3395 arrays_hashcode_elvload(vtmp[idx], Address(ary1, index, Address::times(elsize), 8 * idx * elsize), eltype);
3396 }
3397 // vresult = vresult * vnext + ary1[index+8*idx:index+8*idx+7];
3398 for (int idx = 0; idx < 4; idx++) {
3399 vpmulld(vresult[idx], vresult[idx], vnext, Assembler::AVX_256bit);
3400 arrays_hashcode_elvcast(vtmp[idx], eltype);
3401 vpaddd(vresult[idx], vresult[idx], vtmp[idx], Assembler::AVX_256bit);
3402 }
3403 // index += 32;
3404 addl(index, 32);
3405 // index < bound;
3406 cmpl(index, bound);
3407 jcc(Assembler::less, UNROLLED_VECTOR_LOOP_BEGIN);
3408 // }
3409
3410 lea(ary1, Address(ary1, bound, Address::times(elsize)));
3411 subl(cnt1, bound);
3412 // release bound
3413
3414 // vresult *= IntVector.fromArray(I256, power_of_31_backwards, 1);
3415 for (int idx = 0; idx < 4; idx++) {
3416 lea(tmp2, ExternalAddress(StubRoutines::x86::arrays_hashcode_powers_of_31() + ((8 * idx + 1) * sizeof(jint))));
3417 arrays_hashcode_elvload(vcoef[idx], Address(tmp2, 0), T_INT);
3418 vpmulld(vresult[idx], vresult[idx], vcoef[idx], Assembler::AVX_256bit);
3419 }
3420 // result += vresult.reduceLanes(ADD);
3421 for (int idx = 0; idx < 4; idx++) {
3422 reduceI(Op_AddReductionVI, 256/(sizeof(jint) * 8), result, result, vresult[idx], vtmp[(idx * 2 + 0) % 4], vtmp[(idx * 2 + 1) % 4]);
3423 }
3424
3425 // } else if (cnt1 < 32) {
3426
3427 bind(SHORT_UNROLLED_BEGIN);
3428 // int i = 1;
3429 movl(index, 1);
3430 cmpl(index, cnt1);
3431 jcc(Assembler::greaterEqual, SHORT_UNROLLED_LOOP_EXIT);
3432
3433 // for (; i < cnt1 ; i += 2) {
3434 bind(SHORT_UNROLLED_LOOP_BEGIN);
3435 movl(tmp3, 961);
3436 imull(result, tmp3);
3437 arrays_hashcode_elload(tmp2, Address(ary1, index, Address::times(elsize), -elsize), eltype);
|
3395 arrays_hashcode_elvload(vtmp[idx], Address(ary1, index, Address::times(elsize), 8 * idx * elsize), eltype);
3396 }
3397 // vresult = vresult * vnext + ary1[index+8*idx:index+8*idx+7];
3398 for (int idx = 0; idx < 4; idx++) {
3399 vpmulld(vresult[idx], vresult[idx], vnext, Assembler::AVX_256bit);
3400 arrays_hashcode_elvcast(vtmp[idx], eltype);
3401 vpaddd(vresult[idx], vresult[idx], vtmp[idx], Assembler::AVX_256bit);
3402 }
3403 // index += 32;
3404 addl(index, 32);
3405 // index < bound;
3406 cmpl(index, bound);
3407 jcc(Assembler::less, UNROLLED_VECTOR_LOOP_BEGIN);
3408 // }
3409
3410 lea(ary1, Address(ary1, bound, Address::times(elsize)));
3411 subl(cnt1, bound);
3412 // release bound
3413
3414 // vresult *= IntVector.fromArray(I256, power_of_31_backwards, 1);
3415 lea(tmp2, ExternalAddress(StubRoutines::x86::arrays_hashcode_powers_of_31() + (0 * sizeof(jint))));
3416 for (int idx = 0; idx < 4; idx++) {
3417 arrays_hashcode_elvload(vcoef[idx], Address(tmp2, (int)((8 * idx + 1) * sizeof(jint))), T_INT);
3418 vpmulld(vresult[idx], vresult[idx], vcoef[idx], Assembler::AVX_256bit);
3419 }
3420 // result += vresult.reduceLanes(ADD);
3421 for (int idx = 0; idx < 4; idx++) {
3422 reduceI(Op_AddReductionVI, 256/(sizeof(jint) * 8), result, result, vresult[idx], vtmp[(idx * 2 + 0) % 4], vtmp[(idx * 2 + 1) % 4]);
3423 }
3424
3425 // } else if (cnt1 < 32) {
3426
3427 bind(SHORT_UNROLLED_BEGIN);
3428 // int i = 1;
3429 movl(index, 1);
3430 cmpl(index, cnt1);
3431 jcc(Assembler::greaterEqual, SHORT_UNROLLED_LOOP_EXIT);
3432
3433 // for (; i < cnt1 ; i += 2) {
3434 bind(SHORT_UNROLLED_LOOP_BEGIN);
3435 movl(tmp3, 961);
3436 imull(result, tmp3);
3437 arrays_hashcode_elload(tmp2, Address(ary1, index, Address::times(elsize), -elsize), eltype);
|