Standard Performance: Go

Reference: cinar/indicator/v2 (pure Go), compared against the tulip_rs_go cgo binding.

Only indicators where cinar is a genuine param-compatible equivalent are shown (cinar v2's stream API fixes some parameters or computes fewer output rows; those indicators have tulip_rs_go timings in the database but no comparable competitor).

Indicator	Rust native (ns)	tulip_rs_go (ns)	cinar (ns)	Speedup vs cinar
ad	4,799	5,478	9,095,226	1,660×
adosc	5,657	6,626	23,164,241	3,496×
ao	5,427	6,366	22,578,745	3,547×
apo	4,872	5,452	5,162,216	947×
aroon	18,771	17,936	13,797,216	769×
atr	4,825	5,455	16,812,744	3,082×
bbands	6,490	10,015	19,839,429	1,981×
bop	2,440	3,137	7,441,486	2,372×
cci	78,669	79,522	33,112,153	416×
chaikinmf	7,201	10,203	22,629,489	2,218×
chandelierexit	18,411	18,221	23,706,438	1,301×
dema	6,336	6,405	14,623,870	2,283×
donchianchannel	13,219	14,053	18,828,184	1,340×
dpo	2,701	3,093	18,967,214	6,133×
elderray	6,065	7,441	16,205,745	2,178×
ema	4,784	5,434	3,270,234	602×
emv	2,449	3,134	23,471,574	7,491×
fisher	48,681	64,177	33,741,271	526×
hma	9,349	9,323	18,461,147	1,980×
kama	7,436	7,623	29,176,244	3,828×
max	5,203	6,644	12,656,791	1,905×
mfi	7,987	9,726	24,488,713	2,518×
min	7,022	7,290	11,092,422	1,522×
nvi	2,678	3,345	22,079,337	6,601×
obv	3,513	4,181	5,771,541	1,380×
ppo	5,062	6,154	22,736,404	3,695×
qstick	2,571	3,237	15,645,711	4,833×
roc	2,437	3,022	3,651,393	1,208×
rsi	4,994	6,005	22,302,800	3,714×
sma	2,558	3,071	10,903,048	3,550×
smaenvelope	7,103	8,596	20,025,959	2,330×
stddev	3,795	7,780	8,482,636	1,090×
stoch	20,383	22,865	18,320,716	801×
supertrend	12,074	13,478	35,116,577	2,606×
tema	6,789	7,406	26,263,849	3,546×
tr	1,393	2,096	10,745,694	5,126×
trima	5,626	5,991	17,223,907	2,875×
trix	6,410	7,345	21,044,608	2,865×
typprice	1,086	1,721	10,168,992	5,910×
ultosc	15,579	16,061	33,914,080	2,112×
vwma	3,501	5,458	19,700,709	3,610×
wcprice	1,075	1,697	8,078,139	4,762×
wilders	4,777	5,413	1,919,589	355×
willr	17,282	15,429	24,589,898	1,594×
wma	4,943	5,408	4,593,886	849×

Notable results

tulip_rs_go beats cinar on 45 of 45 compared indicators (cinar is wired where it is param-compatible; the remaining indicators have no pure-Go competitor run). Median speedup: 2,283× — cinar v2's goroutine/channel stream API pays a per-bar scheduling cost the cgo path does not.

Largest wins vs cinar:

Indicator	Speedup vs cinar
emv	7,491×
nvi	6,601×
dpo	6,133×
typprice	5,910×
tr	5,126×
qstick	4,833×
wcprice	4,762×
kama	3,828×
rsi	3,714×
ppo	3,695×

Smallest margins (where cinar's implementation is closest, or computes fewer output rows):

Indicator	Rust native (ns)	tulip_rs_go (ns)	cinar (ns)	Speedup vs cinar
wilders	4,777	5,413	1,919,589	355×
cci	78,669	79,522	33,112,153	416×
fisher	48,681	64,177	33,741,271	526×
ema	4,784	5,434	3,270,234	602×
aroon	18,771	17,936	13,797,216	769×
stoch	20,383	22,865	18,320,716	801×

cgo boundary overhead — the gap between Rust native and tulip_rs_go columns reflects the fixed per-call cost of the cgo boundary (argument marshalling, pointer handoff), 1.2–1.6× for the fastest-running indicators:

Indicator	Rust native (ns)	tulip_rs_go (ns)	Overhead
roc	2,437	3,022	~1.2×
emv	2,449	3,134	~1.3×
bop	2,440	3,137	~1.3×
tr	1,393	2,096	~1.5×
wcprice	1,075	1,697	~1.6×
typprice	1,086	1,721	~1.6×