| #include <torch/csrc/jit/passes/peephole.h> |
| #include <torch/csrc/jit/ir_views.h> |
| #include <torch/csrc/jit/jit_log.h> |
| #include <torch/csrc/jit/passes/dead_code_elimination.h> |
| #include <torch/csrc/jit/symbolic_variable.h> |
| |
| namespace torch { |
| namespace jit { |
| |
| // Conservatively compare two optionals. If both are undefined, assume |
| // they aren't equal |
| template <typename T> |
| static bool mustBeEqual(const c10::optional<T>& a, const c10::optional<T>& b) { |
| return a == b && a.has_value(); |
| } |
| |
| // The intent for this optimization pass is to catch all of the small, easy to |
| // catch peephole optimizations you might be interested in doing. |
| // |
| // Right now, it does: |
| // - Eliminate no-op 'expand' nodes |
| // - Simply x.t().t() to x |
| // |
| // TODO: Decide what kind of fixed point strategy we will have |
| // |
| // The parameter `addmm_fusion_enabled` exists because, as it is today, fusing |
| // add + mm has no benefit within PyTorch running ATen ops. However, we rely on |
| // seeing the fused version of addmm for ONNX export, since after ONNX |
| // translation we would see redundant Gemm ops with sub-optimal inputs. This |
| // flag is exposed so that ONNX export can pass `true` to get the fused |
| // behavior, but normal JIT peephole optimization is left alone. |
| void PeepholeOptimizeImpl(Block* block, bool addmm_fusion_enabled) { |
| for (auto it = block->nodes().begin(); it != block->nodes().end(); ++it) { |
| auto* node = *it; |
| |
| for (Block* sub_block : node->blocks()) { |
| PeepholeOptimizeImpl(sub_block, addmm_fusion_enabled); |
| } |
| |
| // XXX: remember that if you want to simplify an expression by combining |
| // multiple nodes into a different one, then you need to check that they all |
| // belong to the given block |
| if (node->matches( |
| "aten::expand(Tensor self, int[] size, *, bool implicit) -> Tensor", |
| /*const_inputs=*/attr::size)) { |
| // x.expand(x.size()) == x |
| if (auto input_type = node->namedInput(attr::self) |
| ->type() |
| ->cast<CompleteTensorType>()) { |
| auto expanded_sizes = node->get<c10::List<int64_t>>(attr::size); |
| if (!expanded_sizes.has_value() || c10::impl::toVector(*expanded_sizes) == input_type->sizes()) { |
| GRAPH_UPDATE( |
| *node, |
| " (x.expand(x.size()) == x) is replaced with ", |
| node->namedInput(attr::self)->debugName()); |
| node->output()->replaceAllUsesWith(node->namedInput(attr::self)); |
| } |
| } |
| } else if (node->matches("aten::t(Tensor self) -> Tensor")) { |
| // x.t().t() == x |
| Node* input_node = node->input()->node(); |
| if (input_node->matches("aten::t(Tensor self) -> Tensor")) { |
| GRAPH_UPDATE( |
| *node, |
| " (x.t().t() == x) is replaced with ", |
| input_node->input()->debugName()); |
| node->output()->replaceAllUsesWith(input_node->input()); |
| } |
| } else if (node->matches( |
| "aten::type_as(Tensor self, Tensor other) -> Tensor")) { |
| // x.type_as(y) == x iff x.type() == y.type() |
| auto self_type = ProfiledTensorType::create(node->input(0)->type()); |
| auto other_type = ProfiledTensorType::create(node->input(1)->type()); |
| if (mustBeEqual(self_type->scalarType(), other_type->scalarType()) && |
| mustBeEqual(self_type->device(), other_type->device())) { |
| GRAPH_UPDATE( |
| *node, |
| " (x.type_as(y) == x) is replaced with ", |
| node->input(0)->debugName()); |
| node->output()->replaceAllUsesWith(node->input(0)); |
| } |
| } else if ( |
| node->matches( |
| "aten::add(Tensor self, Tensor other, *, Scalar alpha) -> Tensor", |
| /*const_inputs=*/attr::alpha)) { |
| // z + x.mm(y) == z.addmm(x, y) == x.mm(y) + z |
| // This optimization has been disabled at the moment, because it's not |
| // helpful at all until we will be able to represent torch.addmm(a, b, c, |
| // out=a). That's because addmm dispatches internally to gemm, which |
| // computes: |
| // C = beta * C + alpha * A @ B |
| // but aten::addmm(a, b, c, 1, 1) is really: |
| // D = beta * C + alpha * A @ B |
| // and because it works out of place on C, we're only trading off an |
| // explicit add for a copy inside the addmm function. Note that it doesn't |
| // even result in fewer reads, because mm won't even load C (because beta |
| // == 0 for it). |
| if (addmm_fusion_enabled && |
| node->get<at::Scalar>(attr::alpha).value().toDouble() == 1.) { |
| // Look for mm from both sides of the add |
| for (size_t mm_side = 0; mm_side < 2; mm_side++) { |
| // Add will accept tensors of mismatched scalar types, as long as one |
| // of them is a scalar. Addmm will throw in that case, so we can only |
| // perform this fusion if we're sure that it is correct, and for that |
| // we need the add_mat_type. An alternative would be to insert a |
| // type_as conditional on the tensor shape being a scalar, but that |
| // might add overhead, and make analysis harder. |
| auto add_mat_type = |
| ProfiledTensorType::create(node->input(1 - mm_side)->type()); |
| // if we don't have the rank, we can't tell if the bias is a scalar |
| if (!add_mat_type->sizes().size()) { |
| continue; |
| } |
| |
| if (node->input(mm_side)->node()->matches( |
| "aten::mm(Tensor self, Tensor mat2) -> Tensor")) { |
| WithInsertPoint guard(node); |
| |
| auto mm_node = node->input(mm_side)->node(); |
| SymbolicVariable add_mat(node->input(1 - mm_side)); |
| SymbolicVariable mat1(mm_node->input(0)); |
| SymbolicVariable mat2(mm_node->input(1)); |
| |
| auto mat1_type = ProfiledTensorType::create(mat1.value()->type()); |
| auto mat_scalar_type = mat1_type->scalarType(); |
| if (!mat_scalar_type) { |
| auto mat2_type = ProfiledTensorType::create(mat2.value()->type()); |
| mat_scalar_type = mat2_type->scalarType(); |
| } |
| |
| // we can't use type_as if we don't know the target type (mm), the |
| // bias needs to be coerced to |
| if (!mat_scalar_type) { |
| continue; |
| } |
| |
| // We insert the type_as if we're sure that the added element is a |
| // scalar, and we either don't know what is the type of the |
| // scalar, or know the type, and know that it's |
| // mismatched. |
| if (add_mat_type->sizes().size() && |
| *add_mat_type->sizes().size() == 0 && |
| !mustBeEqual(add_mat_type->scalarType(), mat_scalar_type)) { |
| add_mat = add_mat.type_as(mat1); |
| } |
| |
| SymbolicVariable addmm_value = add_mat.addmm(mat1, mat2); |
| |
| // Copy shape information from output node |
| ((Value*)addmm_value)->copyMetadata(node->output()); |
| GRAPH_UPDATE( |
| "Fusing ", |
| mm_node->input(0)->debugName(), |
| ", ", |
| mm_node->input(1)->debugName(), |
| " and ", |
| node->input(1 - mm_side)->debugName(), |
| " into ", |
| addmm_value.value()->debugName()); |
| node->output()->replaceAllUsesWith(addmm_value); |
| } |
| } |
| } |
| // TODO: this doesn't work with Scalar-Tensor ops! We should canonicalize |
| // those |
| } else if ( |
| node->matches( |
| "aten::mul(Tensor self, Scalar other) -> Tensor", |
| /*const_inputs=*/attr::other) || |
| node->matches( |
| "aten::div(Tensor self, Scalar other) -> Tensor", |
| /*const_inputs=*/attr::other)) { |
| // x * 1 == x / 1 == x |
| if (node->get<at::Scalar>(attr::other)->toDouble() == 1) { |
| GRAPH_UPDATE( |
| *node, |
| " (x * 1 == x / 1 == x) is replaced with ", |
| node->input(0)->debugName()); |
| node->output()->replaceAllUsesWith(node->input(0)); |
| } |
| } else if ( |
| node->matches( |
| "aten::add(Tensor self, Scalar other, Scalar alpha) -> Tensor", |
| /*const_inputs=*/{attr::alpha, attr::other}) || |
| node->matches( |
| "aten::sub(Tensor self, Scalar other, Scalar alpha) -> Tensor", |
| /*const_inputs=*/{attr::alpha, attr::other})) { |
| // x + 0 == x - 0 == x |
| if (node->get<at::Scalar>(attr::alpha)->toDouble() == 1 && |
| node->get<at::Scalar>(attr::other)->toDouble() == 0) { |
| GRAPH_UPDATE( |
| *node, |
| " (x + 0 == x - 0 == x) is replaced with ", |
| node->input(0)->debugName()); |
| node->output()->replaceAllUsesWith(node->input(0)); |
| } |
| } else if ( |
| node->kind() == aten::Float || node->kind() == aten::Int || |
| node->kind() == prim::ImplicitTensorToNum) { |
| Node* input_node = node->input()->node(); |
| if (input_node->kind() == prim::NumToTensor) { |
| GRAPH_UPDATE( |
| *node, |
| " (x.NumToTensor().ImplicitTensorToNum() == x.NumToTensor()) is replaced with ", |
| node->input()->debugName()); |
| node->output()->replaceAllUsesWith(input_node->input()); |
| } |
| } else if ( |
| node->matches( |
| "aten::_grad_sum_to_size(Tensor(a) self, int[]? size) -> Tensor(a)")) { |
| if (node->input(1)->mustBeNone()) { |
| GRAPH_UPDATE( |
| *node, |
| " (x._grad_sum_to_size(x, None) == x) is replaced with ", |
| node->input(0)->debugName()); |
| node->output()->replaceAllUsesWith(node->input(0)); |
| } else { |
| auto uses = node->output()->uses(); |
| for (Use u : uses) { |
| if (u.user->matches( |
| "aten::_grad_sum_to_size(Tensor(a) self, int[]? size) -> Tensor(a)") && |
| u.user->input(1)->type()->isSubtypeOf(ListType::ofInts())) { |
| GRAPH_UPDATE( |
| *node, |
| " (x._grad_sum_to_size(y)._grad_sum_to_size(z) == x._grad_sum_to_size(z)) is replaced with ", |
| node->inputs().at(0)->debugName()); |
| u.user->replaceInput(0, node->inputs().at(0)); |
| } |
| } |
| } |
| } else if (node->kind() == prim::If) { |
| IfView n(node); |
| // this handles redundant short circuits like "x and True" or "x or False" |
| for (size_t i = 0; i < n.outputs().size(); ++i) { |
| if (n.outputs().at(i)->type() != BoolType::get()) { |
| continue; |
| } |
| bool true_val = |
| constant_as<bool>(n.thenOutputs().at(i)).value_or(false); |
| bool false_val = |
| constant_as<bool>(n.elseOutputs().at(i)).value_or(true); |
| // if an if node's output equals its condition replace output with |
| // condition |
| if (true_val && !false_val) { |
| GRAPH_UPDATE( |
| "Replacing ", |
| n.outputs().at(i)->debugName(), |
| " (True or False) with ", |
| n.cond()->debugName()); |
| n.outputs().at(i)->replaceAllUsesWith(n.cond()); |
| } |
| } |
| } else if ( |
| node->kind() == aten::__is__ || node->kind() == aten::__isnot__) { |
| // if we are comparing a None value with a value that can't be None |
| // replace the output with true if node is __isnot__ or false if node is |
| // __is__ |
| AT_ASSERT(node->inputs().size() == 2); |
| for (size_t check_none_index : {0, 1}) { |
| bool input_must_be_none = |
| node->inputs().at(check_none_index)->mustBeNone(); |
| bool other_must_not_be_none = |
| node->inputs().at(1 - check_none_index)->mustNotBeNone(); |
| if (input_must_be_none && other_must_not_be_none) { |
| WithInsertPoint guard(node); |
| auto output = node->owningGraph()->insertConstant( |
| node->kind() == aten::__isnot__); |
| GRAPH_UPDATE("Folding ", *node, " to ", output->debugName()); |
| node->output()->replaceAllUsesWith(output); |
| } |
| } |
| } else if ( |
| node->kind() == prim::unchecked_unwrap_optional || |
| node->kind() == aten::_unwrap_optional) { |
| // we are unwrapping an input that can't be None, remove the unwrap |
| auto input = node->input(); |
| if (input->mustNotBeNone()) { |
| GRAPH_UPDATE( |
| "Unwrapping ", *node, " as ", node->input(), " can't be optional"); |
| node->output()->replaceAllUsesWith(node->input()); |
| } |
| } else if (node->matches("prim::dtype(Tensor a) -> int")) { |
| auto ptt = ProfiledTensorType::create(node->input()->type()); |
| if (ptt->scalarType()) { |
| WithInsertPoint guard(node); |
| auto output = node->owningGraph()->insertConstant( |
| static_cast<int64_t>(*ptt->scalarType())); |
| GRAPH_UPDATE( |
| "Replacing ", *node, " with a type constant ", output->debugName()); |
| node->output()->replaceAllUsesWith(output); |
| } |
| } else if (node->matches("prim::device(Tensor a) -> Device")) { |
| auto ptt = ProfiledTensorType::create(node->input()->type()); |
| if (ptt->device()) { |
| WithInsertPoint guard(node); |
| auto output = node->owningGraph()->insertConstant(*ptt->device()); |
| GRAPH_UPDATE( |
| "Replacing ", |
| *node, |
| " with a device constant ", |
| output->debugName()); |
| node->output()->replaceAllUsesWith(output); |
| } |
| } else if (node->matches("aten::dim(Tensor self) -> int")) { |
| auto ptt = ProfiledTensorType::create(node->input()->type()); |
| if (auto dim = ptt->sizes().size()) { |
| WithInsertPoint guard(node); |
| auto output = |
| node->owningGraph()->insertConstant(static_cast<int64_t>(*dim)); |
| GRAPH_UPDATE( |
| "Replacing ", |
| *node, |
| " with a \"dim\" constant ", |
| output->debugName()); |
| node->output()->replaceAllUsesWith(output); |
| } |
| } else if (node->matches("prim::is_cuda(Tensor a) -> bool")) { |
| auto ptt = ProfiledTensorType::create(node->input()->type()); |
| if (ptt->device()) { |
| WithInsertPoint guard(node); |
| auto output = |
| node->owningGraph()->insertConstant((*ptt->device()).is_cuda()); |
| GRAPH_UPDATE( |
| "Replacing ", |
| *node, |
| " with a is_cuda constant ", |
| output->debugName()); |
| node->output()->replaceAllUsesWith(output); |
| } |
| } |
| } |
| } |
| |
| void PeepholeOptimize(Block* block, bool addmm_fusion_enabled) { |
| PeepholeOptimizeImpl(block, addmm_fusion_enabled); |
| GRAPH_DUMP("After PeepholeOptimize: ", block->owningGraph()); |
| // Eliminate dead code created by any peephole passes we've just done |
| EliminateDeadCode(block); |
| } |
| |
| void PeepholeOptimize( |
| const std::shared_ptr<Graph>& graph, |
| bool addmm_fusion_enabled) { |
| PeepholeOptimize(graph->block(), addmm_fusion_enabled); |
| } |
| } // namespace jit |
| } // namespace torch |
