make objectid effects total #52119
Merged
make objectid effects total #52119
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
oscardssmith
added
the
needs nanosoldier run
JeffBezanson
approved these changes
Nov 10, 2023
|
Very nice. I think it's well worth it to have object_id be total and not have to worry about rehashing things, especially for types in other packages. |
vchuravy
approved these changes
Nov 10, 2023
|
@nanosoldier |
|
Your benchmark job has completed - no performance regressions were detected. A full report can be found here. |
vtjnash
removed
the
needs nanosoldier run
Avoids the need to rehash most dictionaries on reload. System image data size increase is about 109MB -> 112MB, since there are about 130k Arrays, 75k CodeInstances, 30k Methods, 80k TypeMapEntries, 75k MethodInstance, and 36k Core.Bindings, and other mutable objects.
vtjnash
force-pushed
the
jn/stable-objectid
branch
from
841ebc6
to
e9421ab
Compare
vchuravy
added a commit
that referenced
this pull request
Apr 13, 2024
For GPUCompiler we would like to support a native on disk cache of LLVM IR. One of the longstanding issues has been the cache invalidation of such an on disk cache. With #52233 we now have an integrated cache for the inference results and we can rely on `CodeInstance` to be stable across sessions. Due to #52119 we can also rely on the `objectid` to be stable. My inital thought was to key the native disk cache in GPUCompiler on the objectid of the corresponding CodeInstance (+ some compilation parameters). While discussing this with @rayegun yesterday we noted that having a CodeInstance with the same objectid might not be enough provenance. E.g we are not gurantueed that the CodeInstance is from the same build artifact and the same precise source code. For the package images we are tracking this during loading and validate all contents at once, and we keep explicitly track of the provenance chain. This PR adds a lookup up table where we map from "external_blobs" e.g. loaded images, to the corresponding top module of each image, and uses this to determine the build_id of the package image.
vchuravy
added a commit
that referenced
this pull request
Apr 19, 2024
For GPUCompiler we would like to support a native on disk cache of LLVM IR. One of the longstanding issues has been the cache invalidation of such an on disk cache. With #52233 we now have an integrated cache for the inference results and we can rely on `CodeInstance` to be stable across sessions. Due to #52119 we can also rely on the `objectid` to be stable. My inital thought was to key the native disk cache in GPUCompiler on the objectid of the corresponding CodeInstance (+ some compilation parameters). While discussing this with @rayegun yesterday we noted that having a CodeInstance with the same objectid might not be enough provenance. E.g we are not gurantueed that the CodeInstance is from the same build artifact and the same precise source code. For the package images we are tracking this during loading and validate all contents at once, and we keep explicitly track of the provenance chain. This PR adds a lookup up table where we map from "external_blobs" e.g. loaded images, to the corresponding top module of each image, and uses this to determine the build_id of the package image. (cherry picked from commit d47cbf6)
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.
Avoids the need to rehash most dictionaries on reload.
System image data size increase is about 109MB -> 112MB, since there are about 130k Arrays, 75k CodeInstances, 30k Methods, 80k TypeMapEntries, 75k MethodInstance, and 36k Core.Bindings, and other mutable objects.