YJIT Benchmarks

Details for Benchmarks at 2024-08-07 00:28:25 UTC

YJIT metrics from the yjit-bench suite using Ruby fa3d9fdaed

Using the geomean of the headline benchmarks for x86 YJIT 3.4.0dev is
  • 73.8% faster than CRuby 3.4.0dev
  • 1.8% faster than YJIT 3.3.4
On railsbench it is
  • 77.7% faster than CRuby 3.4.0dev
  • the same speed as YJIT 3.3.4

Performance on Headline Benchmarks

Speed of each Ruby implementation relative to the baseline CRuby measurement. Higher is better.

Memory Usage on Headline Benchmarks

Memory usage of each Ruby implementation relative to the baseline CRuby measurement. Lower is better.

Performance on Other Benchmarks

Speed of each Ruby implementation relative to the baseline CRuby measurement. Higher is better.

Memory Usage on Other Benchmarks

Memory usage of each Ruby implementation relative to the baseline CRuby measurement. Lower is better.

Performance on MicroBenchmarks

Speed of each Ruby implementation relative to the baseline CRuby measurement. Higher is better.

Memory Usage on MicroBenchmarks

Memory usage of each Ruby implementation relative to the baseline CRuby measurement. Lower is better.

Want Raw Graphs and CSV?

Benchmarks Speed Details

Benchmark Memory Usage Details

Number of Iterations and Warmups Tested

bench CRuby 3.3.4 warmups CRuby 3.3.4 iters CRuby 3.4.0dev warmups CRuby 3.4.0dev iters YJIT 3.3.4 warmups YJIT 3.3.4 iters YJIT 3.4.0dev warmups YJIT 3.4.0dev iters
activerecord 10 51 10 48 10 115 10 118
chunky-png 10 18 10 15 10 36 10 36
erubi-rails 10 10 10 10 10 20 10 18
hexapdf 10 10 10 10 10 10 10 10
liquid-c 10 345 10 334 10 478 10 488
liquid-compile 10 349 10 328 10 481 10 491
liquid-render 10 139 10 131 10 354 10 358
lobsters 10 14 10 13 10 24 10 23
mail 10 158 10 150 10 234 10 219
psych-load 10 10 10 10 10 10 10 10
railsbench 10 10 10 10 10 10 10 10
rubocop 10 136 10 128 10 228 10 236
ruby-lsp 10 128 10 121 10 179 10 206
sequel 10 314 10 306 10 426 10 411
binarytrees 10 51 10 49 10 118 10 116
blurhash 10 61 10 57 10 120 10 143
erubi 10 83 10 87 10 96 10 109
etanni 10 59 10 58 10 69 10 67
fannkuchredux 10 10 10 10 10 18 10 18
fluentd 10 10 10 10 10 10 10 10
graphql 10 10 10 10 10 10 10 10
graphql-native 10 35 10 22 10 40 10 25
lee 10 11 10 10 10 19 10 19
matmul 10 10 10 10 10 15 10 16
nbody 10 206 10 195 10 407 10 467
nqueens 10 117 10 105 10 117 10 436
optcarrot 10 10 10 10 10 10 10 10
protoboeuf 10 179 10 173 10 603 10 696
protoboeuf-encode 10 123 10 125 10 275 10 432
rack 10 464 10 461 10 788 10 897
ruby-json 10 10 10 10 10 10 10 10
rubykon 10 10 10 10 10 10 10 10
sudoku 10 10 10 10 10 31 10 30
tinygql 10 24 10 21 10 50 10 51
30k_ifelse 10 10 10 10 10 70 10 69
30k_methods 10 10 10 10 10 25 10 26
cfunc_itself 10 255 10 237 10 1105 10 1212
fib 10 103 10 99 10 616 10 618
getivar 10 257 10 219 10 1258 10 1727
keyword_args 10 92 10 85 10 720 10 947
object-new 10 229 10 198 10 299 10 247
respond_to 10 95 10 86 10 1729 10 2091
ruby-xor 10 31 10 26 10 98 10 132
setivar 10 423 10 417 10 2350 10 4876
setivar_object 10 266 10 264 10 650 10 702
setivar_young 10 265 10 264 10 651 10 696
str_concat 10 329 10 300 10 592 10 529
throw 10 986 10 913 10 1277 10 1137

Different Ruby configurations want different amounts of warmup. With no JIT, CRuby needs hardly any. YJIT and MJIT 3.0 both warm up quite quickly, while MJIT in 3.1 often slows down for a time as it compiles, after an unpredictable delay.

Benchmark YJIT Stats

Hover your cursor over the benchmark names for descriptions of each benchmark.

bench Exit Report Inline Outlined Comp iSeqs Comp Blocks Inval Inval Ratio Bind Alloc Bind Set Const Bumps Compile Time MS
activerecord (click) 1367161 1202949 214 2169 0 0% 0 0 80.906712
chunky-png (click) 288936 234128 83 1054 1 0% 0 0 35.906515
hexapdf (click) 1264595 1077329 505 11813 37 0% 0 0 415.080143
liquid-c (click) 467692 395896 119 1767 5 0% 0 0 57.887986
liquid-compile (click) 392882 325499 151 2083 2 0% 0 0 73.766959
liquid-render (click) 567650 478679 136 2323 8 0% 0 0 76.949021
mail (click) 716259 656367 347 5553 14 0% 0 0 187.997056
railsbench (click) 2895887 2452689 1636 15969 57 0% 0 0 590.321547
rubocop (click) 5442097 4592129 2892 50927 110 0% 4 0 1933.073992
ruby-lsp (click) 794249 686420 309 4745 24 0% 0 0 159.885228
binarytrees (click) 8404 7594 9 66 0 0% 0 0 2.957232
blurhash (click) 50094 43901 30 453 0 0% 0 0 17.70786
erubi (click) 220417 195490 10 112 0 0% 0 0 4.455016
etanni (click) 24958 23898 11 92 0 0% 0 0 3.583739
fannkuchredux (click) 23706 31291 4 252 0 0% 0 0 9.624682
lee (click) 252198 201524 34 531 0 0% 0 0 19.627625
matmul (click) 7957 8968 7 82 0 0% 0 0 3.161267
nbody (click) 15014 19017 10 191 0 0% 0 0 6.130679
nqueens (click) 23822 23823 9 279 0 0% 0 0 10.491034
optcarrot (click) 309010 266612 187 4398 34 0% 0 0 119.16659
protoboeuf (click) 139992 122036 16 1364 0 0% 0 0 48.935642
protoboeuf-encode (click) 212046 174944 18 1189 0 0% 0 0 36.617198
rack (click) 225005 217652 37 408 0 0% 0 0 13.385966
ruby-json (click) 24237 21549 7 166 0 0% 0 0 5.547251
rubykon (click) 128223 123238 136 1480 3 0% 0 0 48.329976
sudoku (click) 45292 24766 7 516 0 0% 0 0 19.258336
tinygql (click) 265435 253683 58 735 5 0% 0 0 24.281361
30k_ifelse (click) 5134313 4729205 9258 50774 0 0% 0 0 1902.056168
30k_methods (click) 2016559 1591611 5777 19335 0 0% 0 0 538.508884
cfunc_itself (click) 7172 6017 8 66 0 0% 0 0 2.380417
fib (click) 4609 4792 7 45 0 0% 0 0 1.858572
getivar (click) 5575 6644 7 70 0 0% 0 0 2.350477
keyword_args (click) 8042 6744 9 68 0 0% 0 0 2.645572
object-new (click) 4039 4024 6 41 0 0% 0 0 1.849194
respond_to (click) 7910 8158 8 81 0 0% 0 0 2.839673
ruby-xor (click) 5219 5772 4 58 0 0% 0 0 1.993456
setivar (click) 4726 5320 7 50 0 0% 0 0 1.911751
setivar_object (click) 5109 5228 7 50 0 0% 0 0 1.972113
setivar_young (click) 5819 5950 8 58 0 0% 0 0 2.223019
str_concat (click) 7415 7947 10 81 0 0% 0 0 3.138698
throw (click) 7538 6222 9 65 0 0% 0 0 2.69359

YJIT stats correspond to the YJIT stats exit report.

Note: currently, all stats are collected on x86_64, not ARM.

Raw JSON data files

All graphs and table data in this page comes from processing these data files, which come from benchmark runs.