RFR #2 (S) CR 8015270: @Contended: fix multiple issues in the layout code

Coleen Phillimore coleen.phillimore at oracle.com
Sat May 25 07:13:04 PDT 2013


This code still looks fine.   One comment, not a suggested change or 
request to see this again.

+  // TODO: We add +1 to always allocate non-zero resource arrays; we need
+  // to figure out if we still need to do this.
    int* nonstatic_oop_offsets;
    unsigned int* nonstatic_oop_counts;
    unsigned int nonstatic_oop_map_count = 0;
+  unsigned int max_nonstatic_oop_maps  = fac->count[NONSTATIC_OOP] + 1;
  
    nonstatic_oop_offsets = NEW_RESOURCE_ARRAY_IN_THREAD(
-            THREAD, int, nonstatic_oop_count + 1);
+            THREAD, int, max_nonstatic_oop_maps);
    nonstatic_oop_counts  = NEW_RESOURCE_ARRAY_IN_THREAD(
-            THREAD, unsigned int, nonstatic_oop_count + 1);
+            THREAD, unsigned int, max_nonstatic_oop_maps);
  

I don't think there's anything to do here.  The code makes it clear 
because we allocate these arrays unconditionally, even if the count of 
nonstatic oops are zero.   NEW_RESOURCE_IN_ARRAY probably doesn't like that.

Coleen

On 5/25/2013 3:44 AM, Aleksey Shipilev wrote:
> On 05/25/2013 11:35 AM, Aleksey Shipilev wrote:
>> Hi,
>>
>> Here's the merged changeset:
>>    http://cr.openjdk.java.net/~shade/8015270/webrev.01/
>>
>> Summary:
>>   - has_nonstatic_fields bug fixed
>>   - oop_map overrun is fixed
>>   - parsed_annotations->in_contended() cleaned up
>>   - asserts added to cover the regression tests
>>
>> Testing:
>>   - all HS regression tests
>>       (have a few environmental failures, but nothing new)
>>   - full JPRT cycle against hotspot-rt
>>       (passes OK)
>>   - vm.quick.testlist (Solaris, Linux, Mac OS X, Windows)
>>       (no new failures)
>>   - vm.runtime.testlist (Solaris, Linux, Windows)
>>       (no new failures)
> Oh, I forgot the mention this. Since vm.quick had successfully run on
> Windows... Coleen, do you still want me to do runThese there?
>
> -Aleksey.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/attachments/20130525/8860abce/attachment.html 


More information about the hotspot-runtime-dev mailing list