RFR 8218142 [lworld] C1 support for array covariance for aaload

Frederic Parain frederic.parain at oracle.com
Thu Jan 31 15:15:15 UTC 2019


Ioi,

I haven't completed the review yet, but I have a question about this
method:

  150 bool Instruction::maybe_flattened_array() const {
  151   if (ValueArrayFlatten) {
  152     ciType* type = declared_type();
  153     if (type != NULL) {
  154       if (type->is_value_array_klass()) {
  155         ciValueKlass* element_klass = 
type->as_value_array_klass()->element_klass()->as_value_klass();
  156         if (!element_klass->is_loaded() || 
element_klass->flatten_array()) {
  157           // For unloaded value arrays, we will add a runtime 
check for flat-ness.
  158           return true;
  159         }
  160       } else if (type->is_obj_array_klass()) {
  161         ciKlass* element_klass = 
type->as_obj_array_klass()->element_klass();
  162         if (element_klass->is_java_lang_Object()) {
  163           // Array covariance (ValueType[] <: Object[])
  164           // We will add a runtime check for flat-ness.
  165           return true;
  166         }
  167       }
  168     }
  169   }
  170
  171   return false;
  172 }

I think this method should also return true when the type
of the element is an interface. Am I missing something?

Fred


On 1/31/19 7:52 AM, Ioi Lam wrote:
> http://cr.openjdk.java.net/~iklam/valhalla/8218142-aaload-covariance.v01/
> https://bugs.openjdk.java.net/browse/JDK-8218142
> 
> When loading from an array declared as Object[], now the C1 code checks 
> if the array is flattened. If so, the loading is done via a runtime call.
> 
> Here's an example of the generated code
> 
>        mov    0x8(%rsi),%eax      ; rsi = array
>        mov    $0x800000000,%r12
>        add    %r12,%rax           ; rax = array klass
>        xor    %r12,%r12           ; why is this here??
>        mov    0xc(%rax),%eax      ; layout_helper
>        sar    $0x1d,%eax
>        cmp    $0xfffffffd,%eax    ; check for Klass::_lh_array_tag_vt_value
>        je      L_3                ; slow case
> 
> L_1   movslq %edx,%rdx           ; non-flattened load
>        mov    0x10(%rsi,%rdx,4),%eax
>        shl    $0x3,%rax
> L_2   ....
> 
> L_3   mov    %rsi,0x8(%rsp)      ; LoadFlattenedArrayStub slow case
>        mov    %rdx,(%rsp)
>        callq  0x00007fe6489c8840  ; runtime_call load_flattened_array
>        jmpq    L_2
> 
> 
> The code is pretty big and sub-optimal. My plan is to also get aastore 
> working first, and then optimize.
> 
> One possibility is to check if the array is actually an Object[] array. 
> That should help with the erased generic types.
> 
>        mov    0x8(%rsi),%eax      ; rsi = array
> +     cmp    $0x00001234, %eax   ; compressed klass for [Ljava/lang/Object;
> +     je     L_1                 ; go to non-flattened load
>        mov    $0x800000000,%r12
> 
> 
> Thanks
> - Ioi


More information about the valhalla-dev mailing list