Skip to content
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.

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

Reuse the same reader across ReadChunk calls so that prefetched blocks are not discared. #82

Open
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

copybara-service[bot]
Copy link

Reuse the same reader across ReadChunk calls so that prefetched blocks are not discared.

In the sequential non-parallel path (ReadRecord->ReadAheadFromBuffer->ReadChunk) ReadChunk reads the whole chunk into memory. Most readers will round up the read to the blocksize and thereby prefetch some data from the next chunk.

Currently ReadChunk recreates the reader each time and the previously prefetched data is lost and must be read again.

Instead use get_backing_reader() directly each time. We can do this since ReadChunk is only called in single threaded context (Initialize, ReadRecord (parallelism disabled)) so there cannot be any concurrent access to the underlying reader.

ReadTrace (before):

offset=64, size=1M <-- 1st chunk
offset=1M, size=1M
...
offset=8M, size=1M <-- block aligned read crosses chunk boundary
offset=8.3M, size=0.7M <-- redundant read for 2nd chunk
offset=9M, size=1M
...

ReadTrace (after):

offset=64, size=1M <-- 1st chunk
offset=1M, size=1M
...
offset=8M, size=1M <-- block aligned read crosses chunk boundary and is reused for the 2nd chunk
offset=9M, size=1M 
...

@google-cla
Copy link

google-cla bot commented Oct 30, 2023

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

…s are not discared.

In the sequential non-parallel path (ReadRecord->ReadAheadFromBuffer->ReadChunk) ReadChunk reads the whole chunk into memory. Most readers will round up the read to the blocksize and thereby prefetch some data from the next chunk.

Currently ReadChunk recreates the reader each time and the previously prefetched data is lost and must be read again.

Instead use get_backing_reader() directly each time. We can do this since ReadChunk is only called in single threaded context (Initialize, ReadRecord (parallelism disabled)) so there cannot be any concurrent access to the underlying reader.

ReadTrace (before):
```
offset=64, size=1M <-- 1st chunk
offset=1M, size=1M
...
offset=8M, size=1M <-- block aligned read crosses chunk boundary
offset=8.3M, size=0.7M <-- redundant read for 2nd chunk
offset=9M, size=1M
...
```

ReadTrace (after):
```
offset=64, size=1M <-- 1st chunk
offset=1M, size=1M
...
offset=8M, size=1M <-- block aligned read crosses chunk boundary and is reused for the 2nd chunk
offset=9M, size=1M
...
```
PiperOrigin-RevId: 577312179
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

0 participants