Change to streaming out the heap snapshot data #52854
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
* Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
This was referenced Jan 11, 2024
|
Since the current changes do not involve documentation, CI errors are irrelevant. |
giordano
changed the title
|
I'm not super familiar with this code, I've only modified it a bit once, but the changes seems pretty reasonable. Did you folks do some testing to see if chrome is still able to process the file that gets regenerated from the json? |
Yes, we tested the snapshot with different sizes from hundreds of MBs to couple GBs, chrome was still able to process the assembled files. |
d-netto
reviewed
Jan 30, 2024
Drvi
added a commit
to RelationalAI/julia
that referenced
this pull request
Feb 1, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Feb 6, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Feb 7, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Feb 14, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Feb 21, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Feb 22, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
Drvi
added a commit
to RelationalAI/julia
that referenced
this pull request
Feb 28, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Mar 1, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Mar 13, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
Drvi
added a commit
to RelationalAI/julia
that referenced
this pull request
Apr 3, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
d-netto
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Apr 16, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Apr 23, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Apr 24, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Apr 30, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
Apr 30, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
May 2, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
IanButterworth
added
the
backport 1.10
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
May 9, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
May 19, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
May 26, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
May 28, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
DelveCI
pushed a commit
to RelationalAI/julia
that referenced
this pull request
May 29, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
KristofferC
added a commit
that referenced
this pull request
May 30, 2024
Backported PRs: - [x] #54010 <!-- Overload `Base.literal_pow` for `AbstractQ` --> - [x] #54143 <!-- Fix `make install` from tarballs --> - [x] #54151 <!-- LinearAlgebra: Correct zero element in `_generic_matvecmul!` for block adj/trans --> - [x] #54233 <!-- set MAX_OS_WRITE on unix --> - [x] #54251 <!-- fix typo in gc_mark_memory8 when chunking a large array --> - [x] #54363 <!-- typeintersect: fix another stack overflow caused by circular constraints --> - [x] #54497 <!-- Make TestLogger thread-safe (introduce a lock) --> - [x] #53796 <!-- Add a missing doc --> - [x] #54465 <!-- typeintersect: conservative typevar subtitution during `finish_unionall` --> - [x] #54514 <!-- typeintersect: followup cleanup for the nothrow path of type instantiation --> Need manual backport: - [ ] #52505 <!-- fix alignment of emit_unbox_store copy --> - [ ] #53373 <!-- fix sysimage-native-code=no option with pkgimages --> - [ ] #53815 <!-- create phantom task for GC threads --> - [ ] #53984 <!-- Profile: fix heap snapshot is valid char check --> - [ ] #54276 <!-- Fix solve for complex `Hermitian` with non-vanishing imaginary part on diagonal --> Contains multiple commits, manual intervention needed: - [ ] #52854 <!-- Change to streaming out the heap snapshot data --> - [ ] #53218 <!-- Fix interpreter_exec.jl test --> - [ ] #53833 <!-- Profile: make heap snapshots viewable in vscode viewer --> - [ ] #54303 <!-- LinearAlgebra: improve type-inference in Symmetric/Hermitian matmul --> - [ ] #52694 <!-- Reinstate similar for AbstractQ for backward compatibility --> Non-merged PRs with backport label: - [ ] #54471 <!-- Actually setup jit targets when compiling packageimages instead of targeting only one --> - [ ] #53452 <!-- RFC: allow Tuple{Union{}}, returning Union{} --> - [ ] #51479 <!-- prevent code loading from lookin in the versioned environment when building Julia -->
Drvi
added a commit
to RelationalAI/julia
that referenced
this pull request
Jun 7, 2024
This PR is to continue the work on the following PR: Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (JuliaLang#51518 ) Here are the commit history: ``` * Streaming the heap snapshot! This should prevent the engine from OOMing while recording the snapshot! Now we just need to sample the files, either online, before downloading, or offline after downloading :) If we're gonna do it offline, we'll want to gzip the files before downloading them. * Allow custom filename; use original API * Support legacy heap snapshot interface. Add reassembly function. * Add tests * Apply suggestions from code review * Update src/gc-heap-snapshot.cpp * Change to always save the parts in the same directory This way you can always recover from an OOM * Fix bug in reassembler: from_node and to_node were in the wrong order * Fix correctness mistake: The edges have to be reordered according to the node order. That's the whole reason this is tricky. But i'm not sure now whether the SoAs approach is actually an optimization.... It seems like we should probably prefer to inline the Edges right into the vector, rather than having to do another random lookup into the edges table? * Debugging messed up edge array idxs * Disable log message * Write the .nodes and .edges as binary data * Remove unnecessary logging * fix merge issues * attempt to add back the orphan node checking logic ``` --------- Co-authored-by: Nathan Daly <nathan.daly@relational.ai> Co-authored-by: Nathan Daly <NHDaly@gmail.com>
46 tasks
KristofferC
added a commit
that referenced
this pull request
Aug 13, 2024
Backported PRs: - [x] #51351 <!-- Remove boxing in pinv --> - [x] #52678 <!-- Profile: Improve module docstring --> - [x] #54201 <!-- Fix generic triangular solves with empty matrices --> - [x] #54605 <!-- Allow libquadmath to also fail as it is not available on all systems --> - [x] #54634 <!-- Fix trampoline assembly for build on clang 18 on apple silicon --> - [x] #54635 <!-- Aggressive constprop in trevc! to stabilize triangular eigvec --> - [x] #54645 <!-- ensure we set the right value to gc_first_tid --> - [x] #54671 <!-- Add boundscheck in bindingkey_eq to avoid OOB access due to data race --> - [x] #54672 <!-- make: Fix `sed` command for LLVM libraries with no symbol versioning --> - [x] #54704 <!-- LazyString in reinterpretarray error messages --> - [x] #54713 <!-- make: use `readelf` for LLVM symbol version detection --> - [x] #54781 <!-- [LinearAlgebra] Improve resilience to unknown libblastrampoline flags --> - [x] #54837 <!-- Do not add type tag size to the `alloc_typed` lowering for GC allocations --> - [x] #54815 <!-- add sticky task warning to `@task` and `schedule` --> - [x] #55141 <!-- Update the aarch64 devdocs to reflect the current state of its support --> - [x] #55178 <!-- Compat for `Base.@nospecializeinfer` --> - [x] #55197 <!-- compat notice for a[begin] indexing --> - [x] #55209 <!-- correction to compat notice for a[begin] --> - [x] #55203 <!-- document mutable struct const fields --> - [x] #54769 <!-- add missing compat entry to edit --> - [x] #54791 <!-- Bump libblastrampoline to v5.10.1 --> - [x] #55070 <!-- LinearAlgebra: LazyString in error messages for Diagonal/Bidiagonal --> - [x] #54624 <!-- more precise aliasing checks for SubArray --> - [x] #54690 <!-- Fix assertion/crash when optimizing function with dead basic block --> - [x] #55084 <!-- Use triple quotes in TOML.print when string contains newline --> Need manual backport: - [ ] #52505 <!-- fix alignment of emit_unbox_store copy --> - [ ] #53373 <!-- fix sysimage-native-code=no option with pkgimages --> - [ ] #53984 <!-- Profile: fix heap snapshot is valid char check --> - [ ] #54276 <!-- Fix solve for complex `Hermitian` with non-vanishing imaginary part on diagonal --> - [ ] #54669 <!-- Improve error message in inplace transpose --> - [ ] #54871 <!-- Make warn missed transformations pass optional --> Contains multiple commits, manual intervention needed: - [ ] #52854 <!-- Change to streaming out the heap snapshot data --> - [ ] #53218 <!-- Fix interpreter_exec.jl test --> - [ ] #53833 <!-- Profile: make heap snapshots viewable in vscode viewer --> - [ ] #54303 <!-- LinearAlgebra: improve type-inference in Symmetric/Hermitian matmul --> - [ ] #52694 <!-- Reinstate similar for AbstractQ for backward compatibility --> - [ ] #54737 <!-- LazyString in interpolated error messages involving types --> - [ ] #54738 <!-- serialization: fix relocatability bug --> - [ ] #55052 <!-- Fix `(l/r)mul!` with `Diagonal`/`Bidiagonal` --> Non-merged PRs with backport label: - [ ] #55220 <!-- `isfile_casesensitive` fixes on Windows --> - [ ] #55169 <!-- `propertynames` for SVD respects private argument --> - [ ] #55013 <!-- [docs] change docstring to match code --> - [ ] #51479 <!-- prevent code loading from lookin in the versioned environment when building Julia --> - [ ] #50813 <!-- More doctests for Sockets and capitalization fix --> - [ ] #50157 <!-- improve docs for `@inbounds` and `Base.@propagate_inbounds` --> - [ ] #41244 <!-- Fix shell `cd` error when working dir has been deleted -->
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR is to continue the work on the following PR:
Prevent OOMs during heap snapshot: Change to streaming out the snapshot data (#51518 )
Here are the commit history: