You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "viirya (via GitHub)" <gi...@apache.org> on 2024/04/07 07:25:00 UTC
[PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
viirya opened a new pull request, #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246
## Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123.
-->
Closes #.
## Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes.
-->
When we export row addresses and sizes to native, they are copied to separate arrays. We should deallocate array buffers after exporting them to reduce memory allocation.
## What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR.
-->
## How are these changes tested?
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code
If tests are not included in your PR, please explain why (for example, are they covered by existing tests)?
-->
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya merged PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on code in PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246#discussion_r1555019827
##########
spark/src/main/scala/org/apache/spark/shuffle/sort/RowPartition.scala:
##########
@@ -32,8 +32,17 @@ class RowPartition(initialSize: Int) {
def getNumRows: Int = rowAddresses.size
- def getRowAddresses: Array[Long] = rowAddresses.toArray
- def getRowSizes: Array[Int] = rowSizes.toArray
+ def getRowAddresses: Array[Long] = {
+ val array = rowAddresses.toArray
+ rowAddresses = null
+ array
+ }
+
+ def getRowSizes: Array[Int] = {
+ val array = rowSizes.toArray
+ rowSizes = null
+ array
+ }
Review Comment:
They won't be accessed again after `getRowAddresses` and `getRowSizes` get called. Once they are called, they are exported to native side for spilling. After that, `RowPartition` will be reset.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "advancedxy (via GitHub)" <gi...@apache.org>.
advancedxy commented on code in PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246#discussion_r1555201096
##########
spark/src/main/scala/org/apache/spark/shuffle/sort/RowPartition.scala:
##########
@@ -32,8 +32,17 @@ class RowPartition(initialSize: Int) {
def getNumRows: Int = rowAddresses.size
- def getRowAddresses: Array[Long] = rowAddresses.toArray
- def getRowSizes: Array[Int] = rowSizes.toArray
+ def getRowAddresses: Array[Long] = {
+ val array = rowAddresses.toArray
+ rowAddresses = null
+ array
+ }
+
+ def getRowSizes: Array[Int] = {
+ val array = rowSizes.toArray
+ rowSizes = null
+ array
+ }
Review Comment:
> They won't be accessed again after getRowAddresses and getRowSizes get called
I see. Maybe I'm being nitpicking. I think caller would be surprise to see the state change of `RowPartition` after calling a getter method. A more appropriate method name or some comments would help.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246#issuecomment-2041349034
cc @sunchao @parthchandra
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "advancedxy (via GitHub)" <gi...@apache.org>.
advancedxy commented on code in PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246#discussion_r1554854256
##########
spark/src/main/scala/org/apache/spark/shuffle/sort/RowPartition.scala:
##########
@@ -32,8 +32,17 @@ class RowPartition(initialSize: Int) {
def getNumRows: Int = rowAddresses.size
- def getRowAddresses: Array[Long] = rowAddresses.toArray
- def getRowSizes: Array[Int] = rowSizes.toArray
+ def getRowAddresses: Array[Long] = {
+ val array = rowAddresses.toArray
+ rowAddresses = null
+ array
+ }
+
+ def getRowSizes: Array[Int] = {
+ val array = rowSizes.toArray
+ rowSizes = null
+ array
+ }
Review Comment:
hmmm, these are writing operations, I don't think it's suitable to put them in the getters? Besides, the `getNumRows` may still access `rowAddresses`.
How about adding a new method?
```java
def destructiveRowAddressAndSizes: (Array[Long], Array[Int])
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246#issuecomment-2041599080
Merged. Thanks.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] fix: Deallocate row addresses and size arrays after exporting [arrow-datafusion-comet]
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on code in PR #246:
URL: https://github.com/apache/arrow-datafusion-comet/pull/246#discussion_r1555020334
##########
spark/src/main/scala/org/apache/spark/shuffle/sort/RowPartition.scala:
##########
@@ -32,8 +32,17 @@ class RowPartition(initialSize: Int) {
def getNumRows: Int = rowAddresses.size
- def getRowAddresses: Array[Long] = rowAddresses.toArray
- def getRowSizes: Array[Int] = rowSizes.toArray
+ def getRowAddresses: Array[Long] = {
+ val array = rowAddresses.toArray
+ rowAddresses = null
+ array
+ }
+
+ def getRowSizes: Array[Int] = {
+ val array = rowSizes.toArray
+ rowSizes = null
+ array
+ }
Review Comment:
i.e.,
```
long addrs[] = getRowAddresses();
long sizes[] = getRowSizes;
nativeSpilling(adds, sizes);
rowPartition.reset();
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org