Cytoscape.js GPU benchmarks

v4 executor 'cpu' vs 'gpu' · times are per-call p50 · 2026-08-10 22:13 · 44e55dc9 (v4) · chromium v24.18.0 · algorithms-gpu profile

CPU	Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz
Cores	8 physical / 16 logical · 3.6 GHz base · 5 GHz max
Memory	62.6 GiB
GPU	CoffeeLake-S GT2 [UHD Graphics 630] i915 Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] (8.0 GiB) amdgpu
Adapter	amd gcn-4
Browser	chromium 149.0.7827.55
OS	Fedora Linux 43 (Workstation Edition) · 6.19.14-200.fc43.x86_64 · x64
Node	v24.18.0 (v8 13.6.233.17-node.50)
Commit	44e55dc9 perf(65.10): the CPU takes two families back — flat hierarchical merges, sparse PageRank, and honest re-routing (+ the wide benchmark page)
Machine id	5cf3f79c

Speedup overview cpu p50 ÷ gpu p50, log scale · right of the 1× line = gpu faster

markovClustering @ 1.024k

647×

markovClustering @ 512

285×

kMedoids @ 4.096k

121×

markovClustering @ 256

70×

fuzzyCMeans @ 65.536k

70×

kMedoids @ 2.048k

53×

fuzzyCMeans @ 16.384k

38×

floydWarshall @ 1.024k

29×

fuzzyCMeans @ 4.096k

27×

kMeans @ 65.536k

22×

kMedoids @ 1.024k

20×

betweennessCentrality @ 2.048k

17×

kMeans @ 16.384k

14×

floydWarshall @ 512

12×

betweennessCentrality @ 1.024k

9.6×

kMeans @ 4.096k

7.6×

betweennessCentrality @ 512

4.0×

affinityPropagation @ 1.024k

3.7×

floydWarshall @ 256

3.4×

affinityPropagation @ 256

2.5×

affinityPropagation @ 512

1.84×

hierarchicalClustering @ 1.024k

1.00×

hierarchicalClustering @ 4.096k

0.93×

hierarchicalClustering @ 2.048k

0.92×

pageRankDense @ 2.048k

0.18×

pageRankDense @ 1.024k

0.18×

pageRankDense @ 512

0.13×

pageRank @ 512

0.06×

pageRank @ 1.024k

0.01×

pageRank @ 2.048k

0.00×

0.001×0.01×0.1×1×10×100×1000×

Scaling speedup vs graph size

algorithms-gpu

group	× @ N=256	× @ N=512	× @ N=1.024k	× @ N=2.048k	× @ N=4.096k	× @ N=16.384k	× @ N=65.536k
markovClustering	70×	285×	647×	–	–	–	–
floydWarshall	3.4×	12×	29×	–	–	–	–
affinityPropagation	2.5×	1.84×	3.7×	–	–	–	–
pageRank	–	0.06×	0.01×	0.00×	–	–	–
pageRankDense	–	0.13×	0.18×	0.18×	–	–	–
betweennessCentrality	–	4.0×	9.6×	17×	–	–	–
kMedoids	–	–	20×	53×	121×	–	–
hierarchicalClustering	–	–	1.00×	0.92×	0.93×	–	–
kMeans	–	–	–	–	7.6×	14×	22×
fuzzyCMeans	–	–	–	–	27×	38×	70×

algorithms-gpu N=256 · 4.5 sv4 (gpu) v4 (cpu)

markovClustering

70×

floydWarshall

3.4×

affinityPropagation

2.5×

1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
markovClustering	cpu	615 ms	615 ms	617 ms	611 ms	5
markovClustering	gpu	8.80 ms	9.16 ms	8.90 ms	8.50 ms	5
markovClustering	gpu first call	28.0 ms	28.0 ms	28.0 ms	28.0 ms	1
floydWarshall	cpu	23.4 ms	23.4 ms	23.4 ms	23.2 ms	5
floydWarshall	gpu	6.90 ms	6.94 ms	7.00 ms	6.80 ms	5
floydWarshall	gpu first call	20.1 ms	20.1 ms	20.1 ms	20.1 ms	1
affinityPropagation	cpu	39.5 ms	40.8 ms	40.1 ms	38.0 ms	5
affinityPropagation	gpu	15.5 ms	15.7 ms	15.8 ms	14.7 ms	5
affinityPropagation	gpu first call	27.9 ms	27.9 ms	27.9 ms	27.9 ms	1

algorithms-gpu N=512 · 32.7 sv4 (gpu) v4 (cpu)

markovClustering

285×

pageRank

0.06×

pageRankDense

0.13×

floydWarshall

12×

betweennessCentrality

4.0×

affinityPropagation

1.84×

100 µs1 ms10 ms100 ms1 s10 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
markovClustering	cpu	4.79 s	4.78 s	4.80 s	4.76 s	5
markovClustering	gpu	16.8 ms	17.9 ms	16.8 ms	16.7 ms	5
markovClustering	gpu first call	25.8 ms	25.8 ms	25.8 ms	25.8 ms	1
pageRank	cpu	400 µs	460 µs	400 µs	300 µs	5
pageRank	gpu	6.90 ms	6.76 ms	6.90 ms	6.30 ms	5
pageRank	gpu first call	13.5 ms	13.5 ms	13.5 ms	13.5 ms	1
pageRankDense	cpu	900 µs	900 µs	900 µs	800 µs	5
pageRankDense	gpu	6.80 ms	7.00 ms	7.20 ms	6.50 ms	5
pageRankDense	gpu first call	31.5 ms	31.5 ms	31.5 ms	31.5 ms	1
floydWarshall	cpu	179 ms	218 ms	180 ms	179 ms	5
floydWarshall	gpu	15.5 ms	15.3 ms	15.5 ms	14.6 ms	5
floydWarshall	gpu first call	16.7 ms	16.7 ms	16.7 ms	16.7 ms	1
betweennessCentrality	cpu	36.0 ms	36.1 ms	36.3 ms	35.5 ms	5
betweennessCentrality	gpu	9.00 ms	9.30 ms	9.30 ms	8.70 ms	5
betweennessCentrality	gpu first call	19.3 ms	19.3 ms	19.3 ms	19.3 ms	1
affinityPropagation	cpu	206 ms	206 ms	207 ms	203 ms	5
affinityPropagation	gpu	112 ms	113 ms	113 ms	111 ms	5
affinityPropagation	gpu first call	117 ms	117 ms	117 ms	117 ms	1

algorithms-gpu N=1,024 · 3.4 minv4 (gpu) v4 (cpu)

markovClustering

647×

pageRank

0.01×

pageRankDense

0.18×

floydWarshall

29×

betweennessCentrality

9.6×

affinityPropagation

3.7×

kMedoids

20×

hierarchicalClustering

1.00×

100 µs1 ms10 ms100 ms1 s10 s100 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
markovClustering	cpu	30.3 s	30.3 s	30.3 s	30.2 s	5
markovClustering	gpu	46.8 ms	50.2 ms	46.9 ms	46.3 ms	5
markovClustering	gpu first call	62.1 ms	62.1 ms	62.1 ms	62.1 ms	1
pageRank	cpu	100 µs	120 µs	100 µs	100 µs	5
pageRank	gpu	17.8 ms	18.0 ms	18.2 ms	17.7 ms	5
pageRank	gpu first call	19.8 ms	19.8 ms	19.8 ms	19.8 ms	1
pageRankDense	cpu	3.60 ms	3.62 ms	3.60 ms	3.60 ms	5
pageRankDense	gpu	20.5 ms	21.6 ms	20.9 ms	20.5 ms	5
pageRankDense	gpu first call	96.6 ms	96.6 ms	96.6 ms	96.6 ms	1
floydWarshall	cpu	1.41 s	1.95 s	2.97 s	1.17 s	5
floydWarshall	gpu	48.0 ms	50.8 ms	50.2 ms	38.8 ms	5
floydWarshall	gpu first call	51.5 ms	51.5 ms	51.5 ms	51.5 ms	1
betweennessCentrality	cpu	149 ms	151 ms	151 ms	149 ms	5
betweennessCentrality	gpu	15.6 ms	16.5 ms	15.7 ms	15.0 ms	5
betweennessCentrality	gpu first call	17.1 ms	17.1 ms	17.1 ms	17.1 ms	1
affinityPropagation	cpu	1.11 s	1.54 s	2.18 s	1.08 s	5
affinityPropagation	gpu	297 ms	296 ms	298 ms	292 ms	5
affinityPropagation	gpu first call	296 ms	296 ms	296 ms	296 ms	1
kMedoids	cpu	82.5 ms	74.5 ms	84.6 ms	52.9 ms	5
kMedoids	gpu	4.10 ms	3.82 ms	4.30 ms	3.00 ms	5
kMedoids	gpu first call	12.7 ms	12.7 ms	12.7 ms	12.7 ms	1
hierarchicalClustering	cpu	27.9 ms	30.0 ms	33.8 ms	23.7 ms	5
hierarchicalClustering	gpu	27.8 ms	27.6 ms	28.1 ms	26.4 ms	5
hierarchicalClustering	gpu first call	37.3 ms	37.3 ms	37.3 ms	37.3 ms	1

algorithms-gpu N=2,048 · 9.7 sv4 (gpu) v4 (cpu)

pageRank

0.00×

pageRankDense

0.18×

betweennessCentrality

17×

kMedoids

53×

hierarchicalClustering

0.92×

100 µs1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
pageRank	cpu	300 µs	260 µs	300 µs	200 µs	5
pageRank	gpu	73.2 ms	72.2 ms	73.8 ms	67.6 ms	5
pageRank	gpu first call	87.8 ms	87.8 ms	87.8 ms	87.8 ms	1
pageRankDense	cpu	14.8 ms	15.7 ms	14.8 ms	14.7 ms	5
pageRankDense	gpu	81.8 ms	81.6 ms	82.8 ms	79.8 ms	5
pageRankDense	gpu first call	395 ms	395 ms	395 ms	395 ms	1
betweennessCentrality	cpu	650 ms	658 ms	651 ms	634 ms	5
betweennessCentrality	gpu	37.5 ms	39.2 ms	40.1 ms	36.1 ms	5
betweennessCentrality	gpu first call	43.7 ms	43.7 ms	43.7 ms	43.7 ms	1
kMedoids	cpu	246 ms	260 ms	310 ms	163 ms	5
kMedoids	gpu	4.60 ms	5.88 ms	4.90 ms	3.70 ms	5
kMedoids	gpu first call	6.00 ms	6.00 ms	6.00 ms	6.00 ms	1
hierarchicalClustering	cpu	99.3 ms	99.8 ms	100 ms	98.1 ms	5
hierarchicalClustering	gpu	108 ms	113 ms	113 ms	105 ms	5
hierarchicalClustering	gpu first call	103 ms	103 ms	103 ms	103 ms	1

algorithms-gpu N=4,096 · 15.0 sv4 (gpu) v4 (cpu)

kMeans

7.6×

fuzzyCMeans

27×

kMedoids

121×

hierarchicalClustering

0.93×

1 ms10 ms100 ms1 s10 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
kMeans	cpu	30.5 ms	31.2 ms	30.9 ms	29.9 ms	5
kMeans	gpu	4.00 ms	4.06 ms	4.20 ms	3.70 ms	5
kMeans	gpu first call	14.1 ms	14.1 ms	14.1 ms	14.1 ms	1
fuzzyCMeans	cpu	156 ms	160 ms	161 ms	153 ms	5
fuzzyCMeans	gpu	5.80 ms	5.78 ms	6.00 ms	5.50 ms	5
fuzzyCMeans	gpu first call	18.8 ms	18.8 ms	18.8 ms	18.8 ms	1
kMedoids	cpu	1.46 s	1.37 s	1.47 s	926 ms	5
kMedoids	gpu	12.0 ms	13.3 ms	12.4 ms	9.80 ms	5
kMedoids	gpu first call	12.0 ms	12.0 ms	12.0 ms	12.0 ms	1
hierarchicalClustering	cpu	420 ms	421 ms	423 ms	417 ms	5
hierarchicalClustering	gpu	452 ms	454 ms	457 ms	444 ms	5
hierarchicalClustering	gpu first call	478 ms	478 ms	478 ms	478 ms	1

algorithms-gpu N=16,384 · 4.7 sv4 (gpu) v4 (cpu)

kMeans

14×

fuzzyCMeans

38×

1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
kMeans	cpu	127 ms	128 ms	128 ms	125 ms	5
kMeans	gpu	9.00 ms	9.24 ms	9.10 ms	7.90 ms	5
kMeans	gpu first call	17.7 ms	17.7 ms	17.7 ms	17.7 ms	1
fuzzyCMeans	cpu	616 ms	621 ms	631 ms	606 ms	5
fuzzyCMeans	gpu	16.3 ms	15.4 ms	18.4 ms	11.4 ms	5
fuzzyCMeans	gpu first call	22.8 ms	22.8 ms	22.8 ms	22.8 ms	1

algorithms-gpu N=65,536 · 20.0 sv4 (gpu) v4 (cpu)

kMeans

22×

fuzzyCMeans

70×

10 ms100 ms1 s10 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
kMeans	cpu	613 ms	613 ms	614 ms	606 ms	5
kMeans	gpu	27.7 ms	27.7 ms	27.8 ms	27.4 ms	5
kMeans	gpu first call	78.4 ms	78.4 ms	78.4 ms	78.4 ms	1
fuzzyCMeans	cpu	2.57 s	2.57 s	2.58 s	2.54 s	5
fuzzyCMeans	gpu	36.8 ms	38.3 ms	37.2 ms	36.4 ms	5
fuzzyCMeans	gpu first call	84.8 ms	84.8 ms	84.8 ms	84.8 ms	1