink