Update RealtimeAccessBarrier to use common apis for copy arrayCritical #20651
Conversation
LinHu2016
force-pushed
the
off-heap-incremental8
branch
from
a347839 to
028341a
Compare
| bool shouldCopy = false; | ||
| if((javaVM->runtimeFlags & J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) == J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) { | ||
| if ((vmThread->javaVM->runtimeFlags & J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) == J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) { |
There was a problem hiding this comment.
please put constant first
if (J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL == ((vmThread->javaVM->runtimeFlags & J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL))
| } | ||
| } | ||
| vmThread->jniCriticalCopyCount += 1; | ||
| arrayObject = (J9IndexableObject *)J9_JNI_UNWRAP_REFERENCE(array); |
There was a problem hiding this comment.
do we have this call already in line 452?
There was a problem hiding this comment.
in a general case, object might have moved while not having VM access yet, so we have to re-read the address after acquiring VM access
but in RT, object don't move, so this indeed is unnecessary
There was a problem hiding this comment.
ok, thank you for clarification
There was a problem hiding this comment.
GC could happens between, we need reload the reference.
There was a problem hiding this comment.
please see Aleks comment above - "in RT, object don't move, so this indeed is unnecessary"
| VM_VMAccess::inlineExitVMToJNI(vmThread); | ||
| } else { | ||
| // acquire access and return a direct pointer | ||
| MM_JNICriticalRegion::enterCriticalRegion(vmThread, false); | ||
| arrayObject = (J9IndexableObject *)J9_JNI_UNWRAP_REFERENCE(array); |
There was a problem hiding this comment.
the same, see line 452.
There was a problem hiding this comment.
GC could happens between, we need reload the reference.
| bool shouldCopy = false; | ||
| if((javaVM->runtimeFlags & J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) == J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) { | ||
| if ((vmThread->javaVM->runtimeFlags & J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) == J9_RUNTIME_ALWAYS_COPY_JNI_CRITICAL) { |
There was a problem hiding this comment.
const part first please
| Assert_MM_invalidJNICall(); | ||
| } | ||
|
|
||
| arrayObject = (J9IndexableObject *)J9_JNI_UNWRAP_REFERENCE(array); |
There was a problem hiding this comment.
duplicate, see line 481
LinHu2016
force-pushed
the
off-heap-incremental8
branch
3 times, most recently
from
dd82ae1 to
9fdac8d
Compare
|
Jenkins test sanity AIX jdk11 |
LinHu2016
force-pushed
the
off-heap-incremental8
branch
from
9fdac8d to
f162fb8
Compare
- Update RealtimeAccessBarrier to use opyArrayCritical, copyBackArrayCritical, copyStringCritical and freeStringCritical common functions (in ObjectAccessBarrier baseclass) to replace similar code in class RealtimeAccessBarrier. - output warning message if off-heap is enabled in gcpolicy:metronome. Signed-off-by: lhu <[email protected]>
LinHu2016
force-pushed
the
off-heap-incremental8
branch
from
f162fb8 to
7411dc0
Compare
|
Jenkins test sanity AIX jdk11 |
depends on : #20648