Eliminate zero_static_objects_list()

Summary:
In a workload with a large amount of code, zero_static_objects_list()
takes a significant amount of time, and furthermore it is in the
single-threaded part of the GC.

This patch uses a slightly fiddly scheme for marking objects on the
static object lists, using a flag in the low 2 bits that flips between
two states to indicate whether an object has been visited during this
GC or not.  We also have to take into account objects that have not
been visited yet, which might appear at any time due to runtime linking.

Test Plan: validate

Reviewers: austin, bgamari, ezyang, rwbarton

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1076

1 files changed, 4 insertions, 2 deletions

diff --git a/rts/CheckUnload.c b/rts/CheckUnload.c
index 2c01113f63..34f976db4d 100644
--- a/rts/CheckUnload.c
+++ b/rts/CheckUnload.c
@@ -271,7 +271,8 @@ void checkUnload (StgClosure *static_objects)
 
   addrs = allocHashTable();
 
-  for (p = static_objects; p != END_OF_STATIC_LIST; p = link) {
+  for (p = static_objects; p != END_OF_STATIC_OBJECT_LIST; p = link) {
+      p = UNTAG_STATIC_LIST_PTR(p);
       checkAddress(addrs, p);
       info = get_itbl(p);
       link = *STATIC_LINK(info, p);
@@ -279,8 +280,9 @@ void checkUnload (StgClosure *static_objects)
 
   // CAFs on revertible_caf_list are not on static_objects
   for (p = (StgClosure*)revertible_caf_list;
-       p != END_OF_STATIC_LIST;
+       p != END_OF_CAF_LIST;
        p = ((StgIndStatic *)p)->static_link) {
+      p = UNTAG_STATIC_LIST_PTR(p);
       checkAddress(addrs, p);
   }