YJIT Benchmarks

Details for Benchmarks at 2023-07-29 19:08:16

Overall YJIT is 57.8% faster than interpreted CRuby!
On Railsbench specifically, YJIT is 64.5% faster than CRuby!

The basic "faster" measurement is the geomean of all "headlining" x86 benchmarks on this page.

Headlining Benchmarks

These are "headlining" because the "overall" speedup above is based on these benchmarks specifically.

Speed of each Ruby implementation (iterations/second) relative to the CRuby interpreter. Higher is better.

Other Benchmarks

Speed of each Ruby implementation (iterations/second) relative to the CRuby interpreter. Higher is better.

MicroBenchmarks

Speed of each Ruby implementation (iterations/second) relative to the CRuby interpreter. Higher is better.

Want Raw Graphs and CSV?

Benchmarks Speed Details

Benchmark Memory Usage Details

Number of Iterations and Warmups Tested

bench No JIT warmups No JIT iters YJIT warmups YJIT iters
activerecord 30 493 30 493
chunky-png 30 30 30 30
erubi-rails 30 1389 30 1389
hexapdf 30 15 30 15
liquid-c 30 346 30 346
liquid-compile 30 334 30 334
liquid-render 30 195 30 195
mail 30 157 30 157
psych-load 30 15 30 15
railsbench 30 15 30 15
ruby-lsp 30 390 30 390
sequel 30 303 30 303
binarytrees 30 85 30 85
erubi 30 78 30 78
etanni 30 54 30 54
fannkuchredux 30 27 30 27
fluentd 30 15 30 15
lee 30 20 30 20
nbody 30 275 30 275
optcarrot 30 15 30 15
rack 30 187 30 187
ruby-json 30 15 30 15
rubykon 30 15 30 15
30k_ifelse 30 56 30 56
30k_methods 30 23 30 23
cfunc_itself 30 544 30 544
fib 30 453 30 453
getivar 30 939 30 939
keyword_args 30 434 30 434
respond_to 30 736 30 736
setivar 30 1666 30 1666
setivar_object 30 475 30 475
setivar_young 30 474 30 474
str_concat 30 471 30 471
throw 30 820 30 820

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
activerecord (click) 733841 688684 51 559 1 0% 0 0 0
chunky-png (click) 301935 271860 86 1063 1 0% 0 0 0
erubi-rails (click) 1818851 1555615 288 2860 1 0% 0 0 0
hexapdf (click) 1298864 1275181 600 11787 22 0% 0 0 0
liquid-c (click) 445312 446889 122 1637 3 0% 0 0 0
liquid-compile (click) 333936 322537 149 2030 2 0% 0 0 0
liquid-render (click) 477611 470710 146 1959 8 0% 0 0 0
mail (click) 653772 616725 346 4683 9 0% 0 0 0
psych-load (click) 238436 223562 64 580 1 0% 0 0 0
railsbench (click) 2606446 2311735 1359 12167 11 0% 0 0 0
ruby-lsp (click) 5683822 5354586 3412 50587 341 0% 4 0 0
sequel (click) 418306 408083 15 94 0 0% 0 0 0
binarytrees (click) 8985 7488 11 75 0 0% 0 0 0
erubi (click) 205338 203095 9 83 0 0% 0 0 0
etanni (click) 29396 27322 11 77 0 0% 0 0 0
fannkuchredux (click) 19511 20370 8 219 0 0% 0 0 0
fluentd (click) 351716 321620 12 104 0 0% 0 0 0
lee (click) 254928 248137 49 693 0 0% 0 0 0
nbody (click) 16636 16627 10 178 0 0% 0 0 0
optcarrot (click) 329352 394210 197 4506 22 0% 0 0 0
rack (click) 219474 203485 29 270 0 0% 0 0 0
ruby-json (click) 25567 22347 11 191 0 0% 0 0 0
rubykon (click) 130302 136912 145 1505 2 0% 0 0 0
30k_ifelse (click) 5410784 4368378 9263 50806 0 0% 0 0 0
30k_methods (click) 1981346 1528576 5781 19360 0 0% 0 0 0
cfunc_itself (click) 6328 4947 8 61 0 0% 0 0 0
fib (click) 4858 4454 8 49 0 0% 0 0 0
getivar (click) 6004 5980 8 74 0 0% 0 0 0
keyword_args (click) 7103 5541 9 63 0 0% 0 0 0
respond_to (click) 6865 6857 8 76 0 0% 0 0 0
setivar (click) 5155 4541 8 54 0 0% 0 0 0
setivar_object (click) 5493 4561 8 54 0 0% 0 0 0
setivar_young (click) 6280 5537 9 62 0 0% 0 0 0
str_concat (click) 6732 6350 10 76 0 0% 0 0 0
throw (click) 3939 3228 10 37 0 0% 0 0 0

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.