Skip to content

[Bug] Fix bug of core local value #4523

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
morningman merged 1 commit into from
Sep 4, 2020
Merged

[Bug] Fix bug of core local value #4523

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 7 additions & 2 deletions be/src/util/core_local.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@
#include <vector>

#include "common/logging.h"
#include "util/spinlock.h"

namespace doris {

Expand Down Expand Up @@ -48,8 +49,11 @@ class CoreDataAllocatorImpl : public CoreDataAllocator {
virtual ~CoreDataAllocatorImpl();
void* get_or_create(size_t id) override {
size_t block_id = id / ELEMENTS_PER_BLOCK;
if (block_id >= _blocks.size()) {
_blocks.resize(block_id + 1);
{
std::lock_guard<SpinLock> l(_lock);
if (block_id >= _blocks.size()) {
_blocks.resize(block_id + 1);
}
}
CoreDataBlock* block = _blocks[block_id];
if (block == nullptr) {
Expand All @@ -61,6 +65,7 @@ class CoreDataAllocatorImpl : public CoreDataAllocator {
}
private:
static constexpr int ELEMENTS_PER_BLOCK = BLOCK_SIZE / ELEMENT_BYTES;
SpinLock _lock; // lock to protect the modification of _blocks
Copy link
Contributor

@chaoyli chaoyli Sep 3, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

CoreData is to bind data to cpu to eliminate the lock.
If add a lock, it's design is meaningless.

Copy link
Contributor Author

@morningman morningman Sep 3, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is only used when creating the core local value (by calling get_or_create() of CoreDataAllocatorImpl).
And when creating the value, there is nothing to do with "cpu-bind".

The so called "cpu-bind" operation happens when accessing the CoreLocalValue,
not related to CoreDataAllocatorImpl

Copy link
Contributor

@chaoyli chaoyli Sep 3, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

OK, I have no problem.

std::vector<CoreDataBlock*> _blocks;
};

Expand Down