[DO NOT MERGE] Generate reports #3
Closed
Conversation
This file contains hidden or 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
Member
|
Version comparison is extremely valuable, well done! The numbers look really srange though, 3-5x degradation from version to version looks like either debug mode or boehm gc. |
Contributor
Author
|
I will double-check that I do not have any debugging enabled by accident and re-run it tonight. |
valdisxp1
force-pushed
the
compare-3.8.5-with-master
branch
from
2b8cd6a to
6800fd5
Compare
Member
|
@valdisxp1 Indeed, master regressed wrt 0.3.8. I did manual bisect based first few benchmarks and found that it regressed in scala-native/scala-native#1295. The issue is that we used to do DCE after lowering and before codegen and that PR mistakenly removed it. |
valdisxp1
force-pushed
the
compare-3.8.5-with-master
branch
from
5ce9d98 to
7bcaa51
Compare
Contributor
Author
|
Including more changes #4 and automation stuff.
scripts/run.py latest@master latest@my-branch
scripts/summary latest@master latest@my-branch |
…master # Conflicts: # scripts/summary.py
Contributor
Author
densh
pushed a commit
to scala-native/scala-native
that referenced
this pull request
Jul 17, 2019
Features: - New `SCALANATIVE_GC` option `commix` to enable the new GC changes. - Concurrent, parallel sweep. Reduced GC pause times by utilizing additional processor cores during sweeping. - Parallel mark. Reduces GC pause times by utilizing additional processor cores during marking. - Configurable number of GC threads with `SCALANATIVE_GC_THREADS` environment variable. Set to number of available processors - 1 up to 8, but no less than 1. - Increasing heap size until mark is <=5% of total time. This can be adjusted with `SCALANATIVE_TIME_RATIO`. I.e. `export SCALANATIVE_TIME_RATIO=0.03` for <= 3%. - Added an option the change the minimum free block ratio `SCALANATIVE_FREE_RATIO`. If there are not enough free blocks after a garbage collection it will increase the heap size. - The minimum heap size setting renamed from `SCALANATIVE_MIN_HEAP_SIZE` to a shorter `SCALANATIVE_MIN_SIZE`. Updated Immix to match. - The maxium heap size setting renamed from `SCALANATIVE_MAX_HEAP_SIZE` to a shorter `SCALANATIVE_MAX_SIZE`. Updated Immix to match. ### Changes to GC Statistics - Stat collection disabled by default to reduce code size. They can be enabled by compiler arguments. See example and table below. - Stats changed to event-based CSV format. - Stats now also saves events for collection start and end. - Stats includes `start_ns` column - time in nanoseconds since application was started. - Stats is compatible with the new concurrent sweep events. Multiple verbosity levels of enabled for stats with compiler arguments. For example: ```scala nativeCompileOptions += "-DENABLE_GC_STATS_BATCHES" ``` > scala-native-benchmarks flags as per suggested changes in scala-native/scala-native-benchmarks#3 > results size for default tests when doing a single run `--runs 1` instead of the default which is 20. Internals: - thread-safe `BlockList`s - thread-safe `BlockRange`s - thread-safe block counts - thread-safe `GreyList`s - new block flags in `BlockMeta` - removed marking `Stack`s Debugging : - asserts in Immix can now be enabled by setting `DEBUG_ASSERT` flag ```scala nativeCompileOptions += "-DDEBUG_ASSERT" ``` This similar to the way you could enable trace level printing and debug symbols. ```scala // trace level printing in Immix nativeCompileOptions += "-DDEBUG_PRINT" // enable debug symbols for C code nativeCompileOptions += "-g" ``` - More integrity check asserts for sweeping. ## Refactoring: - Moved `Block_Sweep` to `Allocator_Sweep` in order to match `LargeAllocator_Sweep`. - Moved `Heap_AllocSmall` and `HeapAllocLarge` to their respective allocators - Renamed `block_superblock_middle` to ` block_superblock_tail` in order to avoid confusion. This block flag is used for both middle and last blocks
ekrich
pushed a commit
to ekrich/scala-native
that referenced
this pull request
May 21, 2021
Features: - New `SCALANATIVE_GC` option `commix` to enable the new GC changes. - Concurrent, parallel sweep. Reduced GC pause times by utilizing additional processor cores during sweeping. - Parallel mark. Reduces GC pause times by utilizing additional processor cores during marking. - Configurable number of GC threads with `SCALANATIVE_GC_THREADS` environment variable. Set to number of available processors - 1 up to 8, but no less than 1. - Increasing heap size until mark is <=5% of total time. This can be adjusted with `SCALANATIVE_TIME_RATIO`. I.e. `export SCALANATIVE_TIME_RATIO=0.03` for <= 3%. - Added an option the change the minimum free block ratio `SCALANATIVE_FREE_RATIO`. If there are not enough free blocks after a garbage collection it will increase the heap size. - The minimum heap size setting renamed from `SCALANATIVE_MIN_HEAP_SIZE` to a shorter `SCALANATIVE_MIN_SIZE`. Updated Immix to match. - The maxium heap size setting renamed from `SCALANATIVE_MAX_HEAP_SIZE` to a shorter `SCALANATIVE_MAX_SIZE`. Updated Immix to match. ### Changes to GC Statistics - Stat collection disabled by default to reduce code size. They can be enabled by compiler arguments. See example and table below. - Stats changed to event-based CSV format. - Stats now also saves events for collection start and end. - Stats includes `start_ns` column - time in nanoseconds since application was started. - Stats is compatible with the new concurrent sweep events. Multiple verbosity levels of enabled for stats with compiler arguments. For example: ```scala nativeCompileOptions += "-DENABLE_GC_STATS_BATCHES" ``` > scala-native-benchmarks flags as per suggested changes in scala-native/scala-native-benchmarks#3 > results size for default tests when doing a single run `--runs 1` instead of the default which is 20. Internals: - thread-safe `BlockList`s - thread-safe `BlockRange`s - thread-safe block counts - thread-safe `GreyList`s - new block flags in `BlockMeta` - removed marking `Stack`s Debugging : - asserts in Immix can now be enabled by setting `DEBUG_ASSERT` flag ```scala nativeCompileOptions += "-DDEBUG_ASSERT" ``` This similar to the way you could enable trace level printing and debug symbols. ```scala // trace level printing in Immix nativeCompileOptions += "-DDEBUG_PRINT" // enable debug symbols for C code nativeCompileOptions += "-g" ``` - More integrity check asserts for sweeping. ## Refactoring: - Moved `Block_Sweep` to `Allocator_Sweep` in order to match `LargeAllocator_Sweep`. - Moved `Heap_AllocSmall` and `HeapAllocLarge` to their respective allocators - Renamed `block_superblock_middle` to ` block_superblock_tail` in order to avoid confusion. This block flag is used for both middle and last blocks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
run.pyandsummary.py. This makes testing the performance of new features easier (see Readme.md).See example report:
https://github.com/valdisxp1/scala-native-benchmark-results/blob/gh-pages/reports/summary_20180923_222231_scala-native-0.3.8_vs_scala-native-0.3.9-SNAPSHOT/Readme.md