Merge pull request #3115 from lioncash/sqxtun

Arm64/VectorOps: Elide moves where applicable in 128-bit VSQXTUN2
This commit is contained in:
Ryan Houdek authored and GitHub committed 2023-09-17 14:51:27 -07:00
commit fd1b639ad9
1 file changed
+11 -3
@@ -3419,9 +3419,17 @@ DEF_OP(VSQXTUN2) {
mov(Dst.Q(), VectorLower.Q()); mov(Dst.Q(), VectorLower.Q());
ins(ARMEmitter::SubRegSize::i32Bit, Dst, 1, VTMP2, 0); ins(ARMEmitter::SubRegSize::i32Bit, Dst, 1, VTMP2, 0);
} else { } else {
mov(VTMP1.Q(), VectorLower.Q()); auto Lower = VectorLower;
sqxtun2(SubRegSize, VTMP1, VectorUpper); if (Dst != VectorLower) {
mov(Dst.Q(), VTMP1.Q()); mov(VTMP1.Q(), VectorLower.Q());
Lower = VTMP1;
}
sqxtun2(SubRegSize, Lower, VectorUpper);
if (Dst != VectorLower) {
mov(Dst.Q(), Lower.Q());
}
} }
} }
} }