Skip to content

SIMD Performance

tulip_rs uses 256-bit AVX2 f64x4 SIMD (N = 4 lanes) in two modes:

  • by_assets — process 4 different assets with the same options in a single pass
  • by_options — process 1 asset with 4 different option configurations in a single pass

All times are nanoseconds (ns). Lower is better.

by_assets: 4 Assets Simultaneously

Times represent the total wall-time for all 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
Indicator SIMD 4-Asset (ns) 4× C Tulip (ns) 4× TA-Lib (ns) Speedup vs C
ad 15,601 20,342 20,302 1.30×
adaptivemsw 1,722,745 N/A N/A —
adosc 17,596 37,673 35,405 2.14×
adx 22,027 53,976 153,892 2.45×
adxr 25,162 91,922 155,291 3.65×
ao 13,037 46,891 N/A 3.60×
apo 9,531 34,908 44,924 3.66×
aroon 73,254 154,405 296,452 2.11×
aroonosc 79,726 148,011 281,394 1.86×
atr 12,146 43,301 112,308 3.57×
avgprice 5,745 7,977 13,868 1.39×
bbands 22,512 54,069 92,483 2.40×
bop 9,494 11,177 20,107 1.18×
ccfisher 282,227 N/A N/A —
cci 70,875 298,961 488,984 4.22×
chaikinmf 20,169 N/A N/A —
chandelierexit 94,297 N/A N/A —
cmo 16,079 89,529 N/A 5.57×
cvi 12,936 57,332 N/A 4.43×
cybercycle 53,739 N/A N/A —
dema 8,330 25,931 91,739 3.11×
di 22,313 41,200 224,836 1.85×
dm 17,030 27,013 N/A 1.59×
donchianchannel 93,203 132,320 N/A 1.42×
dpo 12,178 10,873 N/A 0.89×
dx 19,627 26,268 N/A 1.34×
ef 10,430 N/A N/A —
elderray 17,740 60,483 N/A 3.41×
ema 8,465 43,435 43,465 5.13×
emv 9,425 20,135 N/A 2.14×
fisher 117,935 324,838 N/A 2.75×
fosc 18,904 37,947 N/A 2.01×
highpass 8,596 N/A N/A —
hilberttransform 24,373 N/A N/A —
hma 28,342 37,065 N/A 1.31×
homodynediscriminator 303,911 N/A 803,805 —
ichimoku 392,468 N/A N/A —
instantaneoustrendline 370,021 N/A 939,137 —
kama 12,820 34,271 43,387 2.67×
keltnerchannel 25,840 N/A N/A —
kvo 26,528 43,093 N/A 1.62×
linreg 10,046 33,865 N/A 3.37×
macd 21,980 42,290 147,120 1.92×
mama 259,810 N/A 860,886 —
marketfi 12,325 10,850 N/A 0.88×
mass 11,748 48,027 N/A 4.09×
max 21,246 56,872 72,441 2.68×
md 28,251 59,184 N/A 2.09×
medprice 4,098 6,705 9,117 1.64×
mfi 20,360 66,282 77,846 3.26×
min 35,656 109,365 126,155 3.07×
mom 3,390 5,403 7,592 1.59×
msw 358,729 2,193,753 N/A 6.11×
natr 13,737 43,256 113,309 3.15×
nvi 10,018 13,916 N/A 1.39×
obv 9,049 13,368 12,957 1.48×
ppo 9,799 37,828 57,592 3.86×
psar 65,093 48,987 31,590 0.75×
pvi 9,957 13,793 N/A 1.38×
qstick 11,571 12,221 N/A 1.06×
roc 9,641 10,611 19,761 1.10×
rocr 9,836 10,817 19,880 1.10×
roofingfilter 12,195 N/A N/A —
rsi 9,709 37,738 102,948 3.89×
sma 8,590 10,623 19,582 1.24×
smaenvelope 22,137 N/A N/A —
stddev 14,114 42,121 N/A 2.98×
stoch 86,023 195,420 205,485 2.27×
stochrsi 86,010 175,549 N/A 2.04×
supersmoother 12,525 N/A N/A —
supertrend 26,840 N/A N/A —
tema 8,837 27,466 130,085 3.11×
tr 5,885 8,018 16,312 1.36×
trendmode 258,052 N/A N/A —
trima 15,353 28,837 28,733 1.88×
trix 9,718 42,657 N/A 4.39×
trvi 16,770 N/A N/A —
tsf 9,807 33,633 N/A 3.43×
typprice 4,445 7,208 N/A 1.62×
ultosc 31,112 70,681 N/A 2.27×
vhf 78,395 156,888 N/A 2.00×
vidya 37,038 74,576 N/A 2.01×
volatility 23,909 69,862 N/A 2.92×
vortex 22,725 N/A N/A —
vosc 12,772 19,942 N/A 1.56×
vwap 15,343 N/A N/A —
vwma 15,467 19,829 N/A 1.28×
wad 13,386 19,957 N/A 1.49×
wcprice 4,387 7,143 N/A 1.63×
wilders 8,671 42,639 N/A 4.92×
willr 79,254 150,691 159,325 1.90×
wma 12,223 34,022 19,874 2.78×
zlema 8,554 33,275 N/A 3.89×

Compares SIMD (1 pass, 4 assets) against 4 independent sequential Rust calls.

Indicator 4× Rust (ns) SIMD 4-Asset (ns) SIMD / 4×Rust Speedup vs Rust
ad 18,982 15,601 0.8219 1.22×
adaptivemsw 2,051,696 1,722,745 0.8397 1.19×
adosc 26,672 17,596 0.6597 1.52×
adx 42,982 22,027 0.5125 1.95×
adxr 56,165 25,162 0.4480 2.23×
ao 22,572 13,037 0.5776 1.73×
apo 18,837 9,531 0.5061 1.98×
aroon 72,749 73,254 1.0181 0.99×
aroonosc 85,149 79,726 0.9381 1.07×
atr 18,856 12,146 0.6441 1.55×
avgprice 5,632 5,745 1.0201 0.98×
bbands 30,127 22,512 0.7499 1.33×
bop 9,451 9,494 1.0045 1.00×
ccfisher 917,848 282,227 0.3069 3.25×
cci 229,714 70,875 0.3097 3.23×
chaikinmf 28,290 20,169 0.7130 1.40×
chandelierexit 75,915 94,297 1.2479 0.80×
cmo 24,620 16,079 0.6531 1.53×
cvi 24,394 12,936 0.5303 1.89×
cybercycle 65,504 53,739 0.8204 1.22×
dema 24,627 8,330 0.3383 2.96×
di 56,286 22,313 0.3965 2.52×
dm 36,931 17,030 0.4612 2.17×
donchianchannel 51,814 93,203 1.8099 0.55×
dpo 10,429 12,178 1.1677 0.86×
dx 36,734 19,627 0.5343 1.87×
ef 18,860 10,430 0.5531 1.81×
elderray 25,126 17,740 0.7083 1.41×
ema 18,760 8,465 0.4513 2.22×
emv 9,569 9,425 0.9850 1.02×
fisher 201,433 117,935 0.5905 1.69×
fosc 31,159 18,904 0.6067 1.65×
highpass 18,716 8,596 0.4593 2.18×
hilberttransform 67,390 24,373 0.3617 2.77×
hma 33,248 28,342 0.8525 1.17×
homodynediscriminator 873,140 303,911 0.3481 2.87×
ichimoku 265,903 392,468 1.4832 0.67×
instantaneoustrendline 899,968 370,021 0.4111 2.43×
kama 28,165 12,820 0.4552 2.20×
keltnerchannel 26,052 25,840 0.9968 1.00×
kvo 37,282 26,528 0.7116 1.41×
linreg 25,518 10,046 0.3937 2.54×
macd 25,213 21,980 0.8828 1.13×
mama 898,880 259,810 0.2890 3.46×
marketfi 9,476 12,325 1.3007 0.77×
mass 21,682 11,748 0.5418 1.85×
max 20,699 21,246 1.0275 0.97×
md 50,310 28,251 0.5486 1.82×
medprice 4,291 4,098 0.9550 1.05×
mfi 30,439 20,360 0.6690 1.49×
min 30,869 35,656 1.2046 0.83×
mom 3,080 3,390 1.1033 0.91×
msw 484,788 358,729 0.7401 1.35×
natr 19,502 13,737 0.7045 1.42×
nvi 9,730 10,018 1.0296 0.97×
obv 13,779 9,049 0.6567 1.52×
ppo 19,851 9,799 0.4936 2.03×
psar 41,512 65,093 1.6158 0.62×
pvi 9,823 9,957 1.0136 0.99×
qstick 11,236 11,571 1.0298 0.97×
roc 9,460 9,641 1.0192 0.98×
rocr 9,482 9,836 1.0375 0.96×
roofingfilter 42,192 12,195 0.2891 3.46×
rsi 19,436 9,709 0.4995 2.00×
sma 9,900 8,590 0.8677 1.15×
smaenvelope 29,148 22,137 0.7798 1.28×
stddev 14,582 14,114 0.9680 1.03×
stoch 82,291 86,023 1.0452 0.96×
stochrsi 78,192 86,010 1.0976 0.91×
supersmoother 41,408 12,525 0.3025 3.31×
supertrend 58,532 26,840 0.4591 2.18×
tema 27,646 8,837 0.3196 3.13×
tr 6,083 5,885 0.9675 1.03×
trendmode 883,542 258,052 0.2921 3.42×
trima 21,974 15,353 0.6987 1.43×
trix 26,536 9,718 0.3662 2.73×
trvi 24,399 16,770 0.6874 1.45×
tsf 25,382 9,807 0.3865 2.59×
typprice 4,142 4,445 1.0732 0.93×
ultosc 65,093 31,112 0.4780 2.09×
vhf 64,940 78,395 1.2082 0.83×
vidya 47,904 37,038 0.7732 1.29×
volatility 35,290 23,909 0.6775 1.48×
vortex 32,568 22,725 0.7004 1.43×
vosc 15,466 12,772 0.8258 1.21×
vwap 13,805 15,343 1.1114 0.90×
vwma 13,535 15,467 1.1428 0.88×
wad 15,185 13,386 0.8815 1.13×
wcprice 4,248 4,387 1.0327 0.97×
wilders 18,674 8,671 0.4643 2.15×
willr 63,041 79,254 1.2596 0.79×
wma 25,004 12,223 0.4890 2.04×
zlema 23,162 8,554 0.3694 2.71×
Notable results — by_assets

68 of 93 indicators (73%) show a SIMD speedup over 4× sequential Rust. Median speedup for benefiting indicators: ~1.77×.

Category Indicator SIMD Speedup vs 4× Sequential Rust
Top performers roofingfilter 3.46×
mama 3.46×
trendmode 3.42×
supersmoother 3.31×
ccfisher 3.25×
cci 3.23×
tema 3.13×
dema 2.96×
homodynediscriminator 2.87×
hilberttransform 2.77×
trix 2.73×
zlema 2.71×
Notable improvement msw 1.35× (SDFT optimisation)
SIMD slower than sequential donchianchannel 0.55×
ichimoku 0.67×
stochrsi 0.91×

Indicators where SIMD is slower typically involve highly sequential computation or irregular memory access patterns where SIMD setup overhead dominates.

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 26,672 10,173 2.63× 2.51× 2.70×
adx 42,982 15,677 2.74× 2.69× 2.85×
adxr 56,165 27,785 2.02× 1.97× 2.13×
apo 18,837 9,633 2.00× 1.63× 2.33×
aroon 72,749 92,615 0.79× 0.77× 0.81×
aroonosc 85,149 89,298 0.95× 0.94× 0.97×
atr 18,856 8,368 2.26× 2.23× 2.29×
bbands 30,127 21,034 1.43× 1.35× 1.53×
ccfisher 917,848 277,796 3.31× 3.30× 3.32×
cci 229,714 118,510 1.94× 1.92× 1.95×
chaikinmf 28,290 19,634 1.44× 1.43× 1.46×
chandelierexit 75,915 109,012 0.70× 0.69× 0.70×
cmo 24,620 16,706 1.47× 1.47× 1.48×
cvi 24,394 22,256 1.10× 1.09× 1.11×
cybercycle 65,504 20,428 3.21× 3.19× 3.23×
dema 24,627 8,143 3.05× 2.58× 3.22×
di 56,286 17,922 3.14× 3.01× 3.32×
dm 36,931 16,203 2.28× 2.24× 2.32×
donchianchannel 51,814 108,522 0.48× 0.47× 0.48×
dpo 10,429 12,204 0.86× 0.84× 0.87×
dx 36,734 14,157 2.60× 2.53× 2.64×
ef 18,860 11,164 1.69× 1.60× 1.77×
elderray 25,126 17,401 1.46× 1.29× 1.54×
ema 18,760 8,118 2.31× 2.25× 2.40×
fisher 201,433 170,471 1.19× 1.15× 1.23×
fosc 31,159 19,160 1.63× 1.62× 1.63×
highpass 18,716 8,821 2.12× 2.10× 2.14×
hilberttransform 67,390 26,258 2.57× 2.56× 2.58×
hma 33,248 28,869 1.15× 1.15× 1.16×
ichimoku 265,903 365,365 0.73× 0.72× 0.74×
kama 28,165 12,006 2.35× 2.32× 2.36×
keltnerchannel 26,052 21,939 1.19× 1.11× 1.36×
kvo 37,282 20,035 1.86× 1.85× 1.89×
linreg 25,518 10,153 2.52× 2.49× 2.54×
macd 25,213 21,179 1.19× 1.11× 1.29×
mama 918,272 262,591 3.50× 3.50× 3.50×
mass 21,682 15,097 1.44× 1.42× 1.45×
max 20,699 27,547 0.75× 0.74× 0.76×
md 50,310 50,407 1.00× 0.99× 1.01×
mfi 30,439 19,658 1.55× 1.54× 1.56×
min 30,869 47,566 0.65× 0.59× 0.72×
mom 3,080 3,432 0.90× 0.88× 0.92×
msw 484,788 227,555 2.13× 2.10× 2.15×
natr 19,502 11,625 1.71× 1.44× 1.96×
ppo 19,851 10,033 1.98× 1.94× 2.03×
psar 41,512 50,356 0.83× 0.81× 0.84×
qstick 11,236 11,392 0.99× 0.98× 0.99×
roc 9,460 10,169 0.93× 0.85× 0.97×
rocr 9,482 10,216 0.93× 0.90× 0.95×
roofingfilter 42,192 11,512 3.66× 3.64× 3.68×
rsi 19,436 10,293 1.89× 1.82× 1.94×
sma 9,900 8,714 1.14× 1.12× 1.15×
smaenvelope 29,148 21,464 1.36× 1.22× 1.71×
stddev 14,582 14,373 1.01× 1.01× 1.02×
stoch 82,291 110,785 0.75× 0.73× 0.76×
stochrsi 78,192 98,799 0.79× 0.77× 0.82×
supersmoother 41,408 12,005 3.48× 2.97× 3.66×
supertrend 58,532 23,569 2.48× 2.43× 2.54×
tema 27,646 8,032 3.44× 3.40× 3.49×
trendmode 883,542 229,837 3.85× 3.80× 3.90×
trima 21,974 15,493 1.42× 1.41× 1.43×
trix 26,536 10,083 2.64× 2.52× 2.68×
trvi 24,399 23,132 1.06× 1.04× 1.07×
tsf 25,382 10,616 2.40× 2.35× 2.43×
ultosc 65,093 41,681 1.56× 1.56× 1.57×
vhf 64,940 89,020 0.73× 0.71× 0.74×
vidya 47,904 37,858 1.27× 1.26× 1.27×
volatility 35,290 22,541 1.57× 1.55× 1.59×
vortex 32,568 23,928 1.36× 1.25× 1.48×
vosc 15,466 12,798 1.21× 1.21× 1.21×
vwma 13,535 15,543 0.87× 0.86× 0.88×
wilders 18,674 9,048 2.09× 1.80× 2.33×
willr 63,041 85,553 0.74× 0.71× 0.76×
wma 25,004 9,850 2.54× 2.44× 2.61×
zlema 23,162 8,812 2.63× 2.50× 2.70×
Notable results — by_options

56 of 75 indicators (75%) show a SIMD speedup over 4× sequential Rust.

Category Indicator Speedup
Top performers trendmode 3.85×
roofingfilter 3.66×
mama 3.50×
supersmoother 3.48×
tema 3.44×
ccfisher 3.31×
cybercycle 3.21×
di 3.14×
dema 3.05×
Notable improvement msw 2.13× (SDFT optimisation)
SIMD slower than sequential donchianchannel 0.48×
min 0.65×
vhf 0.73×
ichimoku 0.73×

Indicators that don't benefit tend to involve complex branching or irregular memory access patterns that prevent effective vectorisation.