RFR: 8320448: Accelerate IndexOf using AVX2 [v19]

[Sandhya Viswanathan]

On Sat, 4 May 2024 19:35:21 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:
> 
>   Rearrange; add lambdas for clarity

src/hotspot/cpu/x86/stubGenerator_x86_64_string.cpp line 1082:

> 1080: // noMatch - label bound outside to jump to if there is no match
> 1081: // haystack - the address of the first byte of the haystack
> 1082: // hsLen - the sizeof the haystack

Good to specify if the size  (size of needle) and hsLen (size of haystack) is in bytes or elements.

src/hotspot/cpu/x86/stubGenerator_x86_64_string.cpp line 1149:

> 1147: 
> 1148:   if (size == (isU ? 2 : 1)) {
> 1149:     __ vpmovmskb(eq_mask, cmp_0, Assembler::AVX_256bit);

vpmovmskb is being done twice if doEarlyBailout is set to 1 (the setting we have currently).
If it helps to simplify, we could assume that doEarlyBailout is always set to 1 and remove this configurability.

src/hotspot/cpu/x86/stubGenerator_x86_64_string.cpp line 1174:

> 1172: #define lastMask rTmp
> 1173:     __ vpmovmskb(lastMask, cmp_k, Assembler::AVX_256bit);
> 1174:     __ shrq(lastMask);

did you mean to shift the lastMask by shiftVal here?

src/hotspot/cpu/x86/stubGenerator_x86_64_string.cpp line 1185:

> 1183:       if (size > (isU ? 4 : 2)) {
> 1184:         if (doEarlyBailout) {
> 1185:           __ testl(eq_mask, eq_mask);

The masks are 32 bit as we are comparing max 32 byes (256 bits) at a time. So we could consistently do either andl, testl, shrl or andq, testq, shrq.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1593225178
PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1593225488
PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1593227487
PR Review Comment: https://git.openjdk.org/jdk/pull/16753#discussion_r1593229554