3754 arrays_hashcode_elvload(vtmp[idx], Address(ary1, index, Address::times(elsize), 8 * idx * elsize), eltype);
3755 }
3756 // vresult = vresult * vnext + ary1[index+8*idx:index+8*idx+7];
3757 for (int idx = 0; idx < 4; idx++) {
3758 vpmulld(vresult[idx], vresult[idx], vnext, Assembler::AVX_256bit);
3759 arrays_hashcode_elvcast(vtmp[idx], eltype);
3760 vpaddd(vresult[idx], vresult[idx], vtmp[idx], Assembler::AVX_256bit);
3761 }
3762 // index += 32;
3763 addl(index, 32);
3764 // index < bound;
3765 cmpl(index, bound);
3766 jcc(Assembler::less, UNROLLED_VECTOR_LOOP_BEGIN);
3767 // }
3768
3769 lea(ary1, Address(ary1, bound, Address::times(elsize)));
3770 subl(cnt1, bound);
3771 // release bound
3772
3773 // vresult *= IntVector.fromArray(I256, power_of_31_backwards, 1);
3774 for (int idx = 0; idx < 4; idx++) {
3775 lea(tmp2, ExternalAddress(StubRoutines::x86::arrays_hashcode_powers_of_31() + ((8 * idx + 1) * sizeof(jint))));
3776 arrays_hashcode_elvload(vcoef[idx], Address(tmp2, 0), T_INT);
3777 vpmulld(vresult[idx], vresult[idx], vcoef[idx], Assembler::AVX_256bit);
3778 }
3779 // result += vresult.reduceLanes(ADD);
3780 for (int idx = 0; idx < 4; idx++) {
3781 reduceI(Op_AddReductionVI, 256/(sizeof(jint) * 8), result, result, vresult[idx], vtmp[(idx * 2 + 0) % 4], vtmp[(idx * 2 + 1) % 4]);
3782 }
3783
3784 // } else if (cnt1 < 32) {
3785
3786 bind(SHORT_UNROLLED_BEGIN);
3787 // int i = 1;
3788 movl(index, 1);
3789 cmpl(index, cnt1);
3790 jcc(Assembler::greaterEqual, SHORT_UNROLLED_LOOP_EXIT);
3791
3792 // for (; i < cnt1 ; i += 2) {
3793 bind(SHORT_UNROLLED_LOOP_BEGIN);
3794 movl(tmp3, 961);
3795 imull(result, tmp3);
3796 arrays_hashcode_elload(tmp2, Address(ary1, index, Address::times(elsize), -elsize), eltype);
|
3754 arrays_hashcode_elvload(vtmp[idx], Address(ary1, index, Address::times(elsize), 8 * idx * elsize), eltype);
3755 }
3756 // vresult = vresult * vnext + ary1[index+8*idx:index+8*idx+7];
3757 for (int idx = 0; idx < 4; idx++) {
3758 vpmulld(vresult[idx], vresult[idx], vnext, Assembler::AVX_256bit);
3759 arrays_hashcode_elvcast(vtmp[idx], eltype);
3760 vpaddd(vresult[idx], vresult[idx], vtmp[idx], Assembler::AVX_256bit);
3761 }
3762 // index += 32;
3763 addl(index, 32);
3764 // index < bound;
3765 cmpl(index, bound);
3766 jcc(Assembler::less, UNROLLED_VECTOR_LOOP_BEGIN);
3767 // }
3768
3769 lea(ary1, Address(ary1, bound, Address::times(elsize)));
3770 subl(cnt1, bound);
3771 // release bound
3772
3773 // vresult *= IntVector.fromArray(I256, power_of_31_backwards, 1);
3774 lea(tmp2, ExternalAddress(StubRoutines::x86::arrays_hashcode_powers_of_31() + (0 * sizeof(jint))));
3775 for (int idx = 0; idx < 4; idx++) {
3776 arrays_hashcode_elvload(vcoef[idx], Address(tmp2, (int)((8 * idx + 1) * sizeof(jint))), T_INT);
3777 vpmulld(vresult[idx], vresult[idx], vcoef[idx], Assembler::AVX_256bit);
3778 }
3779 // result += vresult.reduceLanes(ADD);
3780 for (int idx = 0; idx < 4; idx++) {
3781 reduceI(Op_AddReductionVI, 256/(sizeof(jint) * 8), result, result, vresult[idx], vtmp[(idx * 2 + 0) % 4], vtmp[(idx * 2 + 1) % 4]);
3782 }
3783
3784 // } else if (cnt1 < 32) {
3785
3786 bind(SHORT_UNROLLED_BEGIN);
3787 // int i = 1;
3788 movl(index, 1);
3789 cmpl(index, cnt1);
3790 jcc(Assembler::greaterEqual, SHORT_UNROLLED_LOOP_EXIT);
3791
3792 // for (; i < cnt1 ; i += 2) {
3793 bind(SHORT_UNROLLED_LOOP_BEGIN);
3794 movl(tmp3, 961);
3795 imull(result, tmp3);
3796 arrays_hashcode_elload(tmp2, Address(ary1, index, Address::times(elsize), -elsize), eltype);
|