Fetch before trying to atomically mark in Compressor (#1438)

Tracing is exceedingly slow on biojava if we don't
test-and-test-and-set. My guess is that we get cache line ping-pong when
many workers try to mark the same popular objects; but every policy I
checked has some sort of test-and-test-and-set pattern, and this seems
like a common sense thing to do.

[Plotty](http://squirrel.anu.edu.au/plotty/hayleyp/plots/p/GkUHfS) for a
detailed before-and-after, although both configurations have some other
changes to Compressor which I will upstream later. But this change still
speeds up upstream - here is before:

```
============================ MMTk Statistics Totals ============================
GC	time.other	time.stw	total-work.time.min	total-work.time.total	total-work.time.max	total-work.count	
14	8575.41	14920.23	inf	0.000	-inf	0	
Total time: 23495.63 ms
------------------------------ End MMTk Statistics -----------------------------
```

And after:

```
============================ MMTk Statistics Totals ============================
GC	time.other	time.stw	total-work.time.total	total-work.time.max	total-work.time.min	total-work.count	
14	8826.60	9548.96	0.000	-inf	inf	0	
Total time: 18375.56 ms
------------------------------ End MMTk Statistics -----------------------------
```

Author: no-defun-allowed

src/policy/compressor/compressorspace.rs

@@ -48,8 +48,6 @@ pub struct CompressorSpace<VM: VMBinding> {
     scheduler: Arc<GCWorkScheduler<VM>>,
 }
 
-pub(crate) const GC_MARK_BIT_MASK: u8 = 1;
-
 impl<VM: VMBinding> SFT for CompressorSpace<VM> {
     fn name(&self) -> &'static str {
         self.get_name()
@@ -271,19 +269,26 @@ impl<VM: VMBinding> CompressorSpace<VM> {
     }
 
     pub fn test_and_mark(object: ObjectReference) -> bool {
-        let old = forwarding::MARK_SPEC.fetch_or_atomic(
-            object.to_raw_address(),
-            GC_MARK_BIT_MASK,
-            Ordering::SeqCst,
-        );
-        (old & GC_MARK_BIT_MASK) == 0
+        forwarding::MARK_SPEC
+            .fetch_update_atomic::<u8, _>(
+                object.to_raw_address(),
+                Ordering::SeqCst,
+                Ordering::Relaxed,
+                |v| {
+                    if v == 0 {
+                        Some(1)
+                    } else {
+                        None
+                    }
+                },
+            )
+            .is_ok()
     }
 
     pub fn is_marked(object: ObjectReference) -> bool {
-        let old_value =
+        let mark_bit =
             forwarding::MARK_SPEC.load_atomic::<u8>(object.to_raw_address(), Ordering::SeqCst);
-        let mark_bit = old_value & GC_MARK_BIT_MASK;
-        mark_bit != 0
+        mark_bit == 1
     }
 
     fn generate_tasks(

src/policy/compressor/forwarding.rs

@@ -1,4 +1,3 @@
-use crate::policy::compressor::GC_MARK_BIT_MASK;
 use crate::util::constants::BYTES_IN_WORD;
 use crate::util::linear_scan::{Region, RegionIterator};
 use crate::util::metadata::side_metadata::spec_defs::{COMPRESSOR_MARK, COMPRESSOR_OFFSET_VECTOR};
@@ -144,7 +143,9 @@ impl<VM: VMBinding> ForwardingMetadata<VM> {
             );
         }
 
-        MARK_SPEC.fetch_or_atomic(last_word_of_object, GC_MARK_BIT_MASK, Ordering::SeqCst);
+        // We only mark the last word as input to computing forwarding
+        // information, so relaxed consistency is okay.
+        MARK_SPEC.fetch_or_atomic::<u8>(last_word_of_object, 1, Ordering::Relaxed);
     }
 
     pub fn calculate_offset_vector(&self, region: CompressorRegion, cursor: Address) {