src/hotspot/cpu/x86/x86_64.ad
Index
Unified diffs
Context diffs
Sdiffs
Wdiffs
Patch
New
Old
Previous File
Next File
*** old/src/hotspot/cpu/x86/x86_64.ad Wed Jul 1 16:48:41 2020
--- new/src/hotspot/cpu/x86/x86_64.ad Wed Jul 1 16:48:41 2020
*** 1643,1657 ****
--- 1643,1654 ----
// Prefer ConN+DecodeN over ConP.
return true;
}
bool Matcher::const_klass_prefer_decode() {
// TODO: Either support matching DecodeNKlass (heap-based) in operand
// or condisider the following:
// Prefer ConNKlass+DecodeNKlass over ConP in simple compressed klass mode.
//return CompressedKlassPointers::base() == NULL;
return true;
+ // Prefer ConP over ConNKlass+DecodeNKlass.
+ return false;
}
// Is it better to copy float constants, or load them directly from
// memory? Intel can load a float constant from a direct address,
// requiring no extra registers. Most RISCs will have to materialize
src/hotspot/cpu/x86/x86_64.ad
Index
Unified diffs
Context diffs
Sdiffs
Wdiffs
Patch
New
Old
Previous File
Next File