mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
d5cdc36943
Pull Request resolved: https://github.com/pytorch/pytorch/pull/156320 Approved by: https://github.com/albanD ghstack dependencies: #156318
73 lines
3.1 KiB
C++
73 lines
3.1 KiB
C++
#pragma once
|
|
|
|
#include <torch/csrc/jit/ir/ir.h>
|
|
|
|
#include <unordered_map>
|
|
#include <vector>
|
|
|
|
namespace torch::jit {
|
|
|
|
/**
|
|
* \brief A structure describing a match of a pattern in a graph.
|
|
*
|
|
* The structure contains an anchor node, from which the match was found, and
|
|
* match-maps for nodes and values. A match-map specifies the correspondence
|
|
* between nodes in the pattern graph (match-map keys) with nodes in the actual
|
|
* graph (match-map values). We keep such maps for both nodes and values.
|
|
*/
|
|
struct Match {
|
|
Node* anchor;
|
|
std::unordered_map<const Node*, Node*> nodes_map;
|
|
std::unordered_map<const Value*, Value*> values_map;
|
|
};
|
|
|
|
/**
|
|
* \brief Find all matches of a \p PATTERN in a \p GRAPH.
|
|
*
|
|
* The function returns a vector of match-descriptors (see description of
|
|
* `struct Match`).
|
|
*
|
|
* Matching rules:
|
|
* - Pattern graph must contain a single block.
|
|
* - Matched subgraphs do not span across different blocks.
|
|
* - No uses outside the match are allowed, except for Param and Return nodes.
|
|
* Basically, we're matching hammocks, not arbitrary subgraphs.
|
|
* - The pattern graph must return only one value (i.e. it must have a single
|
|
* node leading to return).
|
|
* - Nodes that are not used in computation of the return value in the pattern
|
|
* graph are ignored during matching (IOW, we're essentially performing DCE on
|
|
* the pattern).
|
|
* - Pattern graph nodes cannot alias. TODO: the check not implemented yet.
|
|
* - Aliasing nodes in the graph cannot constitute a match (i.e. through all
|
|
* found matches, no nodes in the subgraph alias with each other). TODO: check
|
|
* not implemented yet.
|
|
* - The matcher will not mutate either the pattern graph or the matched graph.
|
|
* The matched graph is taken as non-const so that Match may contain non-const
|
|
* pointers. This enables clients of this API to use Match to drive mutations.
|
|
*
|
|
* Note [Multi-output Patterns]
|
|
* ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
* Subgraph matcher provides limited support for multi-output patterns. With a
|
|
* single output pattern, a single scan through the graph is sufficient to
|
|
* find all the matches: given a starting node (an "anchor"), we can
|
|
* deterministically check whether a pattern matches a subgraph corresponding to
|
|
* this anchor node. For a general case of multi-output patterns, we would have
|
|
* N anchors, which would result in M^N comparisons (M is the size of the
|
|
* graph). Clearly this is computationally prohibitive.
|
|
*
|
|
* To overcome this, we impose some constraints on the multi-output patterns
|
|
* that we accept. We require that checking whether the pattern matches a
|
|
* subgraph would still be fully determined by a single node in the graph. To
|
|
* achieve this, we designate the first output in the pattern as the "main"
|
|
* output and assume that we can traverse up from this node to match the
|
|
* entire pattern.
|
|
*
|
|
* Corrolary 1: the order of outputs in the pattern matters!
|
|
* Corollary 2: patterns cannot contain any nodes not participating in the main
|
|
* output computation.
|
|
*/
|
|
std::vector<Match> TORCH_API
|
|
findPatternMatches(const Graph& pattern, Graph& graph);
|
|
|
|
} // namespace torch::jit
|