Benchmark Results¶

tulip_rs is benchmarked against two established C technical-analysis libraries across four scenarios:

Library	Description
Tulip Indicators (C)	The C library that inspired `tulip_rs`; used as the primary scalar baseline
TA-Lib	The industry-standard C technical-analysis library

All timings are nanoseconds (ns) — lower is better. Ratios > 1.00 mean Rust is faster than the competitor.

Methodology¶

Item	Detail
Data source	`indicator_benchmark` PostgreSQL database — latest run per indicator
Input data	Real OHLCV market data — 6,705 bars per asset (single ticker)
Options	Each indicator is run across 4 option sets (see reference below); reported time is the average across all sets
Build flags	`-C opt-level=3 -C lto=fat -C target-cpu=native`
Benchmark harness	Criterion.rs
SIMD lane count	N = 4 (256-bit AVX2 `f64x4`)
Comparison method	C libraries called through native FFI; no wrappers inside the timed region
Warm-up	Criterion warm-up phase runs before measurement to stabilise CPU frequency and cache state

N/A entries

"N/A" in a competitor column means that library does not implement the indicator. "—" in a ratio column means the ratio is not applicable (library absent or Rust is the only implementation).

Benchmark options reference — all indicators

Times are averaged across the following 4 option sets per indicator. Indicators with no options run a single configuration.

Indicator	Option sets used (averaged)
`ad`	(no options)
`adosc`	(short=2, long=5) · (short=6, long=20) · (short=5, long=15) · (short=10, long=30)
`adx`	period: 5 · 14 · 24 · 30
`adxr`	period: 5 · 14 · 24 · 30
`ao`	(no options)
`apo`	(short=2, long=5) · (short=11, long=21) · (short=5, long=11) · (short=14, long=30)
`aroon`	period: 25 · 35 · 50 · 100
`aroonosc`	period: 25 · 35 · 50 · 100
`atr`	period: 5 · 14 · 25 · 30
`avgprice`	(no options)
`bbands`	(period=5, mult=2) · (period=14, mult=2) · (period=20, mult=2) · (period=50, mult=2)
`bop`	(no options)
`cci`	period: 20 · 25 · 30 · 50
`cmo`	period: 5 · 14 · 20 · 30
`cvi`	period: 5 · 14 · 20 · 30
`dema`	period: 5 · 14 · 20 · 50
`di`	period: 5 · 14 · 20 · 30
`dm`	period: 24 · 14 · 5 · 30
`dpo`	period: 5 · 14 · 20 · 30
`dx`	period: 24 · 14 · 5 · 30
`ema`	period: 14 · 20 · 26 · 50
`emv`	(no options)
`fisher`	period: 25 · 35 · 50 · 100
`fosc`	period: 5 · 14 · 20 · 25
`hma`	period: 5 · 14 · 20 · 50
`kama`	period: 5 · 10 · 14 · 20
`kvo`	(short=2, long=5) · (short=9, long=26) · (short=14, long=30) · (short=20, long=50)
`linreg`	period: 5 · 14 · 20 · 25
`macd`	(fast=5, slow=13, signal=8) · (fast=19, slow=39, signal=9) · (fast=10, slow=30, signal=10) · (fast=6, slow=20, signal=9)
`marketfi`	(no options)
`mass`	period: 25 · 30 · 50 · 100
`max`	period: 25 · 35 · 50 · 100
`md`	period: 5 · 10 · 14 · 25
`medprice`	(no options)
`mfi`	period: 14 · 20 · 25 · 30
`min`	period: 25 · 35 · 50 · 100
`mom`	period: 25 · 30 · 50 · 100
`msw`	period: 20 · 25 · 30 · 50
`natr`	period: 14 · 20 · 25 · 30
`nvi`	(no options)
`obv`	(no options)
`ppo`	(short=2, long=5) · (short=12, long=26) · (short=9, long=20) · (short=8, long=18)
`psar`	(step=0.02, max=0.2) · (step=0.2, max=2.0) · (step=0.04, max=0.4) · (step=0.4, max=4.0)
`pvi`	(no options)
`qstick`	period: 5 · 2 · 8 · 14
`roc`	period: 25 · 30 · 50 · 100
`rocr`	period: 25 · 30 · 50 · 100
`rsi`	period: 14 · 20 · 25 · 30
`sma`	period: 50 · 100 · 200 · 300
`stddev`	period: 20 · 50 · 100 · 200
`stoch`	(k=28, ks=16, d=12) · (k=35, ks=21, d=14) · (k=50, ks=30, d=21) · (k=100, ks=50, d=30)
`stochrsi`	period: 14 · 20 · 25 · 35
`tema`	period: 5 · 14 · 20 · 50
`tr`	(no options)
`trima`	period: 14 · 20 · 25 · 30
`trix`	period: 14 · 15 · 20 · 30
`tsf`	period: 5 · 14 · 20 · 25
`typprice`	(no options)
`ultosc`	(s=2, m=3, l=5) · (s=10, m=14, l=20) · (s=14, m=20, l=50) · (s=20, m=50, l=100)
`vhf`	period: 25 · 35 · 50 · 100
`vidya`	(short=9, long=12, α=0.2) · (short=12, long=26, α=0.2) · (short=14, long=30, α=0.2) · (short=14, long=30, α=0.4)
`volatility`	period: 14 · 20 · 25 · 30
`vosc`	(short=2, long=5) · (short=5, long=20) · (short=10, long=25) · (short=14, long=28)
`vwma`	period: 14 · 20 · 25 · 30
`wad`	(no options)
`wcprice`	(no options)
`wilders`	period: 20 · 25 · 30 · 50
`willr`	period: 25 · 35 · 50 · 100
`wma`	period: 14 · 20 · 25 · 30
`zlema`	period: 5 · 10 · 14 · 20

1. Standard Performance (Single Asset)¶

Single asset, single option set. Ratios show how many times slower the competitor is relative to Rust.

Indicator	Rust (ns)	C Tulip (ns)	TA-Lib (ns)	C / Rust	TA-Lib / Rust
ad	4,768	5,088	5,097	1.07	1.07
adosc	6,712	9,209	8,869	1.37	1.32
adx	12,504	14,521	37,786	1.16	3.02
adxr	14,188	18,412	38,867	1.30	2.74
ao	5,488	11,683	N/A	2.13	—
apo	4,709	8,765	11,296	1.86	2.40
aroon	17,561	79,068	73,312	4.50	4.17
aroonosc	16,517	77,160	69,340	4.67	4.20
atr	4,631	10,741	28,159	2.32	6.08
avgprice	1,408	2,018	3,442	1.43	2.44
bbands	7,628	12,218	21,814	1.60	2.86
bop	2,361	2,857	5,005	1.21	2.12
cci	71,100	73,799	122,201	1.04	1.72
cmo	5,992	23,022	N/A	3.84	—
cvi	6,004	14,277	N/A	2.38	—
dema	6,074	6,475	22,723	1.07	3.74
di	9,188	10,207	55,684	1.11	6.06
dm	7,134	7,237	N/A	1.01	—
dpo	2,595	2,650	N/A	1.02	—
dx	10,509	10,283	N/A	0.98	—
ema	4,587	10,703	10,825	2.33	2.36
emv	2,401	5,096	N/A	2.12	—
fisher	46,046	130,141	N/A	2.83	—
fosc	7,576	9,412	N/A	1.24	—
hma	8,134	9,096	N/A	1.12	—
kama	7,091	8,401	10,586	1.18	1.49
kvo	9,168	9,861	N/A	1.08	—
linreg	6,464	8,396	N/A	1.30	—
macd	6,925	10,862	36,639	1.57	5.29
marketfi	2,336	2,722	N/A	1.17	—
mass	5,403	11,935	N/A	2.21	—
max	4,745	29,789	17,647	6.28	3.72
md	12,978	14,733	N/A	1.14	—
medprice	910	1,514	2,305	1.66	2.53
mfi	7,509	17,473	19,266	2.33	2.57
min	6,616	54,789	30,850	8.28	4.66
mom	747	1,308	1,824	1.75	2.44
msw	643,140	528,111	N/A	0.82	—
natr	4,777	10,514	27,663	2.20	5.79
nvi	2,477	3,474	N/A	1.40	—
obv	3,355	3,347	3,130	1.00	0.93
ppo	5,039	9,472	14,158	1.88	2.81
psar	10,386	12,225	7,743	1.18	0.75
pvi	2,489	3,509	N/A	1.41	—
qstick	2,808	2,993	N/A	1.07	—
roc	2,334	2,622	4,839	1.12	2.07
rocr	2,336	2,641	4,901	1.13	2.10
rsi	8,484	9,463	25,371	1.12	2.99
sma	2,471	2,678	4,859	1.08	1.97
stddev	3,651	10,548	N/A	2.89	—
stoch	17,651	91,364	51,353	5.18	2.91
stochrsi	23,078	44,517	N/A	1.93	—
tema	6,808	6,846	32,390	1.01	4.76
tr	1,497	1,959	3,737	1.31	2.50
trima	5,585	7,211	7,273	1.29	1.30
trix	6,565	10,775	N/A	1.64	—
tsf	6,398	8,369	N/A	1.31	—
typprice	1,186	1,953	N/A	1.65	—
ultosc	15,597	18,127	N/A	1.16	—
vhf	11,543	79,311	N/A	6.87	—
vidya	11,800	18,871	N/A	1.60	—
volatility	8,892	17,495	N/A	1.97	—
vosc	3,847	4,949	N/A	1.29	—
vwma	3,432	4,963	N/A	1.45	—
wad	3,870	4,992	N/A	1.29	—
wcprice	1,023	1,915	N/A	1.87	—
wilders	4,715	10,923	N/A	2.32	—
willr	17,055	81,392	41,018	4.77	2.41
wma	6,415	8,685	5,060	1.35	0.79
zlema	5,968	8,622	N/A	1.44	—

Notable results

Category	Indicator	C / Rust	TA-Lib / Rust
Largest advantage vs C Tulip	`min`	8.28×	4.66×
	`vhf`	6.87×	—
	`max`	6.28×	3.72×
	`stoch`	5.18×	2.91×
	`aroonosc`	4.67×	4.20×
Largest advantage vs TA-Lib	`atr`	2.32×	6.08×
	`di`	1.11×	6.06×
	`natr`	2.20×	5.79×
	`macd`	1.57×	5.29×
Rust slower than competitor	`dx` vs C Tulip	0.98×	—
	`msw` vs C Tulip	0.82×	—
	`obv` vs TA-Lib	—	0.93×
	`psar` vs TA-Lib	—	0.75×
	`wma` vs TA-Lib	—	0.79×

Rust beats C Tulip on all but 2 indicators (dx, msw) and beats TA-Lib on all but 3 (obv, psar, wma). Median C / Rust ratio: ~1.32×.

2. SIMD by_assets: 4 Assets Simultaneously¶

Processes 4 different assets with identical options in a single SIMD pass. All times represent the total wall-time for 4 assets.

SIMD / 4× Rust — ratio < 1.00 means SIMD is faster than 4 sequential Rust calls
Speedup vs Rust — how many times faster SIMD is compared to 4× sequential Rust
Speedup vs C — how many times faster SIMD is compared to 4× sequential C Tulip

Indicator	4× Rust (ns)	4× C Tulip (ns)	4× TA-Lib (ns)	SIMD 4-Asset (ns)	SIMD / 4×Rust	Speedup vs Rust	Speedup vs C
ad	19,073	20,352	20,389	15,712	0.82	1.21×	1.30×
adosc	26,848	36,835	35,476	17,547	0.65	1.53×	2.10×
adx	50,016	58,084	151,143	31,608	0.63	1.58×	1.84×
adxr	56,751	73,649	155,470	33,835	0.60	1.68×	2.18×
ao	21,950	46,731	N/A	13,373	0.61	1.64×	3.50×
apo	18,837	35,059	45,183	9,148	0.49	2.06×	3.83×
aroon	70,243	316,271	293,248	85,031	1.21	0.83×	3.72×
aroonosc	66,066	308,641	277,361	83,965	1.27	0.79×	3.67×
atr	18,525	42,962	112,634	14,244	0.77	1.30×	3.02×
avgprice	5,632	8,070	13,769	5,611	1.00	1.00×	1.44×
bbands	30,511	48,872	87,257	22,781	0.75	1.34×	2.15×
bop	9,444	11,428	20,020	9,603	1.02	0.98×	1.19×
cci	284,401	295,196	488,802	69,337	0.24	4.10×	4.26×
cmo	23,966	92,089	N/A	15,883	0.66	1.51×	5.80×
cvi	24,017	57,106	N/A	12,876	0.54	1.87×	4.43×
dema	24,296	25,900	90,893	9,436	0.39	2.57×	2.74×
di	36,751	40,827	222,735	25,642	0.70	1.43×	1.59×
dm	28,538	28,949	N/A	16,059	0.56	1.78×	1.80×
dpo	10,379	10,601	N/A	11,955	1.15	0.87×	0.89×
dx	42,035	41,131	N/A	28,961	0.69	1.45×	1.42×
ema	18,348	42,812	43,301	8,466	0.46	2.17×	5.06×
emv	9,605	20,382	N/A	9,630	1.00	1.00×	2.12×
fisher	184,183	520,563	N/A	121,943	0.66	1.51×	4.27×
fosc	30,302	37,648	N/A	19,000	0.63	1.60×	1.98×
hma	32,536	36,382	N/A	28,391	0.87	1.15×	1.28×
kama	28,364	33,606	42,345	12,887	0.45	2.20×	2.61×
kvo	36,673	39,446	N/A	26,299	0.72	1.39×	1.50×
linreg	25,856	33,582	N/A	9,869	0.38	2.62×	3.40×
macd	27,698	43,446	146,556	21,614	0.78	1.28×	2.01×
marketfi	9,343	10,887	N/A	11,844	1.27	0.79×	0.92×
mass	21,611	47,740	N/A	11,780	0.55	1.83×	4.05×
max	18,980	119,155	70,587	22,648	1.19	0.84×	5.26×
md	51,911	58,930	N/A	27,800	0.54	1.87×	2.12×
medprice	3,641	6,055	9,221	3,993	1.10	0.91×	1.52×
mfi	30,037	69,892	77,064	20,502	0.68	1.47×	3.41×
min	26,463	219,156	123,400	45,373	1.71	0.58×	4.83×
mom	2,989	5,233	7,297	3,302	1.10	0.91×	1.59×
msw	2,572,559	2,112,444	N/A	1,681,722	0.65	1.53×	1.26×
natr	19,109	42,058	110,653	16,645	0.87	1.15×	2.53×
nvi	9,909	13,896	N/A	9,865	1.00	1.00×	1.41×
obv	13,421	13,386	12,518	8,953	0.67	1.50×	1.49×
ppo	20,156	37,888	56,633	9,893	0.49	2.04×	3.83×
psar	41,543	48,901	30,971	70,083	1.69	0.59×	0.70×
pvi	9,956	14,037	N/A	11,866	1.19	0.84×	1.18×
qstick	11,232	11,970	N/A	11,263	1.00	1.00×	1.06×
roc	9,336	10,487	19,356	10,196	1.09	0.92×	1.03×
rocr	9,342	10,564	19,605	9,699	1.04	0.96×	1.09×
rsi	33,935	37,853	101,483	9,861	0.29	3.44×	3.84×
sma	9,884	10,711	19,436	8,501	0.86	1.16×	1.26×
stddev	14,605	42,191	N/A	14,332	0.98	1.02×	2.94×
stoch	70,602	365,457	205,413	93,950	1.33	0.75×	3.89×
stochrsi	92,313	178,069	N/A	100,933	1.09	0.91×	1.76×
tema	27,231	27,383	129,560	8,363	0.31	3.26×	3.27×
tr	5,987	7,834	14,947	5,862	0.98	1.02×	1.34×
trima	22,342	28,842	29,091	15,684	0.70	1.42×	1.84×
trix	26,259	43,100	N/A	9,829	0.37	2.67×	4.38×
tsf	25,592	33,478	N/A	10,107	0.40	2.53×	3.31×
typprice	4,742	7,811	N/A	4,415	0.93	1.07×	1.77×
ultosc	62,388	72,510	N/A	29,472	0.47	2.12×	2.46×
vhf	46,174	317,244	N/A	86,450	1.87	0.53×	3.67×
vidya	47,200	75,485	N/A	37,451	0.79	1.26×	2.02×
volatility	35,566	69,979	N/A	23,840	0.67	1.49×	2.94×
vosc	15,388	19,796	N/A	12,867	0.84	1.20×	1.54×
vwma	13,728	19,852	N/A	15,524	1.13	0.88×	1.28×
wad	15,480	19,966	N/A	13,680	0.88	1.13×	1.46×
wcprice	4,092	7,660	N/A	4,647	1.14	0.88×	1.65×
wilders	18,859	43,693	N/A	8,869	0.47	2.13×	4.93×
willr	68,219	325,567	164,072	85,118	1.25	0.80×	3.83×
wma	25,658	34,740	20,240	9,896	0.39	2.59×	3.51×
zlema	23,873	34,488	N/A	8,736	0.37	2.73×	3.95×

Notable results

60 of 80 indicators (75%) show a SIMD speedup over 4× sequential Rust.
Median speedup for benefiting indicators: ~1.40×.

Category	Indicator	SIMD Speedup vs 4× Sequential Rust
Top performers	`cci`	4.10×
	`rsi`	3.44×
	`tema`	3.26×
	`zlema`	2.73×
	`trix`	2.67×
	`wma`	2.59×
	`linreg`	2.62×
SIMD slower than sequential	`vhf`	0.53×
	`min`	0.58×
	`psar`	0.59×
	`stoch`	0.75×
	`aroonosc`	0.79×

Indicators where SIMD is slower typically involve either highly sequential computation (e.g., psar) or irregular memory access patterns where SIMD setup overhead dominates.

3. SIMD by_options: 4 Option Sets Simultaneously¶

Processes 1 asset with 4 different option configurations in a single SIMD pass. There is no C or TA-Lib equivalent for this mode — the comparison baseline is 4 independent sequential Rust calls.

Speedup — 4× Sequential Rust / SIMD 4-Options
Best / Worst — speedup range across the different option combinations tested

Indicator	4× Sequential Rust (ns)	SIMD 4-Options (ns)	Speedup	Best	Worst
adosc	26,848	9,864	2.72×	2.76×	2.69×
adx	50,016	29,746	1.68×	1.68×	1.68×
adxr	56,751	34,068	1.67×	1.69×	1.65×
apo	18,837	8,448	2.25×	2.40×	1.86×
aroon	70,243	91,485	0.77×	0.78×	0.75×
aroonosc	66,066	86,488	0.77×	0.77×	0.75×
atr	18,525	10,020	1.85×	1.94×	1.77×
bbands	30,511	22,448	1.36×	1.46×	1.29×
cci	284,401	122,043	2.33×	2.37×	2.31×
cmo	23,966	16,843	1.42×	1.44×	1.41×
cvi	24,017	21,747	1.11×	1.11×	1.10×
dema	24,296	9,120	2.69×	3.08×	2.41×
di	36,751	21,405	1.72×	1.76×	1.65×
dm	28,538	15,874	1.80×	1.87×	1.71×
dpo	10,379	12,403	0.84×	0.85×	0.83×
dx	42,035	28,349	1.48×	1.50×	1.46×
ema	18,348	8,067	2.27×	2.43×	2.14×
fisher	184,183	173,805	1.06×	1.09×	1.04×
fosc	30,302	19,223	1.58×	1.59×	1.57×
hma	32,536	28,348	1.15×	1.16×	1.14×
kama	28,364	12,838	2.21×	2.24×	2.20×
kvo	36,673	19,805	1.85×	1.91×	1.81×
linreg	25,856	10,246	2.52×	2.56×	2.49×
macd	27,698	21,760	1.27×	1.40×	1.17×
mass	21,611	15,203	1.42×	1.45×	1.40×
max	18,980	29,224	0.65×	0.68×	0.61×
md	51,911	51,534	1.01×	1.03×	0.99×
mfi	30,037	19,498	1.54×	1.56×	1.51×
min	26,463	56,215	0.47×	0.50×	0.45×
mom	2,989	3,278	0.91×	0.91×	0.91×
msw	2,572,559	2,026,360	1.27×	1.29×	1.24×
natr	19,109	11,040	1.73×	1.74×	1.72×
ppo	20,156	9,679	2.08×	2.10×	2.05×
psar	41,543	55,459	0.75×	0.77×	0.73×
qstick	11,232	11,206	1.00×	1.03×	0.98×
roc	9,336	11,928	0.78×	0.94×	0.73×
rocr	9,342	10,664	0.88×	0.95×	0.77×
rsi	33,935	9,894	3.43×	3.46×	3.40×
sma	9,884	9,058	1.09×	1.11×	1.07×
stddev	14,605	14,488	1.01×	1.02×	1.00×
stoch	70,602	116,190	0.61×	0.62×	0.59×
stochrsi	92,313	115,244	0.80×	0.84×	0.77×
tema	27,231	7,890	3.45×	3.51×	3.42×
trima	22,342	15,704	1.42×	1.46×	1.38×
trix	26,259	9,919	2.65×	2.68×	2.61×
tsf	25,592	12,264	2.09×	2.50×	1.84×
ultosc	62,388	39,726	1.57×	1.59×	1.55×
vhf	46,174	99,461	0.47×	0.48×	0.45×
vidya	47,200	37,966	1.24×	1.26×	1.22×
volatility	35,566	22,771	1.56×	1.60×	1.52×
vosc	15,388	12,799	1.20×	1.22×	1.18×
vwma	13,728	15,734	0.87×	0.90×	0.85×
wilders	18,859	8,456	2.24×	2.32×	2.02×
willr	68,219	104,286	0.65×	0.67×	0.63×
wma	25,658	10,055	2.55×	2.56×	2.54×
zlema	23,873	8,924	2.68×	2.68×	2.67×

Notable results

39 of 56 indicators (70%) show a SIMD speedup over 4× sequential Rust.

Category	Indicator	Speedup
Top performers	`tema`	3.45×
	`rsi`	3.43×
	`adosc`	2.72×
	`dema`	2.69× (up to 3.08× best-case)
	`zlema`	2.68×
	`trix`	2.65×
	`wma`	2.55×
SIMD slower than sequential	`min`	0.47×
	`vhf`	0.47×
	`stoch`	0.61×
	`max`	0.65×
	`psar`	0.75×

Indicators that don't benefit tend to involve complex branching or irregular memory access patterns that prevent effective vectorisation (min, max, stoch, psar, vhf).

4. Optional Outputs: Single-Pass Computation Advantage¶

tulip_rs computes the primary indicator and all intermediate outputs in a single pass through the data. C Tulip and TA-Lib require separate function calls for each intermediate result.

The C Equiv. Total column sums the C Tulip time for the primary indicator plus each sub-indicator called separately. The Rust time already includes every optional output.

Indicator	Optional outputs (sub-indicators)	Rust all-outputs (ns)	C Equiv. Total (ns)	C / Rust	TA-Lib Equiv. Total (ns)	TA-Lib / Rust
adosc	short_ema, long_ema, ad	8,072	35,703	4.42×	35,616	4.41×
adx	dx, atr, tr	14,473	37,504	2.59×	69,682	4.81×
adxr	adx, dx, atr, tr	22,598	55,916	2.47×	108,549	4.80×
ao	short_sma, long_sma, medprice	9,918	18,553	1.87×	—	—
apo	short_ema, long_ema	7,005	30,171	4.31×	32,946	4.70×
aroonosc	aroon_down, aroon_up	17,963	156,228	8.70×	142,652	7.94×
atr	tr	5,144	12,700	2.47×	31,896	6.20×
cci	sma, md, typprice	73,949	93,163	1.26×	127,060	1.72×
dema	ema	6,303	17,178	2.73×	33,548	5.32×
di	atr, tr	13,431	22,907	1.71×	87,580	6.52×
dpo	sma	2,774	5,328	1.92×	—	—
dx	atr, tr	11,213	22,983	2.05×	—	—
emv	medprice	2,638	6,610	2.51×	—	—
fosc	tsf, linreg, linregslope, linregintercept	10,371	26,177	2.52×	—	—
kvo	short_ema, long_ema	9,411	31,267	3.32×	—	—
linreg	linregslope, linregintercept	6,410	8,396	1.31×	—	—
macd	short_ema, long_ema	11,728	32,268	2.75×	58,289	4.97×
md	sma	13,168	17,411	1.32×	—	—
mfi	typprice	8,830	19,426	2.20×	19,266	2.18×
natr	atr, tr	6,178	23,214	3.76×	59,559	9.64×
ppo	short_ema, long_ema	5,784	30,878	5.34×	35,808	6.19×
roc	mom	2,360	3,930	1.67×	6,663	2.82×
stddev	sma	3,744	13,226	3.53×	—	—
stochrsi	rsi	23,902	53,980	2.26×	—	—
tema	dema, ema	7,639	24,024	3.15×	65,938	8.63×
trix	tema, dema, ema	8,933	34,799	3.90×	—	—
tsf	linreg, linregslope, linregintercept	7,715	16,765	2.17×	—	—
vidya	short_sma, long_sma, short_stddev, long_stddev	12,050	45,323	3.76×	—	—
vosc	short_sma, long_sma	6,074	10,305	1.70×	—	—
wma	sma	4,862	11,363	2.34×	9,919	2.04×

Footnote calculations

C Tulip times taken from the Standard benchmark (single call per sub-indicator):

#	Indicator	Calculation
adosc C	C	adosc(9,209) + ema(10,703) + ema(10,703) + ad(5,088) = 35,703
adosc TA	TA-Lib	adosc(8,869) + ema(10,825) + ema(10,825) + ad(5,097) = 35,616
adx C	C	adx(14,521) + dx(10,283) + atr(10,741) + tr(1,959) = 37,504
adx TA	TA-Lib	adx(37,786) + atr(28,159) + tr(3,737) = 69,682 (dx not in TA-Lib)
adxr C	C	adxr(18,412) + adx(14,521) + dx(10,283) + atr(10,741) + tr(1,959) = 55,916
adxr TA	TA-Lib	adxr(38,867) + adx(37,786) + atr(28,159) + tr(3,737) = 108,549
ao C	C	ao(11,683) + sma(2,678) + sma(2,678) + medprice(1,514) = 18,553
apo C	C	apo(8,765) + ema(10,703) + ema(10,703) = 30,171
apo TA	TA-Lib	apo(11,296) + ema(10,825) + ema(10,825) = 32,946
aroonosc C	C	aroonosc(77,160) + aroon(79,068) = 156,228 (aroon returns both up and down in one call)
aroonosc TA	TA-Lib	aroonosc(69,340) + aroon(73,312) = 142,652
atr C	C	atr(10,741) + tr(1,959) = 12,700
atr TA	TA-Lib	atr(28,159) + tr(3,737) = 31,896
cci C	C	cci(73,799) + sma(2,678) + md(14,733) + typprice(1,953) = 93,163
cci TA	TA-Lib	cci(122,201) + sma(4,859) = 127,060 (md and typprice not available separately)
dema C	C	dema(6,475) + ema(10,703) = 17,178
dema TA	TA-Lib	dema(22,723) + ema(10,825) = 33,548
di C	C	di(10,207) + atr(10,741) + tr(1,959) = 22,907
di TA	TA-Lib	di(55,684) + atr(28,159) + tr(3,737) = 87,580
dpo C	C	dpo(2,650) + sma(2,678) = 5,328
dx C	C	dx(10,283) + atr(10,741) + tr(1,959) = 22,983
emv C	C	emv(5,096) + medprice(1,514) = 6,610
fosc C	C	fosc(9,412) + tsf(8,369) + linreg(8,396) = 26,177 (slope/intercept have no standalone C equivalent)
kvo C	C	kvo(9,861) + ema(10,703) + ema(10,703) = 31,267
linreg C	C	linreg(8,396) (slope and intercept not available as standalone in C Tulip)
macd C	C	macd(10,862) + ema(10,703) + ema(10,703) = 32,268
macd TA	TA-Lib	macd(36,639) + ema(10,825) + ema(10,825) = 58,289
md C	C	md(14,733) + sma(2,678) = 17,411
mfi C	C	mfi(17,473) + typprice(1,953) = 19,426
mfi TA	TA-Lib	mfi(19,266) (typprice not available separately in TA-Lib)
natr C	C	natr(10,514) + atr(10,741) + tr(1,959) = 23,214
natr TA	TA-Lib	natr(27,663) + atr(28,159) + tr(3,737) = 59,559
ppo C	C	ppo(9,472) + ema(10,703) + ema(10,703) = 30,878
ppo TA	TA-Lib	ppo(14,158) + ema(10,825) + ema(10,825) = 35,808
roc C	C	roc(2,622) + mom(1,308) = 3,930
roc TA	TA-Lib	roc(4,839) + mom(1,824) = 6,663
stddev C	C	stddev(10,548) + sma(2,678) = 13,226
stochrsi C	C	stochrsi(44,517) + rsi(9,463) = 53,980
tema C	C	tema(6,846) + dema(6,475) + ema(10,703) = 24,024
tema TA	TA-Lib	tema(32,390) + dema(22,723) + ema(10,825) = 65,938
trix C	C	trix(10,775) + tema(6,846) + dema(6,475) + ema(10,703) = 34,799
tsf C	C	tsf(8,369) + linreg(8,396) = 16,765 (slope/intercept have no standalone C equivalent)
vidya C	C	vidya(18,871) + sma(2,678) + sma(2,678) + stddev(10,548) + stddev(10,548) = 45,323
vosc C	C	vosc(4,949) + sma(2,678) + sma(2,678) = 10,305
wma C	C	wma(8,685) + sma(2,678) = 11,363
wma TA	TA-Lib	wma(5,060) + sma(4,859) = 9,919

Optional Outputs Ranked by Advantage¶

Rank	vs	Indicator	Ratio	Key sub-indicators
1	TA-Lib	`natr`	9.64×	atr, tr
2	C	`aroonosc`	8.70×	aroon_down, aroon_up
3	TA-Lib	`tema`	8.63×	dema, ema
4	TA-Lib	`di`	6.52×	atr, tr
5	TA-Lib	`atr`	6.20×	tr
6	TA-Lib	`ppo`	6.19×	short_ema, long_ema
7	C	`ppo`	5.34×	short_ema, long_ema
8	TA-Lib	`dema`	5.32×	ema
9	TA-Lib	`adx`	4.81×	dx, atr, tr
10	TA-Lib	`adxr`	4.80×	adx, dx, atr, tr

5. Summary and Key Findings¶

Standard (single asset)¶

Rust beats C Tulip for all but 2 indicators: dx (0.98×) and msw (0.82×)
Rust beats TA-Lib for all but 3 indicators: obv (0.93×), psar (0.75×), wma (0.79×)
Median C / Rust ratio: ~1.32× (Rust is ~32% faster on average)
Largest wins vs C Tulip: min (8.28×), vhf (6.87×), max (6.28×), stoch (5.18×)
Largest wins vs TA-Lib: atr (6.08×), di (6.06×), natr (5.79×), macd (5.29×)

SIMD by_assets¶

75% of indicators (60 / 80) show a SIMD speedup over 4× sequential Rust
Top performers: cci (4.10×), rsi (3.44×), tema (3.26×), zlema (2.73×)
Indicators that don't benefit are mostly trivial or have inherently sequential dependencies (psar, min, max, stoch, vhf)

SIMD by_options¶

70% of indicators (39 / 56) show a SIMD speedup over 4× sequential Rust
Top performers: tema (3.45×), rsi (3.43×), adosc (2.72×), dema (2.69×), zlema (2.68×)
Same non-benefiting indicators as by_assets — the bottleneck is algorithmic, not data layout

Optional Outputs Single-Pass¶

31 indicators offer optional outputs computed in a single pass
C Tulip would require 2–5× more compute time to obtain the same information via separate calls
TA-Lib multipliers are often higher still due to its slower base times
Standouts: aroonosc (8.70× vs C), natr (9.64× vs TA-Lib), tema (8.63× vs TA-Lib)

Combined Advantage: a Worked Example¶

Consider computing tema with all its sub-indicators (dema, ema) across 4 assets simultaneously with SIMD:

Implementation	Time
`tulip_rs` SIMD 4-asset (all outputs, one pass)	8,363 ns
C Tulip — 4 × (tema + dema + ema) sequential	96,096 ns
TA-Lib — 4 × (tema + dema + ema) sequential	264,152 ns

Combined speedup

~11.5× faster than C Tulip and ~31.6× faster than TA-Lib for the same result.

6. Streaming / Stateful: Single-Bar Update¶

tulip_rs indicators implement a from_state path: once the initial lookback has been computed, each subsequent bar is updated using only the saved state — no need to reprocess the full history. This is what state.batch_indicator() uses under the hood when called with a single new bar.

Batch (ns) — full indicator computation over the complete 6,705-bar dataset (from Section 1)
Streaming 1-bar (ns) — time to update one new bar using saved state
Batch / Streaming — how many times faster the streaming path is vs a full batch recompute

Indicator	Batch Rust (ns)	Streaming 1-bar (ns)	Batch / Streaming
ad	4,768	25	190.7×
adosc	6,712	28	239.7×
adx	12,504	87	143.7×
adxr	14,188	82	173.0×
ao	5,488	15	365.9×
apo	4,709	24	196.2×
aroon	17,561	37	474.6×
aroonosc	16,517	34	485.8×
atr	4,631	23	201.3×
avgprice	1,408	23	61.2×
bbands	7,628	34	224.4×
bop	2,361	24	98.4×
cci	71,100	35	2,031×
cmo	5,992	29	206.6×
cvi	6,004	17	353.2×
dema	6,074	23	264.1×
di	9,188	113	81.3×
dm	7,134	68	104.9×
dpo	2,595	25	103.8×
dx	10,509	72	145.9×
ema	4,587	15	305.8×
emv	2,401	25	96.0×
fisher	46,046	37	1,245×
fosc	7,576	31	244.4×
hma	8,134	32	254.2×
kama	7,091	26	272.7×
kvo	9,168	31	295.7×
linreg	6,464	30	215.5×
macd	6,925	21	329.8×
marketfi	2,336	23	101.6×
mass	5,403	19	284.4×
max	4,745	29	163.6×
md	12,978	31	418.6×
medprice	910	23	39.6×
mfi	7,509	28	268.2×
min	6,616	28	236.3×
mom	747	26	28.7×
msw	643,140	117	5,498×
natr	4,777	28	170.6×
nvi	2,477	24	103.2×
obv	3,355	23	145.9×
ppo	5,039	27	186.6×
psar	10,386	25	415.4×
pvi	2,489	23	108.2×
qstick	2,808	28	100.3×
roc	2,334	26	89.8×
rocr	2,336	27	86.5×
rsi	8,484	70	121.2×
sma	2,471	24	102.9×
stddev	3,651	31	117.8×
stoch	17,651	41	430.5×
stochrsi	23,078	45	513.0×
tema	6,808	26	261.8×
tr	1,497	23	65.1×
trima	5,585	26	214.8×
trix	6,565	28	234.5×
tsf	6,398	30	213.3×
typprice	1,186	15	79.1×
ultosc	15,597	30	519.9×
vhf	11,543	30	384.8×
vidya	11,800	35	337.1×
volatility	8,892	30	296.4×
vosc	3,847	34	113.1×
vwma	3,432	32	107.3×
wad	3,870	24	161.3×
wcprice	1,023	23	44.5×
wilders	4,715	15	314.3×
willr	17,055	36	473.8×
wma	6,415	18	356.4×
zlema	5,968	26	229.5×

Key findings

All 72 indicators implement the stateful streaming path
Median streaming time: ~26 ns per bar update
Fastest — ao, ema, typprice, wilders at 15 ns/bar
Slowest — msw at 117 ns/bar (still faster than the quickest batch indicator, mom at 747 ns)
The streaming path is 100–5,500× faster than full batch recompute depending on the indicator

Rank	Indicator	Batch / Streaming	Why streaming wins so much
1	`msw`	5,498×	Mesa Sine Wave requires FFT-style computation over the full window
2	`cci`	2,031×	Large rolling sum + mean deviation computed over every bar in batch
3	`fisher`	1,245×	Complex normalisation over entire input range
4	`ultosc`	520×	Three-period weighted TR average; all periods recomputed in batch
5	`stochrsi`	513×	RSI + stochastic — two full-window passes per batch call
6	`aroonosc`	486×	Full lookback window scan for highest/lowest on every batch call
7	`aroon`	475×	Same as `aroonosc`
8	`willr`	474×	Rolling high/low window scan
9	`stoch`	431×	Rolling high/low window scan
10	`psar`	415×	Parabolic SAR iterates entire history in batch