3547 arrays_hashcode_elvload(vtmp[idx], Address(ary1, index, Address::times(elsize), 8 * idx * elsize), eltype);
3548 }
3549 // vresult = vresult * vnext + ary1[index+8*idx:index+8*idx+7];
3550 for (int idx = 0; idx < 4; idx++) {
3551 vpmulld(vresult[idx], vresult[idx], vnext, Assembler::AVX_256bit);
3552 arrays_hashcode_elvcast(vtmp[idx], eltype);
3553 vpaddd(vresult[idx], vresult[idx], vtmp[idx], Assembler::AVX_256bit);
3554 }
3555 // index += 32;
3556 addl(index, 32);
3557 // index < bound;
3558 cmpl(index, bound);
3559 jcc(Assembler::less, UNROLLED_VECTOR_LOOP_BEGIN);
3560 // }
3561
3562 lea(ary1, Address(ary1, bound, Address::times(elsize)));
3563 subl(cnt1, bound);
3564 // release bound
3565
3566 // vresult *= IntVector.fromArray(I256, power_of_31_backwards, 1);
3567 for (int idx = 0; idx < 4; idx++) {
3568 lea(tmp2, ExternalAddress(StubRoutines::x86::arrays_hashcode_powers_of_31() + ((8 * idx + 1) * sizeof(jint))));
3569 arrays_hashcode_elvload(vcoef[idx], Address(tmp2, 0), T_INT);
3570 vpmulld(vresult[idx], vresult[idx], vcoef[idx], Assembler::AVX_256bit);
3571 }
3572 // result += vresult.reduceLanes(ADD);
3573 for (int idx = 0; idx < 4; idx++) {
3574 reduceI(Op_AddReductionVI, 256/(sizeof(jint) * 8), result, result, vresult[idx], vtmp[(idx * 2 + 0) % 4], vtmp[(idx * 2 + 1) % 4]);
3575 }
3576
3577 // } else if (cnt1 < 32) {
3578
3579 bind(SHORT_UNROLLED_BEGIN);
3580 // int i = 1;
3581 movl(index, 1);
3582 cmpl(index, cnt1);
3583 jcc(Assembler::greaterEqual, SHORT_UNROLLED_LOOP_EXIT);
3584
3585 // for (; i < cnt1 ; i += 2) {
3586 bind(SHORT_UNROLLED_LOOP_BEGIN);
3587 movl(tmp3, 961);
3588 imull(result, tmp3);
3589 arrays_hashcode_elload(tmp2, Address(ary1, index, Address::times(elsize), -elsize), eltype);
|
3547 arrays_hashcode_elvload(vtmp[idx], Address(ary1, index, Address::times(elsize), 8 * idx * elsize), eltype);
3548 }
3549 // vresult = vresult * vnext + ary1[index+8*idx:index+8*idx+7];
3550 for (int idx = 0; idx < 4; idx++) {
3551 vpmulld(vresult[idx], vresult[idx], vnext, Assembler::AVX_256bit);
3552 arrays_hashcode_elvcast(vtmp[idx], eltype);
3553 vpaddd(vresult[idx], vresult[idx], vtmp[idx], Assembler::AVX_256bit);
3554 }
3555 // index += 32;
3556 addl(index, 32);
3557 // index < bound;
3558 cmpl(index, bound);
3559 jcc(Assembler::less, UNROLLED_VECTOR_LOOP_BEGIN);
3560 // }
3561
3562 lea(ary1, Address(ary1, bound, Address::times(elsize)));
3563 subl(cnt1, bound);
3564 // release bound
3565
3566 // vresult *= IntVector.fromArray(I256, power_of_31_backwards, 1);
3567 lea(tmp2, ExternalAddress(StubRoutines::x86::arrays_hashcode_powers_of_31() + (0 * sizeof(jint))));
3568 for (int idx = 0; idx < 4; idx++) {
3569 arrays_hashcode_elvload(vcoef[idx], Address(tmp2, (int)((8 * idx + 1) * sizeof(jint))), T_INT);
3570 vpmulld(vresult[idx], vresult[idx], vcoef[idx], Assembler::AVX_256bit);
3571 }
3572 // result += vresult.reduceLanes(ADD);
3573 for (int idx = 0; idx < 4; idx++) {
3574 reduceI(Op_AddReductionVI, 256/(sizeof(jint) * 8), result, result, vresult[idx], vtmp[(idx * 2 + 0) % 4], vtmp[(idx * 2 + 1) % 4]);
3575 }
3576
3577 // } else if (cnt1 < 32) {
3578
3579 bind(SHORT_UNROLLED_BEGIN);
3580 // int i = 1;
3581 movl(index, 1);
3582 cmpl(index, cnt1);
3583 jcc(Assembler::greaterEqual, SHORT_UNROLLED_LOOP_EXIT);
3584
3585 // for (; i < cnt1 ; i += 2) {
3586 bind(SHORT_UNROLLED_LOOP_BEGIN);
3587 movl(tmp3, 961);
3588 imull(result, tmp3);
3589 arrays_hashcode_elload(tmp2, Address(ary1, index, Address::times(elsize), -elsize), eltype);
|