[MonoVM] Reduce P/Invoke GC transition asserts in release builds #59029
Conversation
ghost
added
area-VM-meta-mono
community-contribution
|
Tagging subscribers to this area: Issue DetailsProfiling shows that large part of the GC transition overhead (~30%) in #58939 is caused by assert-style checks. Disabling them seems to be the best bang-for-the-buck option for reducing the overhead without fundamental changes to the code.
|
| static void | ||
| check_thread_state (MonoThreadInfo* info) | ||
| { | ||
| #ifdef ENABLE_CHECKED_BUILD_THREAD |
There was a problem hiding this comment.
@filipnavara I think this one should be ENABLE_CHECKED_BUILD_GC - I think other thread state machine stuff in "GC", too. (for example assert_gc_safe_mode).
I think we should also turn on the GC checked build in debug builds:
https://github.com/dotnet/runtime/blob/main/src/mono/CMakeLists.txt#L694-L695
There was a problem hiding this comment.
Hmm. But to be honest I can't tell the difference - looks like we use both defines for closely related things.
There was a problem hiding this comment.
I mostly tried to match the asserts in the same file but I can change it. I will update the PR to enable it on debug builds.
There was a problem hiding this comment.
After checking how the configuration options are set I came to the conclusion to change it to the general ENABLE_CHECKED_BUILD. That one is defined in all debug builds, the other ones can be selectively enabled.
|
On Mac Catalyst ARM64 the test case in #58939 goes from |
|
will you consider to backport to release/6.0? |
|
/backport to release/6.0-rc2 |
|
Started backporting to release/6.0-rc2: https://github.com/dotnet/runtime/actions/runs/1246192307 |
…elease builds (#59269) Backport of #59029 Profiling shows that large part of the GC transition overhead (~30%) in #58939 is caused by assert-style checks. Disabling them seems to be the best bang-for-the-buck option for reducing the overhead without fundamental changes to the code. Co-authored-by: Filip Navara <[email protected]>
ghost
locked as resolved and limited conversation to collaborators
5 participants
Profiling shows that large part of the GC transition overhead (~30%) in #58939 is caused by assert-style checks. Disabling them seems to be the best bang-for-the-buck option for reducing the overhead without fundamental changes to the code.