YJIT Benchmarks

Details for Benchmarks at 2025-01-24 03:43:20 GMT

YJIT metrics from the yjit-bench suite using Ruby 881924f259.

¶Performance on Headline Benchmarks

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

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

¶Memory Usage on Headline Benchmarks

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

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

¶Performance on Other Benchmarks

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

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

¶Memory Usage on Other Benchmarks

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

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

¶Performance on MicroBenchmarks

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

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

¶Memory Usage on MicroBenchmarks

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

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

¶Want Raw Graphs and CSV?

• Headlining benchmarks (x86_64)
• All benchmarks (x86_64)
• CSV speed results (x86_64)
• Headlining benchmarks (aarch64)
• All benchmarks (aarch64)
• CSV speed results (aarch64)

¶Benchmarks Speed Details

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

¶Benchmark Memory Usage Details

Select Platform

Xeon x86_64 AWS Graviton ARM64

X

bench	CRuby 3.3.6 mem (MiB)	CRuby 3.5.0dev mem (MiB)	YJIT 3.3.6 mem (MiB)	YJIT 3.5.0dev mem (MiB)	Inline Code	Outlined Code	YJIT Mem overhead
activerecord	63	64	72	65	2	2	1.2%
chunky-png	57	43	56	43	1	1	1.1%
erubi-rails	106	105	117	105	2	2	-0.4%
hexapdf	157	125	288	166	2	2	32.7%
liquid-c	36	28	39	33	1	1	15.4%
liquid-compile	32	26	35	29	1	1	9.9%
liquid-render	34	28	39	34	1	1	18.6%
lobsters	279	272	316	317	8	7	16.3%
mail	52	49	58	51	1	1	3.9%
psych-load	32	26	34	28	1	1	6.6%
railsbench	107	97	121	115	4	3	17.9%
rubocop	91	82	124	122	6	5	47.5%
ruby-lsp	71	60	76	59	1	1	-0.8%
sequel	37	29	40	33	1	1	11.7%
binarytrees	26	21	26	22	1	1	2.6%
blurhash	21	14	22	15	1	1	5.7%
erubi	28	36	30	38	1	1	5.3%
etanni	23	19	24	20	1	1	2.5%
fannkuchredux	20	13	20	13	1	1	5.2%
fluentd	192	284	205	215	1	1	-24.4%
graphql	39	34	41	36	1	1	5.9%
graphql-native	39	35	43	34	1	1	-2.6%
lee	30	27	31	29	1	1	6.9%
matmul	27	21	28	22	1	1	2.5%
nbody	20	13	20	13	1	1	4.2%
nqueens	20	13	21	15	1	1	10.1%
optcarrot	61	48	68	50	1	1	3.9%
protoboeuf	36	31	40	27	1	1	-12.1%
protoboeuf-encode	63	46	69	78	1	1	66.7%
rack	31	22	32	25	1	1	13.1%
ruby-json	20	14	21	14	1	1	4.0%
rubykon	24	17	26	18	1	1	6.1%
sudoku	20	13	21	14	1	1	5.4%
tinygql	26	19	29	22	1	1	15.5%
30k_ifelse	61	110	91	118	6	6	7.5%
30k_methods	49	53	59	56	3	2	5.9%
cfunc_itself	20	13	20	13	1	1	4.1%
fib	20	13	20	13	1	1	4.2%
getivar	20	13	20	13	1	1	4.2%
keyword_args	20	13	20	13	1	1	4.1%
loops-times	22	15	24	21	1	1	38.3%
object-new	20	13	20	13	1	1	4.1%
respond_to	20	13	20	13	1	1	4.5%
ruby-xor	21	14	21	14	1	1	3.7%
setivar	20	13	20	13	1	1	4.6%
setivar_object	20	13	20	13	1	1	4.1%
setivar_young	20	13	20	13	1	1	4.2%
str_concat	46	39	60	40	1	1	1.4%
throw	20	13	20	13	1	1	4.1%

Memory is shown in mebibytes (1024 * 1024 bytes.)

Older YJIT allocated an additional 256MiB for generated code. Current YJIT allocates executable memory on demand, so this overhead should no longer be present.

¶Number of Iterations and Warmups Tested

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.

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.

• aarch64_prev_ruby_no_jit
• aarch64_prev_ruby_yjit
• aarch64_prod_ruby_no_jit
• aarch64_prod_ruby_with_yjit
• aarch64_yjit_stats
• x86_64_prev_ruby_no_jit
• x86_64_prev_ruby_yjit
• x86_64_prod_ruby_no_jit
• x86_64_prod_ruby_with_yjit
• x86_64_yjit_stats