RFR: 8310268: RISC-V: misaligned memory access in String.Compare intrinsic [v6]

[Fei Yang]

On Mon, 24 Jul 2023 14:16:02 GMT, Vladimir Kempik <vkempik at openjdk.org> wrote:

>> Please review this fix. it eliminates misaligned loads in String.compare on risc-v
>> 
>> for small compares ( <= 72 bytes), the instrinsic in c2_MacroAssembler.cpp is used,
>> it reads ( in case of UU/LL) 8 bytes per loop, and at then end, it reads tail - misaligned load of last 8 bytes from the string.
>> 
>> so if string length is not 8x bytes long then last load is misaligned, also it performs read/compare of some data which already was processed.
>> 
>> I have changed that to compare only last length%8 bytes using SHORT_STRING part of intrinsic for UL/LU. But for UU/LL I have made an optimised version.
>> 
>> Thanks to optimisations for conditional branching at line [947](https://github.com/openjdk/jdk/pull/14534/files#diff-35eb1d2f1e2f0514dd46bd7fbad49ff2c87703d5a3041a6433956df00a3fe6e6R947) I’ve got no perf drop on thead ( with +AvoidUnalignedAccesses) which supports misaligned access.
>> 
>> Improvements to inflate_XX() methods gives 3-5% improvements for UL/LU cases on thead, almost no perf change on hifive.
>> 
>> for large strings, the instrinsics from stubGenerator.cpp is used
>> for UU/LL - generate_compare_long_string_same_encoding, I have just replaced misaligned ld with load_long_misaligned. Since this tail reading is not on hot path, this give some small penalty for thead when -XX:+AvoidUnalignedAccesses.
>> 
>> large LU/UL comparision is done in compare_long_string_different_encoding in sutbGenerator.cpp:
>> These changes are partially based on feilongjiang's patch, but I have changed tail reading to prevent reading past the end of string. I have observed no perf difference between feilongjiang's and my version.
>> 
>> This also enables regression test for string.Compare which previously was aarch64-only
>> 
>> Testing: tier1 and tier2 clean on hifive.
>> 
>> JMH testing, hifive:
>> before:
>> 
>> Benchmark                                   (delta)  (size)  Mode  Cnt     Score    Error  Units
>> StringCompareToDifferentLength.compareToLL        2      24  avgt    9     6.474 ±  1.475  ms/op
>> StringCompareToDifferentLength.compareToLL        2      36  avgt    9   125.823 ±  1.947  ms/op
>> StringCompareToDifferentLength.compareToLL        2      72  avgt    9    10.512 ±  0.236  ms/op
>> StringCompareToDifferentLength.compareToLL        2     128  avgt    9    13.032 ±  0.821  ms/op
>> StringCompareToDifferentLength.compareToLL        2     256  avgt    9    18.983 ±  0.318  ms/op
>> StringCompareToDifferentLength.compareToLL        2     512  avgt    9    29.925 ± ...
>
> Vladimir Kempik has updated the pull request incrementally with one additional commit since the last revision:
> 
>   remove unused branch

Hi, would you mind one more tweak? Thanks.

src/hotspot/cpu/riscv/stubGenerator_riscv.cpp line 2364:

> 2362:              tmpU = isLU ? tmp5 : tmp1, // where to keep U for comparison
> 2363:              tmpL = isLU ? tmp1 : tmp5; // where to keep L for comparison
> 2364: 

I see `cnt1`(x12) is unused now. We could make use of this to further eliminate use of both `tmp4` (x7) and `tmp5` (x31).
Let:

             tmpU = isLU ? tmp2 : tmp1, // where to keep U for comparison
             tmpL = isLU ? tmp1 : tmp2; // where to keep L for comparison

And use `cnt1` and `tmp2` instead where `tmp4` and `tmp5` are used respectively. This would help save one `mv` and saving & restoring of `tmp4` and `tmp5`. Also remember to change `tmpLval` from `tmp4` (x7) to `cnt1`(x12) in compare_string_8_x_LU.

src/hotspot/cpu/riscv/stubGenerator_riscv.cpp line 2391:

> 2389: 
> 2390:       __ addi(t0, cnt2, wordSize);
> 2391:       __ addi(cnt2, cnt2, wordSize*2); // amount of characters left to process

Indentation: s/__ addi(cnt2, cnt2, wordSize*2);/__ addi(cnt2, cnt2, wordSize * 2);/

src/hotspot/cpu/riscv/stubGenerator_riscv.cpp line 2410:

> 2408:       __ bnez(tmp3, CALCULATE_DIFFERENCE);
> 2409: 
> 2410:       __ addi(strL, strL, wordSize/2); // Address of last 4 bytes in Latin1 string

Indentation: s/__ addi(strL, strL, wordSize/2);/__ addi(strL, strL, wordSize / 2);/

-------------

Changes requested by fyang (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/14534#pullrequestreview-1544726633
PR Review Comment: https://git.openjdk.org/jdk/pull/14534#discussion_r1273057036
PR Review Comment: https://git.openjdk.org/jdk/pull/14534#discussion_r1273058345
PR Review Comment: https://git.openjdk.org/jdk/pull/14534#discussion_r1273059841