Optimise toFastHex for engine_getBlobsV2

[siladu]

Replaces tuweni impl, removing the megamorphic calls.
engine_getBlobsV2 uses this since returned blob data is bottlenecked on this conversion.
Other getPayload calls will be updated to use it in a followup PR.

1 blob = 128 KB = 131072 bytes + (1 * 128 proofs * 48 bytes) = 137,216 bytes
9 blobs = 9 * 131072 = 1,179,648 bytes + (9 * 128 proofs * 48 bytes) = 1,234,944 bytes (~1.2 MB)
21 blobs = 21 * 131072 = 2,752,512 bytes + (21 * 128 proofs * 48 bytes) = 2,881,536 bytes (~2.7 MB)
72 blobs = 72 * 131072 = 9,437,184 bytes + (72 * 128 proofs * 48 bytes) = 9,879,552 bytes (~9.4 MB)
128 blobs = 128 * 131072 = 16,777,216 bytes + (128 * 128 proofs * 48 bytes) = 17,563,648 bytes (~16.7 MB)

Benchmark                        (caseName)  Mode  Cnt          Score        Error  Units
# before
ToFastHexBenchmark.tuweni       SIZE_1_BLOB  avgt   15     796406.136 ±   4813.701  ns/op
ToFastHexBenchmark.tuweni      SIZE_9_BLOBS  avgt   15    7944644.764 ±  62536.148  ns/op
ToFastHexBenchmark.tuweni     SIZE_21_BLOBS  avgt   15   18469751.559 ± 130136.114  ns/op
ToFastHexBenchmark.tuweni     SIZE_72_BLOBS  avgt   15   63279210.037 ± 279280.513  ns/op
ToFastHexBenchmark.tuweni    SIZE_128_BLOBS  avgt   15  112467237.963 ± 311872.696  ns/op
# this PR
ToFastHexBenchmark.hexUtils     SIZE_1_BLOB  avgt   15     340773.984 ±   9617.594  ns/op
ToFastHexBenchmark.hexUtils    SIZE_9_BLOBS  avgt   15    3040690.635 ±  15067.695  ns/op
ToFastHexBenchmark.hexUtils   SIZE_21_BLOBS  avgt   15    6998597.390 ±  20288.077  ns/op
ToFastHexBenchmark.hexUtils   SIZE_72_BLOBS  avgt   15   24413063.387 ± 787985.528  ns/op
ToFastHexBenchmark.hexUtils  SIZE_128_BLOBS  avgt   15   43078088.889 ± 650158.811  ns/op

5-6x faster throughput

Before

This PR

---

Hoodi performance showing control, initial StringBuilder optimisation, final char array impl

[ahamlat]

Could you share JMH benchmarks to check before and after to see if the new implementation is better in terms of performance ?

[siladu]

Input size is byte length representing blob (+proof) sizes used in responses for engine_getBlobsV2

1 blob = 128 KB = 131072 bytes + (1 * 128 proofs * 48 bytes) = 137,216 bytes
9 blobs = 9 * 131072 = 1,179,648 bytes + (9 * 128 proofs * 48 bytes) = 1,234,944 bytes (~1.2 MB)
21 blobs = 21 * 131072 = 2,752,512 bytes + (21 * 128 proofs * 48 bytes) = 2,881,536 bytes (~2.7 MB)
72 blobs = 72 * 131072 = 9,437,184 bytes + (72 * 128 proofs * 48 bytes) = 9,879,552 bytes (~9.4 MB)
128 blobs = 128 * 131072 = 16,777,216 bytes + (128 * 128 proofs * 48 bytes) = 17,563,648 bytes (~16.7 MB)

Benchmark                          (inputSize)  Mode  Cnt          Score         Error  Units
ToFastHexBenchmark.hexUtils             137216  avgt   15     347772.399 ±   11241.508  ns/op
ToFastHexBenchmark.hexUtils            1234944  avgt   15    3102369.431 ±   57631.705  ns/op
ToFastHexBenchmark.hexUtils            2881536  avgt   15    7237844.530 ±   81062.748  ns/op
ToFastHexBenchmark.hexUtils            9879552  avgt   15   24894359.969 ±  906055.271  ns/op
ToFastHexBenchmark.hexUtils           17563648  avgt   15   43469924.536 ±  337368.591  ns/op
ToFastHexBenchmark.tuweni               137216  avgt   15     808014.190 ±    2489.838  ns/op
ToFastHexBenchmark.tuweni              1234944  avgt   15    8084398.389 ±  199896.569  ns/op
ToFastHexBenchmark.tuweni              2881536  avgt   15   18816737.733 ±  402653.828  ns/op
ToFastHexBenchmark.tuweni              9879552  avgt   15   64368279.170 ±  385450.292  ns/op
ToFastHexBenchmark.tuweni             17563648  avgt   15  114858759.548 ± 2828625.738  ns/op

[siladu]

Using char array instead of StringBuilder

Benchmark                          (inputSize)  Mode  Cnt          Score        Error  Units
ToFastHexBenchmark.hexUtils             137216  avgt   15     128046.093 ±    547.659  ns/op
ToFastHexBenchmark.hexUtils            1234944  avgt   15    1383249.368 ± 366222.851  ns/op
ToFastHexBenchmark.hexUtils            2881536  avgt   15    3871263.657 ±  72802.766  ns/op
ToFastHexBenchmark.hexUtils            9879552  avgt   15   13176864.399 ± 109785.248  ns/op
ToFastHexBenchmark.hexUtils           17563648  avgt   15   23378568.582 ± 160927.048  ns/op
ToFastHexBenchmark.tuweni               137216  avgt   15     823368.888 ±   5551.835  ns/op
ToFastHexBenchmark.tuweni              1234944  avgt   15    8236590.901 ± 204055.673  ns/op
ToFastHexBenchmark.tuweni              2881536  avgt   15   19031861.216 ±  98560.324  ns/op
ToFastHexBenchmark.tuweni              9879552  avgt   15   65268289.048 ± 560765.539  ns/op
ToFastHexBenchmark.tuweni             17563648  avgt   15  115915573.133 ± 617974.569  ns/op

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.