8191924: Adjust DelegatingClassLoader's metadata space sizing algorithm
Summary: Reduce memory usage for delegating class loader's metadata by adjusting sizing alogrithm
Reviewed-by: adinn, stuefe, coleenp
--- a/src/hotspot/share/memory/metaspace.cpp Wed Mar 14 14:09:33 2018 +0100
+++ b/src/hotspot/share/memory/metaspace.cpp Tue Mar 13 16:21:45 2018 -0400
@@ -1248,9 +1248,9 @@
// Maximum number of small chunks to allocate to a SpaceManager
static uint const _small_chunk_limit;
- // Maximum number of specialize chunks to allocate for anonymous
+ // Maximum number of specialize chunks to allocate for anonymous and delegating
// metadata space to a SpaceManager
- static uint const _anon_metadata_specialize_chunk_limit;
+ static uint const _anon_and_delegating_metadata_specialize_chunk_limit;
// Sum of all space in allocated chunks
size_t _allocated_blocks_words;
@@ -1418,7 +1418,7 @@
};
uint const SpaceManager::_small_chunk_limit = 4;
-uint const SpaceManager::_anon_metadata_specialize_chunk_limit = 4;
+uint const SpaceManager::_anon_and_delegating_metadata_specialize_chunk_limit = 4;
const char* SpaceManager::_expand_lock_name =
"SpaceManager chunk allocation lock";
@@ -3358,11 +3358,11 @@
// Anonymous metadata space is usually small, with majority within 1K - 2K range and
// rarely about 4K (64-bits JVM).
// Instead of jumping to SmallChunk after initial chunk exhausted, keeping allocation
- // from SpecializeChunk up to _anon_metadata_specialize_chunk_limit (4) reduces space waste
- // from 60+% to around 30%.
- if (_space_type == Metaspace::AnonymousMetaspaceType &&
+ // from SpecializeChunk up to _anon_or_delegating_metadata_specialize_chunk_limit (4)
+ // reduces space waste from 60+% to around 30%.
+ if ((_space_type == Metaspace::AnonymousMetaspaceType || _space_type == Metaspace::ReflectionMetaspaceType) &&
_mdtype == Metaspace::NonClassType &&
- sum_count_in_chunks_in_use(SpecializedIndex) < _anon_metadata_specialize_chunk_limit &&
+ sum_count_in_chunks_in_use(SpecializedIndex) < _anon_and_delegating_metadata_specialize_chunk_limit &&
word_size + Metachunk::overhead() <= SpecializedChunk) {
return SpecializedChunk;
}