-
Bug
-
Resolution: Fixed
-
P2
-
18
-
b29
-
x86
Issue | Fix Version | Assignee | Priority | Status | Resolution | Resolved In Build |
---|---|---|---|---|---|---|
JDK-8279170 | 19 | Jatin Bhateja | P2 | Resolved | Fixed | b04 |
JDK-8279751 | 18.0.1 | Jatin Bhateja | P2 | Resolved | Fixed | b02 |
--N: o284 XorVMask === _ o310 o282 [[o311 ]] #vectormask[16]:{byte}
--N: o310 VectorLoadMask === _ o309 [[o284 ]] #vectormask[16]:{byte}
KREG 0 KREG
KREG_K1 0 KREG_K1
KREG_K2 0 KREG_K2
KREG_K3 0 KREG_K3
KREG_K4 0 KREG_K4
KREG_K5 0 KREG_K5
KREG_K6 0 KREG_K6
KREG_K7 0 KREG_K7
--N: o282 VectorMaskCast === _ o261 [[o284 ]] #vectormask[16]:{byte}
VEC 0 vmaskcast
LEGVEC 100 MoveVec2Leg
--N: o261 ReplicateB === _ o260 [[o312 o282 ]] #vectorx[16]:{byte}
RXMM0 0 RXMM0
VEC 0 VEC
LEGVEC 0 LEGVEC
VECX 0 VECX
LEGVECX 0 LEGVECX
X86 backend implementation does not enable maskAll instruction patterns for 32 bit JVM, due to which operations fall backs over replicateB operation which broadcasts the mask value in a vector and after unboxing-boxing optimization this vector eventually reaches to XorVMask which cribs since one operand is held in opmask register and other in vector.
- backported by
-
JDK-8279170 Enable X86 maskAll instruction pattern for 32 bit JVM.
-
- Resolved
-
-
JDK-8279751 Enable X86 maskAll instruction pattern for 32 bit JVM.
-
- Resolved
-