Files
test/source/blender/functions/FN_multi_function.hh

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

151 lines
4.8 KiB
C++
Raw Normal View History

/* SPDX-FileCopyrightText: 2023 Blender Authors
*
* SPDX-License-Identifier: GPL-2.0-or-later */
#pragma once
/** \file
* \ingroup fn
*
* A `MultiFunction` encapsulates a function that is optimized for throughput (instead of latency).
* The throughput is optimized by always processing many elements at once, instead of each element
* separately. This is ideal for functions that are evaluated often (e.g. for every particle).
*
* By processing a lot of data at once, individual functions become easier to optimize for humans
* and for the compiler. Furthermore, performance profiles become easier to understand and show
* better where bottlenecks are.
*
* Every multi-function has a name and an ordered list of parameters. Parameters are used for input
* and output. In fact, there are three kinds of parameters: inputs, outputs and mutable (which is
* combination of input and output).
*
* To call a multi-function, one has to provide three things:
* - `Params`: This references the input and output arrays that the function works with. The
* arrays are not owned by Params.
* - `IndexMask`: An array of indices indicating which indices in the provided arrays should be
* touched/processed.
* - `Context`: Further information for the called function.
*
* A new multi-function is generally implemented as follows:
* 1. Create a new subclass of MultiFunction.
* 2. Implement a constructor that initialized the signature of the function.
* 3. Override the `call` function.
*/
#include "BLI_hash.hh"
#include "FN_multi_function_context.hh"
#include "FN_multi_function_params.hh"
namespace blender::fn::multi_function {
class MultiFunction : NonCopyable, NonMovable {
private:
const Signature *signature_ref_ = nullptr;
public:
virtual ~MultiFunction() = default;
/**
* The result is the same as using #call directly but this method has some additional features.
* - Automatic multi-threading when possible and appropriate.
* - Automatic index mask offsetting to avoid large temporary intermediate arrays that are mostly
* unused.
*/
BLI: refactor IndexMask for better performance and memory usage Goals of this refactor: * Reduce memory consumption of `IndexMask`. The old `IndexMask` uses an `int64_t` for each index which is more than necessary in pretty much all practical cases currently. Using `int32_t` might still become limiting in the future in case we use this to index e.g. byte buffers larger than a few gigabytes. We also don't want to template `IndexMask`, because that would cause a split in the "ecosystem", or everything would have to be implemented twice or templated. * Allow for more multi-threading. The old `IndexMask` contains a single array. This is generally good but has the problem that it is hard to fill from multiple-threads when the final size is not known from the beginning. This is commonly the case when e.g. converting an array of bool to an index mask. Currently, this kind of code only runs on a single thread. * Allow for efficient set operations like join, intersect and difference. It should be possible to multi-thread those operations. * It should be possible to iterate over an `IndexMask` very efficiently. The most important part of that is to avoid all memory access when iterating over continuous ranges. For some core nodes (e.g. math nodes), we generate optimized code for the cases of irregular index masks and simple index ranges. To achieve these goals, a few compromises had to made: * Slicing of the mask (at specific indices) and random element access is `O(log #indices)` now, but with a low constant factor. It should be possible to split a mask into n approximately equally sized parts in `O(n)` though, making the time per split `O(1)`. * Using range-based for loops does not work well when iterating over a nested data structure like the new `IndexMask`. Therefor, `foreach_*` functions with callbacks have to be used. To avoid extra code complexity at the call site, the `foreach_*` methods support multi-threading out of the box. The new data structure splits an `IndexMask` into an arbitrary number of ordered `IndexMaskSegment`. Each segment can contain at most `2^14 = 16384` indices. The indices within a segment are stored as `int16_t`. Each segment has an additional `int64_t` offset which allows storing arbitrary `int64_t` indices. This approach has the main benefits that segments can be processed/constructed individually on multiple threads without a serial bottleneck. Also it reduces the memory requirements significantly. For more details see comments in `BLI_index_mask.hh`. I did a few tests to verify that the data structure generally improves performance and does not cause regressions: * Our field evaluation benchmarks take about as much as before. This is to be expected because we already made sure that e.g. add node evaluation is vectorized. The important thing here is to check that changes to the way we iterate over the indices still allows for auto-vectorization. * Memory usage by a mask is about 1/4 of what it was before in the average case. That's mainly caused by the switch from `int64_t` to `int16_t` for indices. In the worst case, the memory requirements can be larger when there are many indices that are very far away. However, when they are far away from each other, that indicates that there aren't many indices in total. In common cases, memory usage can be way lower than 1/4 of before, because sub-ranges use static memory. * For some more specific numbers I benchmarked `IndexMask::from_bools` in `index_mask_from_selection` on 10.000.000 elements at various probabilities for `true` at every index: ``` Probability Old New 0 4.6 ms 0.8 ms 0.001 5.1 ms 1.3 ms 0.2 8.4 ms 1.8 ms 0.5 15.3 ms 3.0 ms 0.8 20.1 ms 3.0 ms 0.999 25.1 ms 1.7 ms 1 13.5 ms 1.1 ms ``` Pull Request: https://projects.blender.org/blender/blender/pulls/104629
2023-05-24 18:11:41 +02:00
void call_auto(const IndexMask &mask, Params params, Context context) const;
virtual void call(const IndexMask &mask, Params params, Context context) const = 0;
virtual uint64_t hash() const
{
2021-03-25 16:01:28 +01:00
return get_default_hash(this);
}
virtual bool equals(const MultiFunction & /*other*/) const
{
return false;
}
int param_amount() const
2020-07-12 12:38:03 +02:00
{
return signature_ref_->params.size();
2020-07-12 12:38:03 +02:00
}
IndexRange param_indices() const
{
return signature_ref_->params.index_range();
}
ParamType param_type(int param_index) const
{
return signature_ref_->params[param_index].type;
}
StringRefNull param_name(int param_index) const
{
return signature_ref_->params[param_index].name;
}
StringRefNull name() const
{
return signature_ref_->function_name;
}
virtual std::string debug_name() const;
const Signature &signature() const
{
BLI_assert(signature_ref_ != nullptr);
return *signature_ref_;
}
/**
* Information about how the multi-function behaves that help a caller to execute it efficiently.
*/
struct ExecutionHints {
/**
* Suggested minimum workload under which multi-threading does not really help.
* This should be lowered when the multi-function is doing something computationally expensive.
*/
int64_t min_grain_size = 10000;
/**
* Indicates that the multi-function will allocate an array large enough to hold all indices
* passed in as mask. This tells the caller that it would be preferable to pass in smaller
* indices. Also maybe the full mask should be split up into smaller segments to decrease peak
* memory usage.
*/
bool allocates_array = false;
/**
* Tells the caller that every execution takes about the same time. This helps making a more
* educated guess about a good grain size.
*/
bool uniform_execution_time = true;
};
ExecutionHints execution_hints() const;
protected:
/* Make the function use the given signature. This should be called once in the constructor of
* child classes. No copy of the signature is made, so the caller has to make sure that the
* signature lives as long as the multi function. It is ok to embed the signature into the child
* class. */
void set_signature(const Signature *signature)
{
/* Take a pointer as argument, so that it is more obvious that no copy is created. */
BLI_assert(signature != nullptr);
signature_ref_ = signature;
}
virtual ExecutionHints get_execution_hints() const;
};
inline ParamsBuilder::ParamsBuilder(const MultiFunction &fn, const IndexMask *mask)
: ParamsBuilder(fn.signature(), *mask)
{
}
} // namespace blender::fn::multi_function
namespace blender {
namespace mf = fn::multi_function;
}