saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-29 20:14:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fix CUDA 10.2 compile error due to inlined_containers.h inclusion (#10702)

Browse source 
Fix CUDA 10.2 compile error due to inlined_containers.h inclusion
 into a common CUDA header.
 Use NumberOfNodes() to reserve space in a hash table
 Prefer separate call to reserve() rather than passing in the
 hash table constructor. They have somewhat different meaning.
This commit is contained in:
Dmitri Smirnov 2022-02-28 19:56:44 -08:00 committed by GitHub
parent 3243c9579f
commit e23a224518
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 12 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
include/onnxruntime/core/framework/tensor_shape.h
View file

@ -10,8 +10,8 @@
#include <gsl/gsl>
#include "onnxruntime_config.h"
// I have to bring it here because including inline_containers.h
// causes CUDA 10.2 compilers to fail
// Need to include abseil inlined_vector.h header directly here
// as hash tables cause CUDA 10.2 compilers to fail. inlined_vector.h is fine.
#ifdef _MSC_VER
#pragma warning(push)
// C4127: conditional expression is constant

2
include/onnxruntime/core/graph/graph_nodes.h
View file

@ -77,8 +77,6 @@ class ValidNodes {
bool empty() const noexcept { return nodes_->empty(); }
size_t size() const noexcept { return nodes_->size(); }
/**
@class NodeIterator
Iterator to provide const and non-const access to valid Node instances in a Graph.

3
onnxruntime/contrib_ops/cpu/unique.cc
View file

@ -46,7 +46,8 @@ Status Unique<float>::Compute(OpKernelContext* ctx) const {
// XXX: Refactoring for less memory allocations. unordered_map
// used originally for float uniqueness, is this correct?
using IndexingMap = InlinedHashMap<float, ElementData>;
IndexingMap mapped_indices(num_elements);
IndexingMap mapped_indices;
mapped_indices.reserve(num_elements);
// processing
for (int64_t i = 0; i < num_elements; ++i) {

3
onnxruntime/core/framework/allocation_planner.cc
View file

@ -500,7 +500,8 @@ class PlannerImpl {
// Note: for every ml-value, its definition must appear before all its uses in a topological sort of a valid model
using GraphInputsSet = InlinedHashSet<std::string_view>;
const auto& graph_inputs_nodes = graph_viewer_.GetInputsIncludingInitializers();
GraphInputsSet graph_inputs(graph_inputs_nodes.size());
GraphInputsSet graph_inputs;
graph_inputs.reserve(graph_inputs_nodes.size());
for (auto& graph_input : graph_inputs_nodes) {
graph_inputs.insert(graph_input->Name());
}

2
onnxruntime/core/framework/session_state.cc
View file

@ -859,7 +859,7 @@ void SessionState::UpdateToBeExecutedNodes(gsl::span<int const> fetch_mlvalue_id
InlinedVector<const Node*> nodes;
nodes.reserve(fetch_mlvalue_idxs.size());
InlinedHashSet<NodeIndex> reachable_nodes;
reachable_nodes.reserve(graph_.Nodes().size());
reachable_nodes.reserve(graph_.NumberOfNodes());
for (auto idx : fetch_mlvalue_idxs) {
std::string node_arg_name;

6
onnxruntime/core/providers/cpu/tensor/slice_helper.h
View file

@ -33,7 +33,8 @@ inline Status PrepareForComputeHelper(const gsl::span<const int64_t>& raw_starts
// Iterate through the provided axes and override the start/end ranges
using AxesSet = InlinedHashSet<int64_t>;
const auto axes_count = axes.size();
AxesSet unique_axes(axes_count);
AxesSet unique_axes;
unique_axes.reserve(axes_count);
const auto dimension_count = compute_metadata.input_dimensions_.size();
for (size_t axis_index = 0; axis_index < axes_count; ++axis_index) {
@ -91,7 +92,8 @@ inline Status PrepareForComputeHelper(const gsl::span<const int64_t>& raw_starts
// Iterate through the provided axes and override the start/end/steps ranges
using AxesSet = InlinedHashSet<int64_t>;
const auto axes_count = axes.size();
AxesSet unique_axes(axes_count);
AxesSet unique_axes;
unique_axes.reserve(axes_count);
const auto dimension_count = compute_metadata.input_dimensions_.size();
for (size_t axis_index = 0; axis_index < axes_count; ++axis_index) {

3
onnxruntime/core/providers/cuda/cuda_allocator.h
View file

@ -5,7 +5,6 @@
#include <unordered_set>
#include "core/framework/allocator.h"
#include "core/common/inlined_containers.h"
#include "core/platform/ort_mutex.h"
namespace onnxruntime {
@ -48,7 +47,7 @@ class CUDAExternalAllocator : public CUDAAllocator {
ExternalAlloc alloc_;
ExternalFree free_;
ExternalEmptyCache empty_cache_;
InlinedHashSet<void*> reserved_;
std::unordered_set<void*> reserved_;
};
//TODO: add a default constructor