[GitHub] [spark] attilapiros commented on a change in pull request #34846: [SPARK-37593][CORE] Optimize HeapMemoryAllocator to avoid memory waste in humongous allocation when using G1GC

[GitBox <gi...@apache.org>]

attilapiros commented on a change in pull request #34846:
URL: https://github.com/apache/spark/pull/34846#discussion_r769417875



##########
File path: core/src/main/scala/org/apache/spark/util/Utils.scala
##########
@@ -3204,6 +3204,14 @@ private[spark] object Utils extends Logging {
     }
     files.toSeq
   }
+
+  val isG1GarbageCollector: Boolean = {
+    ManagementFactory.getGarbageCollectorMXBeans

Review comment:
       When I looked for how to find out what the garbage collector type is I bumped into this several times:
   
   ```java
     HotSpotDiagnosticMXBean diagnostic = ManagementFactoryHelper.getDiagnosticMXBean();
   
       VMOption option = diagnostic.getVMOption("UseG1GC");
       if (option.getValue().equals("false")) {
         ...
       }
   ```
   For example at the [OpenJDK tests](https://www.programcreek.com/java-api-examples/?code=AdoptOpenJDK%2Fopenjdk-jdk8u%2Fopenjdk-jdk8u-master%2Fhotspot%2Ftest%2Fgc%2Farguments%2FTestG1HeapRegionSize.java
   ). 
   
   Is there any reason why a different solution have been chosen here?
   

##########
File path: core/src/main/scala/org/apache/spark/memory/MemoryManager.scala
##########
@@ -255,7 +256,17 @@ private[spark] abstract class MemoryManager(
     }
     val size = ByteArrayMethods.nextPowerOf2(maxTungstenMemory / cores / safetyFactor)
     val default = math.min(maxPageSize, math.max(minPageSize, size))
-    conf.get(BUFFER_PAGESIZE).getOrElse(default)
+    val sizeAsBytes = conf.get(BUFFER_PAGESIZE).getOrElse(default)
+    // If we are using G1 GC, it's better to take the LONG_ARRAY_OFFSET into consideration
+    // so that the requested memory size is power of 2 and can be divided by G1 region size
+    // to reduce memory waste within one G1 region
+    if (Utils.isG1GarbageCollector &&
+      tungstenMemoryMode == MemoryMode.ON_HEAP &&
+      sizeAsBytes % (1024 * 1024) == 0) {
+      sizeAsBytes - Platform.LONG_ARRAY_OFFSET

Review comment:
       If I get this right in case of G1GC the best would be if we choose a pageSize where the following holds:
   
   ```
     G1HeapRegionSize % (pageSize + Platform.LONG_ARRAY_OFFSET) == 0
   ```
   And when When BUFFER_PAGESIZE is not set we are free to choose it as:
   ```
   pageSize = G1HeapRegionSize - Platform.LONG_ARRAY_OFFSET;
   ```
   
   And with the above code we just try to calculate G1HeapRegionSize with our own way. But what about accessing this value?
   Like the same way used in the [OpenJDK tests](https://www.programcreek.com/java-api-examples/?code=AdoptOpenJDK%2Fopenjdk-jdk8u%2Fopenjdk-jdk8u-master%2Fhotspot%2Ftest%2Fgc%2Farguments%2FTestG1HeapRegionSize.java):
   
   ```java
     HotSpotDiagnosticMXBean diagnostic = ManagementFactoryHelper.getDiagnosticMXBean();
     option = diagnostic.getVMOption("G1HeapRegionSize");
   ```
   
   As I see the OpenJDK test was executed like:
   ```
   run main/othervm -Xmx64m TestG1HeapRegionSize 1048576
   ```
   So `diagnostic.getVMOption("G1HeapRegionSize")` gives back the calculated region size.
   
   




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org