RFR: 8320448: Accelerate IndexOf using AVX2 [v41]

[Sandhya Viswanathan]

On Fri, 24 May 2024 23:15:26 GMT, Scott Gibbons <sgibbons at openjdk.org> wrote:

>> Re-write the IndexOf code without the use of the pcmpestri instruction, only using AVX2 instructions.  This change accelerates String.IndexOf on average 1.3x for AVX2.  The benchmark numbers:
>> 
>> 
>> Benchmark	                                               Score		Latest		
>> StringIndexOf.advancedWithMediumSub   343.573		317.934		0.925375393x
>> StringIndexOf.advancedWithShortSub1	  1039.081		1053.96		1.014319384x
>> StringIndexOf.advancedWithShortSub2	      55.828		110.541		1.980027943x
>> StringIndexOf.constantPattern	                9.361		11.906		1.271872663x
>> StringIndexOf.searchCharLongSuccess	        4.216		4.218		1.000474383x
>> StringIndexOf.searchCharMediumSuccess	3.133		3.216		1.02649218x
>> StringIndexOf.searchCharShortSuccess	3.76		        3.761		1.000265957x
>> StringIndexOf.success	                                9.186		9.713		1.057369911x
>> StringIndexOf.successBig	                      14.341		46.343		3.231504079x
>> StringIndexOfChar.latin1_AVX2_String	  6220.918		12154.52		1.953814533x
>> StringIndexOfChar.latin1_AVX2_char	  5503.556		5540.044		1.006629895x
>> StringIndexOfChar.latin1_SSE4_String	  6978.854		6818.689		0.977049957x
>> StringIndexOfChar.latin1_SSE4_char	  5657.499		5474.624		0.967675646x
>> StringIndexOfChar.latin1_Short_String	  7132.541		6863.359		0.962260014x
>> StringIndexOfChar.latin1_Short_char	16013.389	      16162.437		1.009307711x
>> StringIndexOfChar.latin1_mixed_String	  7386.123	      14771.622		1.999915517x
>> StringIndexOfChar.latin1_mixed_char	  9901.671		9782.245		0.987938803
>
> Scott Gibbons has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Fix test; review comments

src/hotspot/cpu/x86/c2_stubGenerator_x86_64_string.cpp line 316:

> 314:   // Set up jump tables.  Used when needle size <= NUMBER_OF_CASES
> 315:   setup_jump_tables(ae, L_returnRBP, L_checkRangeAndReturn, L_bigCaseFixupAndReturn,
> 316:                     &big_jump_table, &small_jump_table, _masm);

We could directly use L_returnError here instead of L_returnRBP.

src/hotspot/cpu/x86/c2_stubGenerator_x86_64_string.cpp line 476:

> 474: // Used to check and return value in rbp - usually error
> 475:   __ bind(L_returnRBP);
> 476:   __ movq(rax, rbp);

This seems spurious as rax is being overwritten at line 489. Did you intend to return -1?

src/hotspot/cpu/x86/c2_stubGenerator_x86_64_string.cpp line 1816:

> 1814:         byte_compare_helper(i + 1, L_loopTop, L_fixup, needle, needle_val, hs_ptr, eq_mask, set_bit,
> 1815:                             rTmp4, ae, _masm);
> 1816:       }

L_checkRange on NoMatch could be L_error instead.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1614172379
PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1614172021
PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1614175081