Cytoscape.js GPU benchmarks

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
OS	Fedora Linux 43 (Workstation Edition) · 7.1.10-100.fc43.x86_64 · x64
Node	v24.18.0 (v8 13.6.233.17-node.50)
Commit	9ed49abc docs(113): the performance review recorded — what the sweep measured and what it fixed
Machine id	5cf3f79c

Speedup overview v3 p50 ÷ gpu p50, log scale · right of the 1× line = gpu faster

sweep: edges() @ 2k

283798×

sweep: nodes() @ 2k

169915×

core: edges() @ 2k

93566×

core: nodes() @ 2k

52815×

sweep: elements() @ 2k

29159×

core: elements() @ 2k

20919×

core: graph scratch() set + get (v3 restyles the graph) @ 2k

16146×

core: removeScratch() @ 2k

9890×

core: graph data() set + get (v3 restyles the graph) @ 2k

8660×

compare: same() @ 2k

8169×

trav: successors() @ 2k

735×

trav: predecessors() @ 2k

725×

mut-bulk: hide + show @ 2k

264×

flags: show + hide + visible @ 2k

228×

parent drag (shift subtree + bb settle) @ 2k

183×

compound: commonAncestors @ 2k

121×

degree: nodes.totalDegree() @ 2k

112×

viewport: reset() @ 2k

107×

mut-bulk: shift @ 2k

105×

flags: panify + unpanify + pannable @ 2k

94×

flags: selectify + unselectify + selectable @ 2k

94×

flags: grabify + ungrabify + grabbable @ 2k

94×

flags: lock + unlock + locked @ 2k

93×

degree: nodes.maxDegree() @ 2k

92×

86×

core: options() @ 2k

85×

degree: node.degree() @ 2k

85×

reparent round-trip @ 2k

77×

degree: indegree + outdegree (2 calls) @ 2k

75×

collection: same + anySame + contains (3 calls) @ 2k

71×

algo: closenessCentrality (one root) @ 2k

71×

algo: pageRank (20 iters) @ 500

70×

algo: closenessCentrality (one root) @ 500

68×

mut-bulk: positions(obj) @ 2k

62×

algo: bfs (whole graph) @ 500

61×

dimensions: shift + silentShift (round-trip) @ 2k

60×

sweep: nodes(":selected") vs nodes({selected}) @ 2k

59×

algo: bfs (whole graph) @ 2k

58×

algo: dfs (whole graph) @ 2k

58×

viewport: viewport({ zoom, pan }) @ 2k

57×

algo: dfs (whole graph) @ 500

57×

algo: dijkstra + pathTo @ 2k

57×

mut-bulk: lock + unlock @ 2k

53×

algo: closenessCentrality (weighted, one root) @ 500

52×

algo: dijkstra + pathTo @ 500

51×

dimensions: positions( fn ) over 100 @ 2k

51×

child drag (move + parent re-derive) @ 2k

50×

degree: minDegree + maxDegree (2 calls) @ 2k

44×

collection: component() @ 2k

43×

traverse: components() @ 2k

41×

algo: degreeCentralityNormalized @ 2k

37×

37×

algo: degreeCentralityNormalized @ 500

37×

sweep: selected nodes ($("node:selected") vs filter({group, selected})) @ 2k

36×

algo: tarjan SCC @ 500

35×

query: structural, parents (:parent vs { parent: true }) @ 2k

35×

algo: degreeCentrality (one root) @ 500

35×

algo: tarjan SCC @ 2k

34×

algo: degreeCentrality (one root) @ 2k

34×

bounds: getFitViewport() (compute without committing) @ 2k

33×

bounds: cy.fit() (whole-graph scan) @ 2k

33×

query: structural, orphans (:orphan vs { child: false }) @ 2k

32×

31×

collection: silentPositions( fn ) over 100 @ 2k

30×

collection: intersection @ 2k

29×

collection: difference @ 2k

29×

algo: kruskal @ 2k

29×

scn: select-all + fit (listeners on) @ 2k

27×

load: init from the definition form (construct + dispose) @ 2k

26×

emit: position write, no listeners @ 2k

26×

algo: kruskal @ 500

26×

set: difference() @ 2k

25×

degree: minIndegree + maxOutdegree (2 calls) @ 2k

25×

degree: maxIndegree + minOutdegree (2 calls) @ 2k

25×

set: intersection() @ 2k

25×

algo: hopcroft-tarjan biconnected @ 500

25×

core: id lookup ($("#id") vs $id()) @ 2k

25×

trav: band roots() @ 2k

24×

algo: hopcroft-tarjan biconnected @ 2k

24×

mut-bulk: positions(fn) @ 2k

23×

mut: position set @ 2k

23×

algo: bellmanFord @ 500

22×

data: bulk write, numeric column @ 2k

21×

data: bulk write, object column (plain-array fallback) @ 2k

20×

algo: bellmanFord @ 2k

19×

core: add + remove @ 2k

19×

viewport: center( eles ) @ 2k

19×

19×

data: bulk write, one new dictionary entry per pass @ 2k

17×

data: bulk write, dictionary string column (4 values) @ 2k

17×

flags: select + unselect + selected (band) @ 2k

16×

dimensions: silentPosition @ 2k

16×

scn: drag (select band of 100 + 8-step shift, position listener) @ 2k

16×

16×

flags: active/activate/unactivate (3 calls) @ 2k

16×

query: membership ([bucket in ...] vs { in }) @ 2k

16×

viewport: minZoom/maxZoom get (2 calls) @ 2k

15×

collection: absoluteComplement @ 2k

15×

query: data equality ([bucket = 3] vs { data: { bucket: 3 } }) @ 2k

14×

core: selected ($(":selected") vs filter({selected})) @ 2k

14×

traversal: roots + leaves (2 calls) @ 2k

13×

style: renderedStyle (one prop) @ 2k

13×

query: structural, childless (:childless vs { parent: false }) @ 2k

13×

mut-bulk: data set @ 2k

13×

sweep: selected ($(":selected") vs filter({selected})) @ 2k

13×

algo: betweennessCentrality (unweighted) @ 500

13×

dimensions: padding + paddedWidth + paddedHeight (3 calls) @ 2k

12×

query: two keys AND-ed @ 2k

12×

collection: toArray() @ 2k

12×

dimensions: relativePosition get @ 2k

12×

algo: fuzzyCMeans @ 500

11×

algo: betweennessCentrality (unweighted) @ 2k

11×

query: data comparison ([weight > 3] vs { gt }) @ 2k

11×

core: batch + per-id data() (was batchData, round 90) @ 2k

9.5×

emit: 1 core listener @ 2k

8.9×

collection: removeScratch() @ 2k

8.8×

core: removeAllListeners() @ 2k

8.7×

algo: betweennessCentrality (weighted) @ 500

8.7×

sweep: all nodes ($("node") vs filter({group})) @ 2k

8.6×

algo: hierarchicalClustering (threshold) @ 500

8.3×

emit: delegated (v3 selector vs v4 predicate) @ 2k

8.2×

mut-bulk: select + unselect @ 2k

8.1×

8.1×

scn: explore (expand 2-hop + select + fit) @ 2k

8.0×

emit: 1 element (ref-qualified) listener @ 2k

8.0×

query: predicate function (both sides) @ 2k

7.9×

data: removeData (whole collection) @ 2k

7.8×

algo: kMedoids @ 500

7.7×

element data: scratch get + set @ 2k

7.5×

7.5×

core: collection() @ 2k

7.2×

core: collection() @ 2k

7.2×

compound: orphans + nonorphans (2 calls) @ 2k

7.1×

scn: edit (remove + re-add 256 nodes + cascade, add/remove listeners) @ 2k

6.8×

6.8×

emit: 10 core listeners @ 2k

6.6×

mut-bulk: remove + re-add (256 nodes + cascade) @ 2k

6.6×

bounds: whole-graph boundingBox() @ 2k

6.4×

core: batch( fn ) (empty) @ 2k

6.3×

style: whole-object style() @ 2k

5.9×

core: startBatch + endBatch + batching @ 2k

5.9×

style: sheet swap (compile + applyAll) @ 2k

5.8×

style: bypass set + removeStyle round-trip @ 2k

5.7×

collection: union @ 2k

5.7×

sweep: filter(fn) @ 2k

5.3×

algo: affinityPropagation @ 500

5.3×

set: union() @ 2k

5.2×

traversal: allAreNeighbors @ 2k

5.2×

algo: hierholzer (directed) @ 500

5.2×

traverse: incomers() @ 2k

5.1×

trav: outgoers() @ 2k

5.1×

traverse: outgoers() @ 2k

5.1×

trav: band neighborhood() @ 2k

5.1×

compare: contains() @ 2k

5.0×

trav: incomers() @ 2k

4.9×

core: animated() @ 2k

4.8×

compound: children + descendants + siblings (3 calls) @ 2k

4.8×

core: nodes ($("node") vs filter({group})) @ 2k

4.7×

viewport: zoom() get @ 2k

4.7×

algo: kMeans (fixed centroids) @ 500

4.6×

collection: length + empty + nonempty (3 calls) @ 2k

4.5×

mut: select + unselect @ 2k

4.3×

core: filter(fn) @ 2k

4.2×

style: bypass set (ele.style name,value) @ 2k

4.1×

trav: neighborhood() @ 2k

3.9×

viewport: width() + height() + size() (3 calls) @ 2k

3.8×

collection: symmetricDifference @ 2k

3.8×

traverse: neighborhood() @ 2k

3.7×

compound: parent + parents + ancestors (3 calls) @ 2k

3.6×

algo: aStar @ 500

3.5×

collection: cy() @ 2k

3.5×

traversal: sources + targets (2 calls) @ 2k

3.4×

algo: aStar @ 2k

3.4×

traversal: connectedNodes @ 2k

3.3×

traversal: parallelEdges + codirectedEdges (2 calls) @ 2k

3.3×

collection: layoutDimensions() @ 2k

3.3×

collection: diff @ 2k

3.1×

collection: byGroup @ 2k

3.1×

flags: takesUpSpace + interactive + transparent (3 calls) @ 2k

3.0×

anim: start + stop, 512-node collection @ 2k

2.9×

collection: animated() @ 2k

2.9×

traversal: openNeighborhood + closedNeighborhood (2 calls) @ 2k

2.9×

algo: hierholzer (directed) @ 2k

2.9×

anim: viewport animate + stop @ 2k

2.9×

trav: band connectedNodes() @ 2k

2.8×

trav: closedNeighborhood() @ 2k

2.8×

traverse: connectedEdges() @ 2k

2.8×

collection: delayAnimation() @ 2k

2.8×

core: window() @ 2k

2.8×

dimensions: rendered dims (4 calls) @ 2k

2.8×

style: numericStyle @ 2k

2.7×

core: container() @ 2k

2.7×

load: cy.json() @ 2k

2.6×

mut: data set @ 2k

2.6×

traversal: isLoop + isSimple (2 calls) @ 2k

2.6×

trav: connectedEdges() @ 2k

2.6×

dimensions: width + height + outerWidth + outerHeight (4 calls) @ 2k

2.5×

core: resize() / invalidateSize() @ 2k

2.5×

algo: closenessCentralityNormalized @ 500

2.5×

iter: map() @ 2k

2.4×

core: on('render') + off('render') (was onRender/offRender, round 90) @ 2k

2.4×

collection: is (none match: the full walk) @ 2k

2.3×

flags: effectiveOpacity @ 2k

2.3×

core: zoom set @ 2k

2.3×

dimensions: renderedPosition get @ 2k

2.3×

data: read the whole object @ 2k

2.3×

collection: slice() @ 2k

2.3×

core: destroyed() @ 2k

2.2×

core: isReady() @ 2k

2.2×

flags: grabbed + hidden + inactive (3 calls) @ 2k

2.2×

emit: on() + off() round-trip @ 2k

2.1×

style: read one prop (width) @ 2k

2.1×

iter: filter(fn) @ 2k

2.1×

collection: min + max ( fn ) (2 calls) @ 2k

2.0×

element data: json() one element @ 2k

2.0×

element data: jsons() over 100 @ 2k

2.00×

core: selectionType get @ 2k

1.99×

dimensions: renderedBoundingBox @ 2k

1.92×

style: first apply on add (256-node band) @ 2k

1.89×

json: element @ 2k

1.88×

core: gating setters round-trip (autolock/autoungrabify/autounselectify) @ 2k

1.87×

viewport: zoom( n ) set @ 2k

1.86×

collection: eq/first/last (3 calls) @ 2k

1.83×

scn: refresh (bulk data write + mapped labels + filter + fit, data listener) @ 2k

1.81×

viewport: pan( p ) set @ 2k

1.80×

collection: element accessors (id/group/isNode/isEdge, 4 calls) @ 2k

1.77×

style: read one prop (color) @ 2k

1.76×

algo: floydWarshall @ 500

1.76×

collection: indexOfId() @ 2k

1.75×

style( 'text-max-width' ) (was #74 in the switch) @ 2k

1.73×

style( 'text-wrap' ) (was #73 in the switch) @ 2k

1.72×

viewport: panBy() @ 2k

1.71×

viewport: renderedExtent() @ 2k

1.65×

bounds: one node boundingBox() @ 2k

1.63×

position: get @ 2k

1.61×

load: add a 256-node band (add + remove) @ 2k

1.59×

collection: some + every (2 calls) @ 2k

1.59×

anim: delay() (no channels touched) @ 2k

1.57×

collection: sort( fn ) @ 2k

1.57×

viewport: extent() @ 2k

1.55×

trav: band sources() @ 2k

1.49×

collection: allAre (all match: the full walk) @ 2k

1.48×

compound: isParent/isChild/isChildless/isOrphan (4 calls) @ 2k

1.48×

trav: band connectedEdges() @ 2k

1.47×

viewport: pan() get @ 2k

1.41×

collection: reduce( fn ) @ 2k

1.38×

collection: indexOf @ 2k

1.35×

style( 'target-distance-from-node' ) (was #150 in the switch) @ 2k

1.35×

data: read one string key (dictionary decode) @ 2k

1.34×

anim: start + stop, one element @ 2k

1.31×

core: pan() get (x32) @ 2k

1.30×

traversal: source() + target() (2 calls) @ 2k

1.28×

traversal: edgesTo @ 2k

1.28×

data: read one numeric key @ 2k

1.27×

style( 'taxi-radius' ) (was #142 in the switch) @ 2k

1.25×

traversal: edgesWith @ 2k

1.25×

core: instanceString + headless + styleEnabled (3 calls) @ 2k

1.25×

core: hasElementWithId + hasCompoundNodes (2 calls) @ 2k

1.18×

collection: removed() + inside() (2 calls) @ 2k

1.17×

core: gating getters (5 calls) @ 2k

1.16×

style( 'background-color' ) (was #4 in the switch) @ 2k

1.13×

core: extent() @ 2k

1.13×

style( 'border-width' ) (was #6 in the switch) @ 2k

1.11×

algo: markovClustering @ 500

1.10×

iter: forEach() @ 2k

1.06×

algo: kargerStein @ 500

1.05×

core: getElementById() @ 2k

1.04×

core: elements() memo hit (was mutableElements, round 90) @ 2k

1.02×

data: get @ 2k

1.02×

1×10×100×1000×10000×100000×1000000×

Scaling speedup vs graph size

algorithms

group	× @ N=500	× @ N=2k
algo: bfs (whole graph)	61×	58×
algo: dfs (whole graph)	57×	58×
algo: dijkstra + pathTo	51×	57×
algo: aStar	3.5×	3.4×
algo: bellmanFord	22×	19×
algo: kruskal	26×	29×
algo: tarjan SCC	35×	34×
algo: hopcroft-tarjan biconnected	25×	24×
algo: hierholzer (directed)	5.2×	2.9×
algo: betweennessCentrality (unweighted)	13×	11×
algo: degreeCentralityNormalized	37×	37×
algo: closenessCentrality (one root)	68×	71×
algo: degreeCentrality (one root)	35×	34×
algo: pageRank (20 iters)	70×	–
algo: closenessCentrality (weighted, one root)	52×	–
algo: floydWarshall	1.76×	–
algo: closenessCentralityNormalized	2.5×	–
algo: markovClustering	1.10×	–
algo: hierarchicalClustering (threshold)	8.3×	–
algo: kMeans (fixed centroids)	4.6×	–
algo: kMedoids	7.7×	–
algo: fuzzyCMeans	11×	–
algo: affinityPropagation	5.3×	–
algo: kargerStein	1.05×	–
algo: betweennessCentrality (weighted)	8.7×	–

core+collection N=2,000 · 2.9 minv4 (gpu) v3

core: getElementById()

1.04×

core: id lookup ($("#id") vs $id())

25×

core: nodes()

52815×

core: edges()

93566×

core: elements()

20919×

core: nodes ($("node") vs filter({group}))

4.7×

core: selected ($(":selected") vs filter({selected}))

14×

core: filter(fn)

4.2×

core: collection()

7.2×

core: pan() get (x32)

1.30×

core: extent()

1.13×

core: add + remove

19×

core: zoom set

2.3×

traverse: neighborhood()

3.7×

traverse: connectedEdges()

2.8×

traverse: outgoers()

5.1×

traverse: incomers()

5.1×

traverse: components()

41×

degree: node.degree()

85×

degree: nodes.totalDegree()

112×

degree: nodes.maxDegree()

92×

data: get

1.02×

position: get

1.61×

json: element

1.88×

set: union()

5.2×

set: intersection()

25×

set: difference()

25×

iter: map()

2.4×

iter: forEach()

1.06×

iter: filter(fn)

2.1×

compare: same()

8169×

compare: contains()

5.0×

mut: data set

2.6×

mut: position set

23×

mut: select + unselect

4.3×

1 ns10 ns100 ns1 µs10 µs100 µs1 ms10 ms100 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
core: getElementById()	v3	93.4 ns	97.4 ns	167 ns	90.9 ns	1761
core: getElementById()	gpu	90.1 ns	93.9 ns	157 ns	87.8 ns	1823
core: id lookup ($("#id") vs $id())	v3	2.09 µs	2.46 µs	3.96 µs	2.00 µs	64
core: id lookup ($("#id") vs $id())	gpu	84.6 ns	87.3 ns	159 ns	81.5 ns	1958
core: nodes()	v3	418 µs	426 µs	556 µs	409 µs	1643
core: nodes()	gpu	7.92 ns	8.42 ns	17.6 ns	7.46 ns	20201
core: edges()	v3	723 µs	742 µs	922 µs	701 µs	942
core: edges()	gpu	7.73 ns	8.25 ns	16.6 ns	7.51 ns	20620
core: elements()	v3	145 µs	152 µs	264 µs	141 µs	4596
core: elements()	gpu	6.94 ns	7.28 ns	23.8 ns	6.49 ns	23343
core: nodes ($("node") vs filter({group}))	v3	539 µs	575 µs	1.25 ms	531 µs	1212
core: nodes ($("node") vs filter({group}))	gpu	114 µs	119 µs	260 µs	84.9 µs	5761
core: selected ($(":selected") vs filter({selected}))	v3	349 µs	370 µs	651 µs	334 µs	1880
core: selected ($(":selected") vs filter({selected}))	gpu	25.8 µs	26.5 µs	26.8 µs	25.1 µs	12
core: filter(fn)	v3	421 µs	434 µs	561 µs	409 µs	1609
core: filter(fn)	gpu	99.9 µs	106 µs	348 µs	96.5 µs	6557
core: collection()	v3	459 ns	578 ns	1.83 µs	447 ns	290
core: collection()	gpu	63.6 ns	67.8 ns	180 ns	60.9 ns	2468
core: pan() get (x32)	v3	44.5 ns	46.3 ns	58.0 ns	42.4 ns	3641
core: pan() get (x32)	gpu	34.3 ns	36.3 ns	77.4 ns	32.9 ns	4714
core: extent()	v3	22.5 ns	40.6 ns	163 ns	21.4 ns	4157
core: extent()	gpu	20.0 ns	28.0 ns	111 ns	18.6 ns	5960
core: add + remove	v3	127 µs	132 µs	311 µs	117 µs	5256
core: add + remove	gpu	6.57 µs	7.57 µs	25.9 µs	6.19 µs	85124
core: zoom set	v3	1.35 µs	1.37 µs	1.57 µs	1.24 µs	120
core: zoom set	gpu	585 ns	602 ns	709 ns	576 ns	273
traverse: neighborhood()	v3	3.58 µs	3.81 µs	4.51 µs	3.41 µs	40
traverse: neighborhood()	gpu	967 ns	1.14 µs	2.23 µs	941 ns	142
traverse: connectedEdges()	v3	1.07 µs	1.21 µs	2.04 µs	1.04 µs	135
traverse: connectedEdges()	gpu	383 ns	412 ns	1.36 µs	381 ns	408
traverse: outgoers()	v3	2.41 µs	2.62 µs	3.46 µs	2.30 µs	61
traverse: outgoers()	gpu	473 ns	551 ns	2.03 µs	435 ns	294
traverse: incomers()	v3	2.37 µs	2.57 µs	3.36 µs	2.26 µs	62
traverse: incomers()	gpu	461 ns	576 ns	2.08 µs	444 ns	281
traverse: components()	v3	79.9 ms	82.2 ms	88.8 ms	76.4 ms	12
traverse: components()	gpu	1.93 ms	2.05 ms	3.55 ms	1.87 ms	323
degree: node.degree()	v3	3.50 µs	3.93 µs	5.08 µs	3.36 µs	40
degree: node.degree()	gpu	41.3 ns	43.0 ns	127 ns	38.9 ns	3970
degree: nodes.totalDegree()	v3	7.24 ms	8.01 ms	11.4 ms	6.72 ms	83
degree: nodes.totalDegree()	gpu	64.9 µs	66.6 µs	105 µs	60.8 µs	10487
degree: nodes.maxDegree()	v3	7.05 ms	7.91 ms	12.2 ms	6.77 ms	81
degree: nodes.maxDegree()	gpu	76.4 µs	80.2 µs	134 µs	71.2 µs	8452
data: get	v3	30.3 ns	32.0 ns	62.0 ns	28.8 ns	5336
data: get	gpu	29.8 ns	31.3 ns	49.5 ns	28.8 ns	5469
position: get	v3	35.2 ns	36.6 ns	75.3 ns	33.9 ns	4674
position: get	gpu	21.9 ns	25.3 ns	49.9 ns	21.0 ns	6584
json: element	v3	300 ns	394 ns	3.54 µs	293 ns	423
json: element	gpu	160 ns	165 ns	255 ns	153 ns	985
set: union()	v3	94.8 µs	106 µs	164 µs	91.9 µs	6388
set: union()	gpu	18.1 µs	18.7 µs	19.2 µs	17.4 µs	12
set: intersection()	v3	89.5 µs	102 µs	156 µs	86.5 µs	6637
set: intersection()	gpu	3.59 µs	3.62 µs	3.98 µs	3.53 µs	42
set: difference()	v3	90.0 µs	100 µs	142 µs	87.0 µs	6735
set: difference()	gpu	3.55 µs	3.58 µs	3.85 µs	3.47 µs	42
iter: map()	v3	27.4 µs	28.9 µs	75.0 µs	26.4 µs	23841
iter: map()	gpu	11.5 µs	11.6 µs	11.8 µs	11.4 µs	11
iter: forEach()	v3	3.51 µs	3.78 µs	3.75 µs	3.37 µs	36
iter: forEach()	gpu	3.32 µs	3.35 µs	3.53 µs	3.25 µs	45
iter: filter(fn)	v3	242 µs	248 µs	316 µs	237 µs	2822
iter: filter(fn)	gpu	115 µs	124 µs	199 µs	109 µs	5600
compare: same()	v3	121 µs	124 µs	177 µs	115 µs	5664
compare: same()	gpu	14.8 ns	15.4 ns	24.9 ns	14.2 ns	11082
compare: contains()	v3	6.22 µs	6.24 µs	6.47 µs	6.08 µs	24
compare: contains()	gpu	1.24 µs	1.26 µs	1.35 µs	1.20 µs	130
mut: data set	v3	617 ns	636 ns	1.05 µs	590 ns	264
mut: data set	gpu	234 ns	246 ns	430 ns	228 ns	690
mut: position set	v3	1.20 µs	1.21 µs	1.34 µs	1.16 µs	136
mut: position set	gpu	52.3 ns	55.2 ns	90.2 ns	50.9 ns	3071
mut: select + unselect	v3	2.53 µs	2.71 µs	3.31 µs	2.42 µs	58
mut: select + unselect	gpu	596 ns	603 ns	691 ns	574 ns	271

materializers N=2,000 · 49.8 sv4 (gpu) v3

sweep: nodes()

169915×

sweep: edges()

283798×

sweep: elements()

29159×

sweep: all nodes ($("node") vs filter({group}))

8.6×

sweep: selected ($(":selected") vs filter({selected}))

13×

sweep: selected nodes ($("node:selected") vs filter({group, selected}))

36×

sweep: nodes(":selected") vs nodes({selected})

59×

sweep: filter(fn)

5.3×

1 ns10 ns100 ns1 µs10 µs100 µs1 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
sweep: nodes()	v3	410 µs	419 µs	496 µs	397 µs	1673
sweep: nodes()	gpu	2.41 ns	2.48 ns	2.95 ns	2.26 ns	66581
sweep: edges()	v3	714 µs	726 µs	830 µs	679 µs	962
sweep: edges()	gpu	2.52 ns	2.62 ns	4.76 ns	2.46 ns	63240
sweep: elements()	v3	139 µs	147 µs	248 µs	135 µs	4727
sweep: elements()	gpu	4.76 ns	4.89 ns	8.99 ns	4.55 ns	34553
sweep: all nodes ($("node") vs filter({group}))	v3	546 µs	569 µs	801 µs	528 µs	1228
sweep: all nodes ($("node") vs filter({group}))	gpu	63.2 µs	76.4 µs	177 µs	59.3 µs	9071
sweep: selected ($(":selected") vs filter({selected}))	v3	339 µs	354 µs	605 µs	329 µs	1977
sweep: selected ($(":selected") vs filter({selected}))	gpu	25.7 µs	26.1 µs	26.2 µs	25.5 µs	12
sweep: selected nodes ($("node:selected") vs filter({group, selected}))	v3	325 µs	334 µs	491 µs	315 µs	2096
sweep: selected nodes ($("node:selected") vs filter({group, selected}))	gpu	9.07 µs	9.18 µs	9.47 µs	9.04 µs	15
sweep: nodes(":selected") vs nodes({selected})	v3	532 µs	541 µs	776 µs	510 µs	1292
sweep: nodes(":selected") vs nodes({selected})	gpu	8.95 µs	8.99 µs	9.11 µs	8.88 µs	15
sweep: filter(fn)	v3	407 µs	417 µs	538 µs	396 µs	1680
sweep: filter(fn)	gpu	76.3 µs	79.1 µs	129 µs	74.5 µs	8830

mutators N=2,000 · 56.1 sv4 (gpu) v3

mut-bulk: select + unselect

8.1×

mut-bulk: hide + show

264×

mut-bulk: lock + unlock

53×

mut-bulk: positions(obj)

62×

mut-bulk: positions(fn)

23×

mut-bulk: shift

105×

mut-bulk: data set

13×

mut-bulk: remove + re-add (256 nodes + cascade)

6.6×

10 µs100 µs1 ms10 ms100 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
mut-bulk: select + unselect	v3	2.12 ms	2.19 ms	3.49 ms	2.05 ms	303
mut-bulk: select + unselect	gpu	261 µs	270 µs	397 µs	253 µs	2594
mut-bulk: hide + show	v3	56.4 ms	61.5 ms	72.6 ms	55.3 ms	12
mut-bulk: hide + show	gpu	214 µs	219 µs	361 µs	209 µs	3088
mut-bulk: lock + unlock	v3	2.04 ms	2.10 ms	3.44 ms	1.95 ms	329
mut-bulk: lock + unlock	gpu	38.3 µs	38.7 µs	39.1 µs	37.8 µs	12
mut-bulk: positions(obj)	v3	2.28 ms	2.34 ms	3.81 ms	2.21 ms	294
mut-bulk: positions(obj)	gpu	36.8 µs	38.4 µs	67.1 µs	34.7 µs	18007
mut-bulk: positions(fn)	v3	3.36 ms	3.53 ms	5.60 ms	3.26 ms	194
mut-bulk: positions(fn)	gpu	146 µs	149 µs	201 µs	141 µs	4616
mut-bulk: shift	v3	3.40 ms	3.52 ms	4.96 ms	3.27 ms	188
mut-bulk: shift	gpu	32.3 µs	33.1 µs	55.1 µs	30.8 µs	20547
mut-bulk: data set	v3	871 µs	902 µs	1.54 ms	846 µs	770
mut-bulk: data set	gpu	65.7 µs	67.6 µs	118 µs	62.9 µs	10161
mut-bulk: remove + re-add (256 nodes + cascade)	v3	41.6 ms	45.6 ms	53.3 ms	40.4 ms	11
mut-bulk: remove + re-add (256 nodes + cascade)	gpu	6.35 ms	6.45 ms	7.51 ms	5.65 ms	102

scenarios N=2,000 · 37.9 sv4 (gpu) v3

scn: explore (expand 2-hop + select + fit)

8.0×

scn: select-all + fit (listeners on)

27×

scn: drag (select band of 100 + 8-step shift, position listener)

16×

scn: edit (remove + re-add 256 nodes + cascade, add/remove listeners)

6.8×

scn: refresh (bulk data write + mapped labels + filter + fit, data listener)

1.81×

10 µs100 µs1 ms10 ms100 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
scn: explore (expand 2-hop + select + fit)	v3	297 µs	349 µs	981 µs	281 µs	1991
scn: explore (expand 2-hop + select + fit)	gpu	37.3 µs	43.2 µs	210 µs	33.7 µs	15793
scn: select-all + fit (listeners on)	v3	25.1 ms	25.2 ms	27.0 ms	23.3 ms	23
scn: select-all + fit (listeners on)	gpu	931 µs	977 µs	1.39 ms	892 µs	688
scn: drag (select band of 100 + 8-step shift, position listener)	v3	1.65 ms	1.69 ms	2.17 ms	1.60 ms	397
scn: drag (select band of 100 + 8-step shift, position listener)	gpu	102 µs	109 µs	348 µs	97.5 µs	6361
scn: edit (remove + re-add 256 nodes + cascade, add/remove listeners)	v3	44.6 ms	44.5 ms	46.5 ms	42.8 ms	11
scn: edit (remove + re-add 256 nodes + cascade, add/remove listeners)	gpu	6.53 ms	6.78 ms	8.84 ms	5.96 ms	97
scn: refresh (bulk data write + mapped labels + filter + fit, data listener)	v3	7.50 ms	7.56 ms	9.27 ms	6.72 ms	84
scn: refresh (bulk data write + mapped labels + filter + fit, data listener)	gpu	4.13 ms	4.26 ms	7.01 ms	3.70 ms	152

traversal N=2,000 · 2.0 minv4 (gpu) v3

trav: neighborhood()

3.9×

trav: closedNeighborhood()

2.8×

trav: connectedEdges()

2.6×

trav: outgoers()

5.1×

trav: incomers()

4.9×

trav: successors()

735×

trav: predecessors()

725×

trav: band neighborhood()

5.1×

trav: band connectedEdges()

1.47×

trav: band roots()

24×

trav: band sources()

1.49×

trav: band connectedNodes()

2.8×

100 ns1 µs10 µs100 µs1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
trav: neighborhood()	v3	3.94 µs	4.05 µs	5.30 µs	3.46 µs	38
trav: neighborhood()	gpu	1.01 µs	1.06 µs	2.00 µs	825 ns	156
trav: closedNeighborhood()	v3	5.83 µs	5.68 µs	6.16 µs	5.17 µs	26
trav: closedNeighborhood()	gpu	2.09 µs	2.24 µs	2.84 µs	2.03 µs	72
trav: connectedEdges()	v3	991 ns	1.10 µs	1.91 µs	939 ns	151
trav: connectedEdges()	gpu	388 ns	404 ns	663 ns	378 ns	419
trav: outgoers()	v3	2.26 µs	2.42 µs	3.03 µs	2.18 µs	66
trav: outgoers()	gpu	441 ns	522 ns	2.01 µs	438 ns	318
trav: incomers()	v3	2.29 µs	2.43 µs	2.95 µs	2.23 µs	66
trav: incomers()	gpu	466 ns	544 ns	2.02 µs	441 ns	303
trav: successors()	v3	432 ms	438 ms	452 ms	429 ms	12
trav: successors()	gpu	587 µs	617 µs	1.75 ms	574 µs	1103
trav: predecessors()	v3	434 ms	435 ms	443 ms	428 ms	12
trav: predecessors()	gpu	598 µs	625 µs	1.82 ms	576 µs	1090
trav: band neighborhood()	v3	187 µs	201 µs	308 µs	179 µs	3473
trav: band neighborhood()	gpu	37.0 µs	37.4 µs	38.7 µs	36.0 µs	12
trav: band connectedEdges()	v3	24.5 µs	24.7 µs	24.9 µs	23.3 µs	12
trav: band connectedEdges()	gpu	16.7 µs	17.1 µs	19.6 µs	16.0 µs	12
trav: band roots()	v3	109 µs	118 µs	175 µs	105 µs	5870
trav: band roots()	gpu	4.49 µs	4.50 µs	4.64 µs	4.41 µs	34
trav: band sources()	v3	6.51 µs	6.52 µs	6.77 µs	6.19 µs	22
trav: band sources()	gpu	4.38 µs	4.43 µs	4.81 µs	4.27 µs	34
trav: band connectedNodes()	v3	14.1 µs	14.7 µs	15.5 µs	13.9 µs	12
trav: band connectedNodes()	gpu	4.97 µs	5.03 µs	5.34 µs	4.87 µs	30

algorithms N=2,000 · 4.6 minv4 (gpu) v3

algo: bfs (whole graph)

58×

algo: dfs (whole graph)

58×

algo: dijkstra + pathTo

57×

algo: aStar

3.4×

algo: bellmanFord

19×

algo: kruskal

29×

algo: tarjan SCC

34×

algo: hopcroft-tarjan biconnected

24×

algo: hierholzer (directed)

2.9×

algo: betweennessCentrality (unweighted)

11×

algo: degreeCentralityNormalized

37×

algo: closenessCentrality (one root)

71×

algo: degreeCentrality (one root)

34×

100 ns1 µs10 µs100 µs1 ms10 ms100 ms1 s10 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
algo: bfs (whole graph)	v3	51.4 ms	53.5 ms	59.5 ms	49.7 ms	9
algo: bfs (whole graph)	gpu	879 µs	908 µs	1.36 ms	868 µs	766
algo: dfs (whole graph)	v3	48.8 ms	49.0 ms	49.9 ms	48.3 ms	10
algo: dfs (whole graph)	gpu	836 µs	850 µs	1.16 ms	800 µs	817
algo: dijkstra + pathTo	v3	52.1 ms	52.3 ms	53.0 ms	51.6 ms	10
algo: dijkstra + pathTo	gpu	920 µs	934 µs	1.16 ms	888 µs	745
algo: aStar	v3	1.64 ms	1.68 ms	2.50 ms	1.57 ms	411
algo: aStar	gpu	483 µs	495 µs	639 µs	466 µs	1412
algo: bellmanFord	v3	227 ms	225 ms	248 ms	205 ms	12
algo: bellmanFord	gpu	11.7 ms	11.8 ms	12.3 ms	11.6 ms	56
algo: kruskal	v3	13.2 ms	14.1 ms	17.2 ms	12.8 ms	45
algo: kruskal	gpu	461 µs	482 µs	953 µs	452 µs	1448
algo: tarjan SCC	v3	22.9 ms	24.1 ms	30.0 ms	21.1 ms	25
algo: tarjan SCC	gpu	667 µs	791 µs	5.37 ms	598 µs	875
algo: hopcroft-tarjan biconnected	v3	84.8 ms	86.0 ms	87.9 ms	81.2 ms	12
algo: hopcroft-tarjan biconnected	gpu	3.53 ms	3.67 ms	4.81 ms	3.45 ms	185
algo: hierholzer (directed)	v3	22.3 ms	23.2 ms	25.6 ms	21.5 ms	26
algo: hierholzer (directed)	gpu	7.76 ms	7.84 ms	9.20 ms	7.46 ms	80
algo: betweennessCentrality (unweighted)	v3	4.07 s	4.08 s	4.12 s	4.04 s	12
algo: betweennessCentrality (unweighted)	gpu	358 ms	371 ms	433 ms	338 ms	12
algo: degreeCentralityNormalized	v3	14.6 ms	15.2 ms	18.1 ms	13.5 ms	42
algo: degreeCentralityNormalized	gpu	392 µs	400 µs	555 µs	380 µs	1708
algo: closenessCentrality (one root)	v3	59.8 ms	61.6 ms	65.9 ms	57.3 ms	12
algo: closenessCentrality (one root)	gpu	842 µs	855 µs	1.03 ms	815 µs	796
algo: degreeCentrality (one root)	v3	7.35 µs	7.12 µs	7.59 µs	6.55 µs	20
algo: degreeCentrality (one root)	gpu	218 ns	223 ns	299 ns	209 ns	759

algorithms N=500 · 10.2 minv4 (gpu) v3

algo: bfs (whole graph)

61×

algo: dfs (whole graph)

57×

algo: dijkstra + pathTo

51×

algo: aStar

3.5×

algo: bellmanFord

22×

algo: kruskal

26×

algo: tarjan SCC

35×

algo: hopcroft-tarjan biconnected

25×

algo: hierholzer (directed)

5.2×

algo: betweennessCentrality (unweighted)

13×

algo: degreeCentralityNormalized

37×

algo: closenessCentrality (one root)

68×

algo: degreeCentrality (one root)

35×

algo: pageRank (20 iters)

70×

algo: closenessCentrality (weighted, one root)

52×

algo: floydWarshall

1.76×

algo: closenessCentralityNormalized

2.5×

algo: markovClustering

1.10×

algo: hierarchicalClustering (threshold)

8.3×

algo: kMeans (fixed centroids)

4.6×

algo: kMedoids

7.7×

algo: fuzzyCMeans

11×

algo: affinityPropagation

5.3×

algo: kargerStein

1.05×

algo: betweennessCentrality (weighted)

8.7×

100 ns1 µs10 µs100 µs1 ms10 ms100 ms1 s10 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
algo: bfs (whole graph)	v3	12.7 ms	13.2 ms	18.1 ms	10.7 ms	49
algo: bfs (whole graph)	gpu	208 µs	213 µs	268 µs	204 µs	3286
algo: dfs (whole graph)	v3	11.7 ms	12.6 ms	15.5 ms	11.4 ms	52
algo: dfs (whole graph)	gpu	204 µs	209 µs	306 µs	200 µs	3343
algo: dijkstra + pathTo	v3	12.1 ms	12.7 ms	16.3 ms	11.4 ms	51
algo: dijkstra + pathTo	gpu	234 µs	241 µs	409 µs	228 µs	2888
algo: aStar	v3	442 µs	450 µs	535 µs	427 µs	1556
algo: aStar	gpu	125 µs	130 µs	202 µs	120 µs	5393
algo: bellmanFord	v3	17.7 ms	17.8 ms	18.5 ms	17.3 ms	36
algo: bellmanFord	gpu	810 µs	837 µs	1.90 ms	784 µs	830
algo: kruskal	v3	3.22 ms	3.50 ms	7.04 ms	3.14 ms	196
algo: kruskal	gpu	126 µs	131 µs	198 µs	121 µs	5338
algo: tarjan SCC	v3	5.35 ms	5.71 ms	8.86 ms	5.21 ms	118
algo: tarjan SCC	gpu	152 µs	159 µs	253 µs	147 µs	4408
algo: hopcroft-tarjan biconnected	v3	20.4 ms	20.9 ms	24.3 ms	18.9 ms	30
algo: hopcroft-tarjan biconnected	gpu	825 µs	856 µs	1.29 ms	803 µs	814
algo: hierholzer (directed)	v3	4.27 ms	4.55 ms	7.52 ms	4.09 ms	144
algo: hierholzer (directed)	gpu	819 µs	845 µs	1.22 ms	795 µs	806
algo: betweennessCentrality (unweighted)	v3	260 ms	262 ms	266 ms	257 ms	12
algo: betweennessCentrality (unweighted)	gpu	20.2 ms	20.2 ms	20.6 ms	19.8 ms	30
algo: degreeCentralityNormalized	v3	3.48 ms	3.80 ms	7.26 ms	3.39 ms	180
algo: degreeCentralityNormalized	gpu	94.5 µs	97.1 µs	145 µs	92.9 µs	7192
algo: closenessCentrality (one root)	v3	13.7 ms	14.3 ms	17.0 ms	13.0 ms	45
algo: closenessCentrality (one root)	gpu	201 µs	205 µs	260 µs	193 µs	3414
algo: degreeCentrality (one root)	v3	7.39 µs	7.28 µs	7.61 µs	6.58 µs	20
algo: degreeCentrality (one root)	gpu	214 ns	223 ns	345 ns	204 ns	763
algo: pageRank (20 iters)	v3	2.75 ms	2.90 ms	6.07 ms	2.67 ms	237
algo: pageRank (20 iters)	gpu	39.3 µs	44.1 µs	144 µs	38.4 µs	15701
algo: closenessCentrality (weighted, one root)	v3	13.4 ms	14.4 ms	18.6 ms	13.1 ms	44
algo: closenessCentrality (weighted, one root)	gpu	256 µs	266 µs	508 µs	244 µs	2635
algo: floydWarshall	v3	245 ms	273 ms	387 ms	242 ms	12
algo: floydWarshall	gpu	140 ms	141 ms	141 ms	138 ms	12
algo: closenessCentralityNormalized	v3	273 ms	273 ms	275 ms	268 ms	12
algo: closenessCentralityNormalized	gpu	111 ms	117 ms	141 ms	110 ms	12
algo: markovClustering	v3	4.10 s	4.35 s	4.63 s	4.07 s	12
algo: markovClustering	gpu	3.73 s	3.74 s	3.76 s	3.72 s	12
algo: hierarchicalClustering (threshold)	v3	34.4 ms	34.5 ms	35.6 ms	33.9 ms	16
algo: hierarchicalClustering (threshold)	gpu	4.14 ms	4.25 ms	5.61 ms	4.03 ms	160
algo: kMeans (fixed centroids)	v3	9.01 ms	9.04 ms	9.70 ms	8.60 ms	74
algo: kMeans (fixed centroids)	gpu	1.98 ms	2.03 ms	3.12 ms	1.90 ms	342
algo: kMedoids	v3	301 ms	294 ms	455 ms	110 ms	12
algo: kMedoids	gpu	39.3 ms	35.8 ms	45.8 ms	13.0 ms	17
algo: fuzzyCMeans	v3	78.8 ms	80.2 ms	82.8 ms	78.1 ms	12
algo: fuzzyCMeans	gpu	6.93 ms	6.99 ms	7.74 ms	6.71 ms	96
algo: affinityPropagation	v3	208 ms	210 ms	217 ms	198 ms	12
algo: affinityPropagation	gpu	39.1 ms	39.6 ms	41.7 ms	38.7 ms	14
algo: kargerStein	v3	322 ms	322 ms	330 ms	312 ms	12
algo: kargerStein	gpu	306 ms	307 ms	312 ms	298 ms	12
algo: betweennessCentrality (weighted)	v3	705 ms	706 ms	714 ms	694 ms	12
algo: betweennessCentrality (weighted)	gpu	81.1 ms	82.9 ms	90.5 ms	77.9 ms	12

mappers N=2,000 · 23.9 s

mapper: bulk write, color — cpu eval

1.31 ms

mapper: bulk write, color — gpu path

175 µs

mapper: bulk write, color + size — cpu eval

1.57 ms

mapper: bulk write, color + size — gpu color / cpu size

1.47 ms

mapper: bulk write, mapped labels (round-5 path) — label refresh

3.97 ms

100 µs1 ms10 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
mapper: bulk write, color	cpu eval	1.31 ms	1.35 ms	1.98 ms	1.25 ms	516
mapper: bulk write, color	gpu path	175 µs	184 µs	328 µs	169 µs	3811
mapper: bulk write, color + size	cpu eval	1.57 ms	1.60 ms	1.74 ms	1.53 ms	429
mapper: bulk write, color + size	gpu color / cpu size	1.47 ms	1.50 ms	1.70 ms	1.46 ms	464
mapper: bulk write, mapped labels (round-5 path)	label refresh	3.97 ms	4.20 ms	6.70 ms	3.39 ms	161

layouts N=2,000 · 2.1 minv4 (gpu) v3

86×

19×

37×

6.8×

16×

31×

8.1×

2.49 ms

7.76 ms

165 ms

10.3 ms

115 ms

110 ms

109 ms

layout: force seed — spectral vs scatter (20 iterations) — spectral (landmark MDS, the default)

110 ms

57.0 ms

457 µs

32.0 µs

544 ns

layout: contract fixed cost (empty impl vs bulk placement) — wrapper + bulk write

39.8 µs

7.5×

100 ns1 µs10 µs100 µs1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
layout: grid	v3	11.0 ms	11.7 ms	16.4 ms	9.86 ms	54
layout: grid	gpu	127 µs	133 µs	247 µs	80.5 µs	4911
layout: preset	v3	5.50 ms	5.74 ms	8.49 ms	5.35 ms	111
layout: preset	gpu	296 µs	316 µs	660 µs	288 µs	2117
layout: preset (fn form)	v3	5.43 ms	5.73 ms	9.01 ms	5.15 ms	117
layout: preset (fn form)	gpu	145 µs	161 µs	502 µs	139 µs	4297
layout: circle	v3	9.43 ms	10.0 ms	15.1 ms	9.01 ms	65
layout: circle	gpu	1.39 ms	1.50 ms	3.54 ms	1.30 ms	453
layout: concentric	v3	32.1 ms	32.5 ms	36.6 ms	28.4 ms	17
layout: concentric	gpu	1.98 ms	2.21 ms	4.52 ms	1.94 ms	306
layout: breadthfirst	v3	392 ms	405 ms	448 ms	386 ms	12
layout: breadthfirst	gpu	12.5 ms	13.1 ms	16.0 ms	12.0 ms	47
layout: random	v3	5.34 ms	5.73 ms	9.57 ms	5.13 ms	116
layout: random	gpu	663 µs	716 µs	1.53 ms	643 µs	965
layout: radial vs breadthfirst-circle (unbalanced tree)	radial	2.49 ms	2.85 ms	4.98 ms	2.35 ms	236
layout: radial vs breadthfirst-circle (unbalanced tree)	breadthfirst circle	7.76 ms	8.16 ms	16.7 ms	7.53 ms	79
layout: flow vs breadthfirst (staged DAG)	flow	165 ms	179 ms	207 ms	162 ms	12
layout: flow vs breadthfirst (staged DAG)	breadthfirst	10.3 ms	10.7 ms	18.5 ms	9.73 ms	60
layout: force (CPU executor, 20 iterations)	gpu	115 ms	123 ms	135 ms	105 ms	12
layout: force constrained vs plain (CPU, 20 iterations)	constrained	110 ms	112 ms	119 ms	103 ms	12
layout: force constrained vs plain (CPU, 20 iterations)	plain	109 ms	111 ms	119 ms	102 ms	12
layout: force seed — spectral vs scatter (20 iterations)	spectral (landmark MDS, the default)	110 ms	113 ms	123 ms	103 ms	12
layout: force seed — spectral vs scatter (20 iterations)	scatter	57.0 ms	56.9 ms	57.3 ms	56.1 ms	9
layout: plumbing (layoutPositions vs setPositions)	finisher	457 µs	493 µs	994 µs	443 µs	1407
layout: plumbing (layoutPositions vs setPositions)	bulk	32.0 µs	41.1 µs	251 µs	29.9 µs	16081
layout: contract fixed cost (empty impl vs bulk placement)	wrapper only	544 ns	973 ns	3.88 µs	527 ns	166
layout: contract fixed cost (empty impl vs bulk placement)	wrapper + bulk write	39.8 µs	40.7 µs	41.8 µs	38.3 µs	12
layout: scope (eles.layout, 10% of nodes)	v3	971 µs	1.05 ms	2.72 ms	946 µs	586
layout: scope (eles.layout, 10% of nodes)	gpu	130 µs	143 µs	266 µs	125 µs	4767

style N=2,000 · 1.7 minv4 (gpu) v3

style: sheet swap (compile + applyAll)

5.8×

style: compile vs compile + apply — compile only (in batch)

24.6 µs

style: compile vs compile + apply — compile + apply

2.25 ms

style: first apply on add (256-node band)

1.89×

style: applyAll, flat vs compound (100 parents — the 14.6 partition) — flat

2.27 ms

style: applyAll, flat vs compound (100 parents — the 14.6 partition) — compound

2.74 ms

style: applyAll — constant vs state-case (the 57.1d partition) — constant sheet

2.26 ms

style: applyAll — constant vs state-case (the 57.1d partition) — state-case sheet

2.27 ms

style: applyAll — state-case vs data-case (what the partition avoids) — state-case sheet

2.29 ms

style: applyAll — state-case vs data-case (what the partition avoids) — data-case sheet

9.93 ms

style: select+unselect 256-band — constant vs state-case sheet — constant sheet (skips restyle)

6.71 µs

style: select+unselect 256-band — constant vs state-case sheet — paint-only state sheet (round-61 diff path)

32.6 µs

style: select+unselect 256-band — constant vs state-case sheet — state-case sheet (all channels; full write)

382 µs

style: read one prop (color)

1.76×

style: read one prop (width)

2.1×

style: numericStyle

2.7×

style: renderedStyle (one prop)

13×

style: whole-object style()

5.9×

style: bypass set (ele.style name,value)

4.1×

style: bypass set + removeStyle round-trip

5.7×

style: select+unselect 256-band under the default sheet — bypass punch-out — no bypasses (diff path; the O(0) gate row)

33.1 µs

style: select+unselect 256-band under the default sheet — bypass punch-out — 128 of 256 bypassed (punched out to merged writes)

226 µs

style: sheet swap (applyAll) — 0 vs 128 bypasses — no bypasses

2.46 ms

style: sheet swap (applyAll) — 0 vs 128 bypasses — 128 bypassed

8.59 ms

10 ns100 ns1 µs10 µs100 µs1 ms10 ms100 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
style: sheet swap (compile + applyAll)	v3	13.3 ms	13.4 ms	15.5 ms	12.5 ms	47
style: sheet swap (compile + applyAll)	gpu	2.28 ms	2.31 ms	2.69 ms	2.22 ms	289
style: compile vs compile + apply	compile only (in batch)	24.6 µs	28.2 µs	35.3 µs	23.8 µs	12
style: compile vs compile + apply	compile + apply	2.25 ms	2.28 ms	2.62 ms	2.18 ms	292
style: first apply on add (256-node band)	v3	1.40 ms	1.66 ms	6.98 ms	1.16 ms	412
style: first apply on add (256-node band)	gpu	739 µs	873 µs	2.67 ms	690 µs	791
style: applyAll, flat vs compound (100 parents — the 14.6 partition)	flat	2.27 ms	2.34 ms	3.02 ms	2.20 ms	296
style: applyAll, flat vs compound (100 parents — the 14.6 partition)	compound	2.74 ms	2.83 ms	4.77 ms	2.64 ms	243
style: applyAll — constant vs state-case (the 57.1d partition)	constant sheet	2.26 ms	2.29 ms	2.70 ms	2.21 ms	303
style: applyAll — constant vs state-case (the 57.1d partition)	state-case sheet	2.27 ms	2.30 ms	2.89 ms	2.17 ms	300
style: applyAll — state-case vs data-case (what the partition avoids)	state-case sheet	2.29 ms	2.32 ms	2.71 ms	2.19 ms	298
style: applyAll — state-case vs data-case (what the partition avoids)	data-case sheet	9.93 ms	10.0 ms	10.8 ms	9.68 ms	63
style: select+unselect 256-band — constant vs state-case sheet	constant sheet (skips restyle)	6.71 µs	6.73 µs	6.86 µs	6.64 µs	21
style: select+unselect 256-band — constant vs state-case sheet	paint-only state sheet (round-61 diff path)	32.6 µs	33.9 µs	55.9 µs	31.2 µs	19871
style: select+unselect 256-band — constant vs state-case sheet	state-case sheet (all channels; full write)	382 µs	397 µs	679 µs	378 µs	1744
style: read one prop (color)	v3	94.0 ns	107 ns	618 ns	85.0 ns	1986392
style: read one prop (color)	gpu	53.4 ns	58.3 ns	97.6 ns	51.3 ns	2923
style: read one prop (width)	v3	113 ns	121 ns	433 ns	96.0 ns	1934561
style: read one prop (width)	gpu	53.3 ns	58.3 ns	309 ns	51.0 ns	2920
style: numericStyle	v3	139 ns	133 ns	357 ns	86.0 ns	1838743
style: numericStyle	gpu	51.5 ns	53.9 ns	116 ns	48.0 ns	3170
style: renderedStyle (one prop)	v3	918 ns	981 ns	2.00 µs	866 ns	540568
style: renderedStyle (one prop)	gpu	68.8 ns	72.9 ns	180 ns	67.1 ns	2346
style: whole-object style()	v3	84.8 µs	90.5 µs	180 µs	82.8 µs	7684
style: whole-object style()	gpu	14.4 µs	16.2 µs	52.3 µs	13.6 µs	41636
style: bypass set (ele.style name,value)	v3	10.6 µs	10.5 µs	10.9 µs	9.84 µs	11
style: bypass set (ele.style name,value)	gpu	2.57 µs	2.86 µs	4.12 µs	2.44 µs	54
style: bypass set + removeStyle round-trip	v3	20.9 µs	22.5 µs	22.5 µs	20.1 µs	12
style: bypass set + removeStyle round-trip	gpu	3.64 µs	3.76 µs	4.59 µs	3.32 µs	40
style: select+unselect 256-band under the default sheet — bypass punch-out	no bypasses (diff path; the O(0) gate row)	33.1 µs	33.7 µs	46.2 µs	31.7 µs	19660
style: select+unselect 256-band under the default sheet — bypass punch-out	128 of 256 bypassed (punched out to merged writes)	226 µs	236 µs	389 µs	219 µs	2956
style: sheet swap (applyAll) — 0 vs 128 bypasses	no bypasses	2.46 ms	2.50 ms	2.96 ms	2.35 ms	276
style: sheet swap (applyAll) — 0 vs 128 bypasses	128 bypassed	8.59 ms	8.89 ms	11.5 ms	8.36 ms	73

style-bundle N=2,000 · 58.5 sv4 (gpu) v3

style( 'background-color' ) (was #4 in the switch)

1.13×

style( 'border-width' ) (was #6 in the switch)

1.11×

style( 'text-wrap' ) (was #73 in the switch)

1.72×

style( 'text-max-width' ) (was #74 in the switch)

1.73×

style( 'taxi-radius' ) (was #142 in the switch)

1.25×

style( 'target-distance-from-node' ) (was #150 in the switch)

1.35×

whole-object style() (gpu-only: v3 returns a different shape) — gpu node

13.4 µs

whole-object style() (gpu-only: v3 returns a different shape) — gpu edge

14.5 µs

colour-valued vs numeric reads (the string build) — gpu background-color (colour)

73.0 ns

colour-valued vs numeric reads (the string build) — gpu border-color (colour)

71.3 ns

colour-valued vs numeric reads (the string build) — gpu border-width (number)

66.7 ns

colour-valued vs numeric reads (the string build) — gpu width (number)

58.6 ns

the getters that ride readProp — gpu numericStyle( width )

61.0 ns

the getters that ride readProp — gpu effectiveOpacity()

21.5 ns

the getters that ride readProp — gpu renderedStyle( width )

77.2 ns

10 ns100 ns1 µs10 µs100 µs

All stats (table view)

group	bench	p50	avg	p99	min	samples
style( 'background-color' ) (was #4 in the switch)	v3	59.6 ns	60.8 ns	73.9 ns	57.2 ns	2821
style( 'background-color' ) (was #4 in the switch)	gpu	52.7 ns	54.5 ns	84.4 ns	50.4 ns	3127
style( 'border-width' ) (was #6 in the switch)	v3	60.4 ns	61.3 ns	75.0 ns	58.1 ns	2798
style( 'border-width' ) (was #6 in the switch)	gpu	54.3 ns	55.5 ns	68.9 ns	52.0 ns	3072
style( 'text-wrap' ) (was #73 in the switch)	v3	61.3 ns	62.6 ns	80.6 ns	59.1 ns	2731
style( 'text-wrap' ) (was #73 in the switch)	gpu	35.6 ns	36.3 ns	46.6 ns	34.0 ns	4732
style( 'text-max-width' ) (was #74 in the switch)	v3	59.0 ns	60.2 ns	74.1 ns	57.0 ns	2855
style( 'text-max-width' ) (was #74 in the switch)	gpu	34.0 ns	34.6 ns	45.0 ns	32.5 ns	4954
style( 'taxi-radius' ) (was #142 in the switch)	v3	60.7 ns	61.9 ns	78.0 ns	58.9 ns	2777
style( 'taxi-radius' ) (was #142 in the switch)	gpu	48.5 ns	51.6 ns	77.7 ns	46.5 ns	3305
style( 'target-distance-from-node' ) (was #150 in the switch)	v3	65.0 ns	68.3 ns	112 ns	62.5 ns	2514
style( 'target-distance-from-node' ) (was #150 in the switch)	gpu	48.3 ns	49.4 ns	67.2 ns	46.8 ns	3468
whole-object style() (gpu-only: v3 returns a different shape)	gpu node	13.4 µs	14.3 µs	30.2 µs	12.8 µs	47075
whole-object style() (gpu-only: v3 returns a different shape)	gpu edge	14.5 µs	15.5 µs	36.7 µs	13.9 µs	43432
colour-valued vs numeric reads (the string build)	gpu background-color (colour)	73.0 ns	75.0 ns	109 ns	70.5 ns	2279
colour-valued vs numeric reads (the string build)	gpu border-color (colour)	71.3 ns	72.8 ns	105 ns	69.2 ns	2351
colour-valued vs numeric reads (the string build)	gpu border-width (number)	66.7 ns	68.7 ns	137 ns	64.2 ns	2487
colour-valued vs numeric reads (the string build)	gpu width (number)	58.6 ns	59.8 ns	76.2 ns	56.2 ns	2858
the getters that ride readProp	gpu numericStyle( width )	61.0 ns	62.3 ns	85.1 ns	59.1 ns	2746
the getters that ride readProp	gpu effectiveOpacity()	21.5 ns	21.8 ns	31.2 ns	20.4 ns	7857
the getters that ride readProp	gpu renderedStyle( width )	77.2 ns	77.9 ns	87.5 ns	74.9 ns	2196

load N=2,000 · 58.5 sv4 (gpu) v3

load: init from the definition form (construct + dispose)

26×

load: v4 ingest forms (construct + dispose) — definition form

4.36 ms

load: v4 ingest forms (construct + dispose) — columnar form

3.30 ms

load: v4 ingest forms (construct + dispose) — wire buffer

2.86 ms

load: v4 ingest forms (construct + dispose) — def clone alone (control)

617 µs

load: conversion — toColumnarElements

696 µs

load: conversion — serializeElements (from columnar)

181 µs

load: conversion — serializeElements (from defs)

870 µs

load: conversion — deserializeElements

4.59 µs

load: export from a live graph — cy.serialize() (wire)

499 µs

load: export from a live graph — cy.json() (v4)

727 µs

load: cy.json()

2.6×

load: add a 256-node band (add + remove)

1.59×

load: add a 256-node band, v4 forms — definition form

2.32 ms

load: add a 256-node band, v4 forms — columnar form

1.92 ms

load: add a 256-node band, v4 forms — wire buffer

1.92 ms

1 µs10 µs100 µs1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
load: init from the definition form (construct + dispose)	v3	133 ms	144 ms	173 ms	127 ms	12
load: init from the definition form (construct + dispose)	gpu	5.02 ms	6.54 ms	20.3 ms	4.26 ms	98
load: v4 ingest forms (construct + dispose)	definition form	4.36 ms	6.55 ms	16.4 ms	4.12 ms	103
load: v4 ingest forms (construct + dispose)	columnar form	3.30 ms	4.72 ms	13.1 ms	3.19 ms	142
load: v4 ingest forms (construct + dispose)	wire buffer	2.86 ms	4.14 ms	12.9 ms	2.80 ms	162
load: v4 ingest forms (construct + dispose)	def clone alone (control)	617 µs	800 µs	2.40 ms	204 µs	825
load: conversion	toColumnarElements	696 µs	745 µs	1.25 ms	676 µs	929
load: conversion	serializeElements (from columnar)	181 µs	367 µs	967 µs	173 µs	1896
load: conversion	serializeElements (from defs)	870 µs	894 µs	1.32 ms	845 µs	777
load: conversion	deserializeElements	4.59 µs	4.42 µs	5.79 µs	3.30 µs	30
load: export from a live graph	cy.serialize() (wire)	499 µs	571 µs	1.97 ms	486 µs	1203
load: export from a live graph	cy.json() (v4)	727 µs	1.08 ms	6.77 ms	686 µs	622
load: cy.json()	v3	2.02 ms	2.34 ms	6.25 ms	1.96 ms	291
load: cy.json()	gpu	762 µs	903 µs	2.60 ms	714 µs	770
load: add a 256-node band (add + remove)	v3	3.69 ms	4.40 ms	10.9 ms	3.22 ms	150
load: add a 256-node band (add + remove)	gpu	2.32 ms	2.62 ms	7.89 ms	1.62 ms	258
load: add a 256-node band, v4 forms	definition form	2.32 ms	2.34 ms	5.29 ms	1.53 ms	291
load: add a 256-node band, v4 forms	columnar form	1.92 ms	2.01 ms	4.26 ms	1.29 ms	337
load: add a 256-node band, v4 forms	wire buffer	1.92 ms	1.97 ms	3.86 ms	1.29 ms	348

spatial N=2,000 · 2.0 minv4 (gpu) v3

pick: hit (the walk; one inside-test) — zoom 1

20.7 µs

pick: hit (the walk; one inside-test) — zoom 2

20.8 µs

pick: miss walk — exact vs 24 px touch halo (57.9) — exact

46.2 µs

pick: miss walk — exact vs 24 px touch halo (57.9) — padPx 24

44.7 µs

pick: shape inside-test, 2000 coincident misses — ellipse

87.5 µs

pick: shape inside-test, 2000 coincident misses — star

234 µs

pick: shape inside-test, 2000 coincident misses — round-hexagon

792 µs

pick: shape inside-test, 2000 coincident misses — barrel

688 µs

pick: shape inside-test, 2000 coincident misses — cut-rectangle

99.6 µs

pick: shape inside-test, 2000 coincident misses — polygon

209 µs

pick: round-hexagon inside-test by zoom (not affine-invariant) — zoom 1

803 µs

pick: round-hexagon inside-test by zoom (not affine-invariant) — zoom 2

791 µs

pick: background miss (full highWater walk) — gpu

44.9 µs

pick: label box in the scan (text-events) — text-events: no

45.0 µs

pick: label box in the scan (text-events) — text-events: yes

123 µs

box: elementsInBox by covered fraction — 10%

149 µs

box: elementsInBox by covered fraction — 50%

197 µs

box: elementsInBox by covered fraction — 100%

331 µs

box: label containment (boxSelectionIncludesLabels) — off

229 µs

box: label containment (boxSelectionIncludesLabels) — on

262 µs

box: overlap vs contain (straight edges) — contain

234 µs

box: overlap vs contain (straight edges) — overlap

463 µs

box: overlap vs contain (curved edges — the segment walk) — contain

1.25 ms

box: overlap vs contain (curved edges — the segment walk) — overlap

1.21 ms

bounds: whole-graph boundingBox()

6.4×

bounds: boundingBox, labels in vs out (v4) — includeLabels: true (default)

1.20 ms

bounds: boundingBox, labels in vs out (v4) — includeLabels: false

656 µs

bounds: cy.fit() (whole-graph scan)

33×

bounds: getFitViewport() (compute without committing)

33×

bounds: boundingBoxAt (v4 only) — gpu

1.47 ms

bounds: one node boundingBox()

1.63×

bounds: labelBoundingBox (v4 only) — gpu

415 ns

bounds: compound + taxi exact scan (rounds 54/92, v4 only) — warm (curve bbs memoized)

19.2 µs

bounds: compound + taxi exact scan (rounds 54/92, v4 only) — cold (geometry write per call)

204 µs

bounds: straight-only scan (ndex-shaped, round 92, v4 only) — warm

34.1 µs

bounds: straight-only scan (ndex-shaped, round 92, v4 only) — cold (geometry write per call)

34.6 µs

100 ns1 µs10 µs100 µs1 ms10 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
pick: hit (the walk; one inside-test)	zoom 1	20.7 µs	21.5 µs	42.5 µs	20.0 µs	31984
pick: hit (the walk; one inside-test)	zoom 2	20.8 µs	21.2 µs	21.6 µs	20.4 µs	12
pick: miss walk — exact vs 24 px touch halo (57.9)	exact	46.2 µs	46.5 µs	47.1 µs	45.7 µs	12
pick: miss walk — exact vs 24 px touch halo (57.9)	padPx 24	44.7 µs	45.9 µs	83.4 µs	42.2 µs	15158
pick: shape inside-test, 2000 coincident misses	ellipse	87.5 µs	90.4 µs	136 µs	83.6 µs	7726
pick: shape inside-test, 2000 coincident misses	star	234 µs	242 µs	293 µs	224 µs	2877
pick: shape inside-test, 2000 coincident misses	round-hexagon	792 µs	800 µs	921 µs	770 µs	875
pick: shape inside-test, 2000 coincident misses	barrel	688 µs	708 µs	885 µs	663 µs	985
pick: shape inside-test, 2000 coincident misses	cut-rectangle	99.6 µs	101 µs	142 µs	96.0 µs	6908
pick: shape inside-test, 2000 coincident misses	polygon	209 µs	214 µs	272 µs	202 µs	3278
pick: round-hexagon inside-test by zoom (not affine-invariant)	zoom 1	803 µs	812 µs	990 µs	774 µs	862
pick: round-hexagon inside-test by zoom (not affine-invariant)	zoom 2	791 µs	799 µs	932 µs	768 µs	876
pick: background miss (full highWater walk)	gpu	44.9 µs	45.3 µs	46.1 µs	44.4 µs	12
pick: label box in the scan (text-events)	text-events: no	45.0 µs	45.3 µs	46.5 µs	44.3 µs	12
pick: label box in the scan (text-events)	text-events: yes	123 µs	128 µs	193 µs	118 µs	5447
box: elementsInBox by covered fraction	10%	149 µs	154 µs	247 µs	145 µs	4538
box: elementsInBox by covered fraction	50%	197 µs	203 µs	374 µs	188 µs	3421
box: elementsInBox by covered fraction	100%	331 µs	340 µs	783 µs	312 µs	2052
box: label containment (boxSelectionIncludesLabels)	off	229 µs	237 µs	374 µs	222 µs	2939
box: label containment (boxSelectionIncludesLabels)	on	262 µs	275 µs	630 µs	251 µs	2540
box: overlap vs contain (straight edges)	contain	234 µs	241 µs	370 µs	231 µs	2898
box: overlap vs contain (straight edges)	overlap	463 µs	470 µs	801 µs	412 µs	1484
box: overlap vs contain (curved edges — the segment walk)	contain	1.25 ms	1.27 ms	1.65 ms	1.21 ms	550
box: overlap vs contain (curved edges — the segment walk)	overlap	1.21 ms	1.24 ms	1.93 ms	1.16 ms	562
bounds: whole-graph boundingBox()	v3	7.72 ms	7.79 ms	8.90 ms	7.28 ms	84
bounds: whole-graph boundingBox()	gpu	1.21 ms	1.23 ms	1.52 ms	1.17 ms	549
bounds: boundingBox, labels in vs out (v4)	includeLabels: true (default)	1.20 ms	1.24 ms	1.86 ms	1.17 ms	562
bounds: boundingBox, labels in vs out (v4)	includeLabels: false	656 µs	670 µs	985 µs	632 µs	1043
bounds: cy.fit() (whole-graph scan)	v3	7.57 ms	7.89 ms	11.8 ms	7.25 ms	85
bounds: cy.fit() (whole-graph scan)	gpu	232 µs	244 µs	450 µs	225 µs	2868
bounds: getFitViewport() (compute without committing)	v3	7.62 ms	7.82 ms	9.77 ms	7.31 ms	85
bounds: getFitViewport() (compute without committing)	gpu	232 µs	242 µs	359 µs	225 µs	2888
bounds: boundingBoxAt (v4 only)	gpu	1.47 ms	1.55 ms	3.01 ms	1.41 ms	438
bounds: one node boundingBox()	v3	1.64 µs	1.65 µs	2.02 µs	1.58 µs	97
bounds: one node boundingBox()	gpu	1.00 µs	1.02 µs	1.33 µs	969 ns	160
bounds: labelBoundingBox (v4 only)	gpu	415 ns	448 ns	780 ns	412 ns	374
bounds: compound + taxi exact scan (rounds 54/92, v4 only)	warm (curve bbs memoized)	19.2 µs	20.0 µs	46.0 µs	18.4 µs	34403
bounds: compound + taxi exact scan (rounds 54/92, v4 only)	cold (geometry write per call)	204 µs	214 µs	334 µs	200 µs	3266
bounds: straight-only scan (ndex-shaped, round 92, v4 only)	warm	34.1 µs	34.4 µs	35.1 µs	33.5 µs	12
bounds: straight-only scan (ndex-shaped, round 92, v4 only)	cold (geometry write per call)	34.6 µs	35.2 µs	35.9 µs	33.9 µs	12

data N=2,000 · 1.4 minv4 (gpu) v3

data: bulk write, numeric column

21×

data: bulk write, dictionary string column (4 values)

17×

data: bulk write, one new dictionary entry per pass

17×

data: bulk write, object column (plain-array fallback)

20×

data: removeData (whole collection)

7.8×

data: read one numeric key

1.27×

data: read one string key (dictionary decode)

1.34×

data: read the whole object

2.3×

query: data equality ([bucket = 3] vs { data: { bucket: 3 } })

14×

query: data comparison ([weight > 3] vs { gt })

11×

query: two keys AND-ed

12×

query: membership ([bucket in ...] vs { in })

16×

query: predicate function (both sides)

7.9×

query: structural, parents (:parent vs { parent: true })

35×

query: structural, childless (:childless vs { parent: false })

13×

query: structural, orphans (:orphan vs { child: false })

32×

10 ns100 ns1 µs10 µs100 µs1 ms10 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
data: bulk write, numeric column	v3	1.79 ms	1.87 ms	3.14 ms	1.69 ms	358
data: bulk write, numeric column	gpu	85.9 µs	90.2 µs	154 µs	82.7 µs	7749
data: bulk write, dictionary string column (4 values)	v3	1.71 ms	1.74 ms	2.10 ms	1.63 ms	388
data: bulk write, dictionary string column (4 values)	gpu	103 µs	109 µs	313 µs	95.6 µs	6435
data: bulk write, one new dictionary entry per pass	v3	1.70 ms	1.73 ms	2.09 ms	1.66 ms	398
data: bulk write, one new dictionary entry per pass	gpu	99.9 µs	104 µs	158 µs	96.4 µs	6704
data: bulk write, object column (plain-array fallback)	v3	1.71 ms	1.80 ms	3.07 ms	1.64 ms	384
data: bulk write, object column (plain-array fallback)	gpu	83.8 µs	86.7 µs	136 µs	79.8 µs	8066
data: removeData (whole collection)	v3	2.53 ms	2.62 ms	4.23 ms	2.45 ms	262
data: removeData (whole collection)	gpu	326 µs	343 µs	704 µs	319 µs	2011
data: read one numeric key	v3	38.3 ns	39.9 ns	117 ns	36.6 ns	4299
data: read one numeric key	gpu	30.3 ns	31.8 ns	81.2 ns	29.1 ns	5247
data: read one string key (dictionary decode)	v3	41.4 ns	43.9 ns	91.4 ns	39.3 ns	3910
data: read one string key (dictionary decode)	gpu	31.0 ns	32.0 ns	68.9 ns	26.7 ns	5355
data: read the whole object	v3	47.7 ns	49.1 ns	75.5 ns	45.4 ns	3501
data: read the whole object	gpu	21.1 ns	22.2 ns	40.5 ns	20.1 ns	7668
query: data equality ([bucket = 3] vs { data: { bucket: 3 } })	v3	682 µs	695 µs	1.14 ms	667 µs	1004
query: data equality ([bucket = 3] vs { data: { bucket: 3 } })	gpu	48.2 µs	52.3 µs	124 µs	43.3 µs	13200
query: data comparison ([weight > 3] vs { gt })	v3	764 µs	783 µs	1.25 ms	732 µs	890
query: data comparison ([weight > 3] vs { gt })	gpu	69.2 µs	71.6 µs	161 µs	56.9 µs	9673
query: two keys AND-ed	v3	796 µs	812 µs	1.09 ms	783 µs	859
query: two keys AND-ed	gpu	66.4 µs	70.2 µs	139 µs	65.3 µs	9892
query: membership ([bucket in ...] vs { in })	v3	1.01 ms	1.02 ms	1.35 ms	967 µs	683
query: membership ([bucket in ...] vs { in })	gpu	65.0 µs	69.8 µs	134 µs	62.3 µs	9744
query: predicate function (both sides)	v3	657 µs	672 µs	929 µs	640 µs	1036
query: predicate function (both sides)	gpu	83.5 µs	86.5 µs	139 µs	79.5 µs	8066
query: structural, parents (:parent vs { parent: true })	v3	530 µs	538 µs	662 µs	509 µs	1299
query: structural, parents (:parent vs { parent: true })	gpu	15.2 µs	16.2 µs	19.0 µs	15.1 µs	12
query: structural, childless (:childless vs { parent: false })	v3	795 µs	815 µs	1.11 ms	773 µs	857
query: structural, childless (:childless vs { parent: false })	gpu	59.6 µs	73.5 µs	184 µs	55.4 µs	9266
query: structural, orphans (:orphan vs { child: false })	v3	499 µs	507 µs	599 µs	484 µs	1335
query: structural, orphans (:orphan vs { child: false })	gpu	15.4 µs	16.1 µs	16.2 µs	15.3 µs	12

events N=2,000 · 1.3 minv4 (gpu) v3

emit: position write, no listeners

26×

emit: 1 core listener

8.9×

emit: 1 element (ref-qualified) listener

8.0×

emit: delegated (v3 selector vs v4 predicate)

8.2×

emit: 10 core listeners

6.6×

emit: on() + off() round-trip

2.1×

emit: flat vs phased (compound bubbling, 14.5) — orphan (flat emit)

295 ns

emit: flat vs phased (compound bubbling, 14.5) — child of 2 ancestors (phased)

845 ns

emit: flat vs phased, no listeners — orphan (flat)

80.6 ns

emit: flat vs phased, no listeners — child (phased)

526 ns

anim: start + stop, one element

1.31×

anim: start against a running animation (round-21 eviction) — disjoint channels (composes)

1.29 µs

anim: start against a running animation (round-21 eviction) — overlapping channels (evicts)

818 ns

anim: start + stop, 512-node collection

2.9×

anim: delay() (no channels touched)

1.57×

anim: controls (v4 only) — pause + resume

463 ns

anim: controls (v4 only) — reverse

893 ns

anim: viewport animate + stop

2.9×

10 ns100 ns1 µs10 µs100 µs1 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
emit: position write, no listeners	v3	1.29 µs	1.31 µs	1.78 µs	1.23 µs	115
emit: position write, no listeners	gpu	49.0 ns	51.8 ns	81.1 ns	46.5 ns	3302
emit: 1 core listener	v3	1.45 µs	1.45 µs	1.59 µs	1.39 µs	107
emit: 1 core listener	gpu	164 ns	168 ns	254 ns	157 ns	936
emit: 1 element (ref-qualified) listener	v3	1.41 µs	1.48 µs	2.21 µs	1.34 µs	106
emit: 1 element (ref-qualified) listener	gpu	178 ns	184 ns	222 ns	172 ns	867
emit: delegated (v3 selector vs v4 predicate)	v3	1.52 µs	1.53 µs	1.63 µs	1.48 µs	101
emit: delegated (v3 selector vs v4 predicate)	gpu	187 ns	190 ns	257 ns	179 ns	839
emit: 10 core listeners	v3	1.97 µs	2.03 µs	2.34 µs	1.87 µs	77
emit: 10 core listeners	gpu	296 ns	302 ns	353 ns	295 ns	508
emit: on() + off() round-trip	v3	459 ns	567 ns	1.40 µs	428 ns	293
emit: on() + off() round-trip	gpu	214 ns	257 ns	909 ns	184 ns	641
emit: flat vs phased (compound bubbling, 14.5)	orphan (flat emit)	295 ns	344 ns	2.74 µs	276 ns	1012114
emit: flat vs phased (compound bubbling, 14.5)	child of 2 ancestors (phased)	845 ns	896 ns	1.34 µs	810 ns	177
emit: flat vs phased, no listeners	orphan (flat)	80.6 ns	90.4 ns	682 ns	78.3 ns	1870
emit: flat vs phased, no listeners	child (phased)	526 ns	578 ns	1.91 µs	505 ns	288
anim: start + stop, one element	v3	826 ns	1.22 µs	3.39 µs	776 ns	421268
anim: start + stop, one element	gpu	630 ns	1.05 µs	6.85 µs	582 ns	464771
anim: start against a running animation (round-21 eviction)	disjoint channels (composes)	1.29 µs	1.91 µs	3.82 µs	1.07 µs	80
anim: start against a running animation (round-21 eviction)	overlapping channels (evicts)	818 ns	1.42 µs	3.63 µs	792 ns	107
anim: start + stop, 512-node collection	v3	294 µs	455 µs	3.01 ms	284 µs	1465
anim: start + stop, 512-node collection	gpu	99.9 µs	113 µs	185 µs	97.4 µs	5487
anim: delay() (no channels touched)	v3	646 ns	962 ns	3.26 µs	628 ns	168
anim: delay() (no channels touched)	gpu	410 ns	704 ns	2.67 µs	399 ns	231
anim: controls (v4 only)	pause + resume	463 ns	745 ns	2.63 µs	451 ns	205
anim: controls (v4 only)	reverse	893 ns	1.52 µs	3.45 µs	881 ns	100
anim: viewport animate + stop	v3	558 ns	837 ns	3.10 µs	543 ns	190
anim: viewport animate + stop	gpu	194 ns	224 ns	825 ns	192 ns	740

store N=2,000 · 1.0 min

id index: set x 2000 (build) — gpu

263 µs

id index: single-key operations — get (probe)

66.1 ns

id index: single-key operations — has

50.9 ns

id index: single-key operations — idAt (memoized hit)

8.96 ns

id index: single-key operations — hashAt

6.98 ns

id index: remove + re-set (churn, drives the round-11 reclaim) — gpu

231 ns

adjacency: rebuild (4000 edges, two counting passes) — gpu

39.2 µs

adjacency: reads — outDegree (O(1))

6.16 ns

adjacency: reads — outEdges

44.9 ns

adjacency: reads — connectedEdges

117 ns

adjacency: addEdge + removeEdge (the overlay path) — gpu

143 ns

blob: write x 2000 (build) — gpu

104 µs

blob: single-record operations — rewrite in place (same length)

29.5 ns

blob: single-record operations — offsetOf

4.24 ns

blob: single-record operations — free + rewrite

151 ns

dirty: mark (per-write span coalescing) — contiguous marks

12.7 ns

dirty: mark (per-write span coalescing) — scattered marks

17.7 ns

dirty: take (a frame's drain) — gpu

587 ns

images: registry bookkeeping — acquire (existing url)

202 ns

images: registry bookkeeping — acquire (new url) + release

531 ns

charts: sheet apply, chart vs no chart — no chart

2.21 ms

charts: sheet apply, chart vs no chart — pie from data (blob record per node)

2.65 ms

charts: data write refreshing chart-values — gpu

1.48 ms

1 ns10 ns100 ns1 µs10 µs100 µs1 ms10 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
id index: set x 2000 (build)	gpu	263 µs	277 µs	443 µs	252 µs	2536
id index: single-key operations	get (probe)	66.1 ns	65.7 ns	114 ns	52.9 ns	2597
id index: single-key operations	has	50.9 ns	51.9 ns	69.6 ns	49.4 ns	3291
id index: single-key operations	idAt (memoized hit)	8.96 ns	9.19 ns	17.4 ns	8.59 ns	18527
id index: single-key operations	hashAt	6.98 ns	7.17 ns	13.9 ns	6.77 ns	23639
id index: remove + re-set (churn, drives the round-11 reclaim)	gpu	231 ns	237 ns	297 ns	227 ns	712
adjacency: rebuild (4000 edges, two counting passes)	gpu	39.2 µs	45.7 µs	209 µs	36.8 µs	15170
adjacency: reads	outDegree (O(1))	6.16 ns	6.27 ns	9.48 ns	5.95 ns	27061
adjacency: reads	outEdges	44.9 ns	45.8 ns	58.1 ns	43.9 ns	3732
adjacency: reads	connectedEdges	117 ns	180 ns	2.12 µs	114 ns	931
adjacency: addEdge + removeEdge (the overlay path)	gpu	143 ns	146 ns	209 ns	136 ns	1159
blob: write x 2000 (build)	gpu	104 µs	126 µs	628 µs	98.6 µs	5539
blob: single-record operations	rewrite in place (same length)	29.5 ns	30.1 ns	40.7 ns	28.7 ns	5679
blob: single-record operations	offsetOf	4.24 ns	4.37 ns	8.73 ns	4.09 ns	38513
blob: single-record operations	free + rewrite	151 ns	171 ns	510 ns	146 ns	998
dirty: mark (per-write span coalescing)	contiguous marks	12.7 ns	13.2 ns	25.6 ns	12.1 ns	12836
dirty: mark (per-write span coalescing)	scattered marks	17.7 ns	18.2 ns	28.1 ns	17.1 ns	9370
dirty: take (a frame's drain)	gpu	587 ns	696 ns	1.30 µs	580 ns	237
images: registry bookkeeping	acquire (existing url)	202 ns	210 ns	399 ns	200 ns	812
images: registry bookkeeping	acquire (new url) + release	531 ns	605 ns	2.82 µs	521 ns	272
charts: sheet apply, chart vs no chart	no chart	2.21 ms	2.30 ms	4.35 ms	2.17 ms	292
charts: sheet apply, chart vs no chart	pie from data (blob record per node)	2.65 ms	2.69 ms	3.21 ms	2.53 ms	255
charts: data write refreshing chart-values	gpu	1.48 ms	1.54 ms	2.18 ms	1.47 ms	451

surface N=2,000 · 8.7 minv4 (gpu) v3

viewport: zoom() get

4.7×

viewport: zoom( n ) set

1.86×

viewport: pan() get

1.41×

viewport: pan( p ) set

1.80×

viewport: panBy()

1.71×

viewport: extent()

1.55×

viewport: renderedExtent()

1.65×

viewport: width() + height() + size() (3 calls)

3.8×

viewport: minZoom/maxZoom get (2 calls)

15×

viewport: center( eles )

19×

viewport: reset()

107×

viewport: viewport({ zoom, pan })

57×

viewport: getCenterPan( eles ) — gpu

6.10 µs

core: instanceString + headless + styleEnabled (3 calls)

1.25×

core: hasElementWithId + hasCompoundNodes (2 calls)

1.18×

core: options()

85×

core: elements() memo hit (was mutableElements, round 90)

1.02×

core: gating getters (5 calls)

1.16×

core: gating setters round-trip (autolock/autoungrabify/autounselectify)

1.87×

core: interaction tuning getters (5 calls) — gpu

7.30 ns

core: graph data() set + get (v3 restyles the graph)

8660×

core: graph scratch() set + get (v3 restyles the graph)

16146×

core: batch( fn ) (empty)

6.3×

core: startBatch + endBatch + batching

5.9×

core: collection()

7.2×

core: compact() (no dead slots — the early-out) — gpu

16.9 µs

collection: element accessors (id/group/isNode/isEdge, 4 calls)

1.77×

collection: length + empty + nonempty (3 calls)

4.5×

collection: eq/first/last (3 calls)

1.83×

collection: slice()

2.3×

collection: toArray()

12×

collection: sort( fn )

1.57×

collection: reduce( fn )

1.38×

collection: min + max ( fn ) (2 calls)

2.0×

collection: some + every (2 calls)

1.59×

collection: same + anySame + contains (3 calls)

71×

collection: allAre (all match: the full walk)

1.48×

collection: is (none match: the full walk)

2.3×

collection: union

5.7×

collection: intersection

29×

collection: difference

29×

collection: symmetricDifference

3.8×

collection: absoluteComplement

15×

collection: byGroup

3.1×

collection: diff

3.1×

collection: indexOf

1.35×

traversal: openNeighborhood + closedNeighborhood (2 calls)

2.9×

traversal: connectedNodes

3.3×

traversal: sources + targets (2 calls)

3.4×

traversal: source() + target() (2 calls)

1.28×

traversal: edgesWith

1.25×

traversal: edgesTo

1.28×

traversal: parallelEdges + codirectedEdges (2 calls)

3.3×

traversal: roots + leaves (2 calls)

13×

traversal: allAreNeighbors

5.2×

traversal: isLoop + isSimple (2 calls)

2.6×

degree: indegree + outdegree (2 calls)

75×

degree: minDegree + maxDegree (2 calls)

44×

degree: minIndegree + maxOutdegree (2 calls)

25×

dimensions: renderedPosition get

2.3×

dimensions: relativePosition get

12×

dimensions: width + height + outerWidth + outerHeight (4 calls)

2.5×

dimensions: rendered dims (4 calls)

2.8×

dimensions: renderedBoundingBox

1.92×

dimensions: midpoint + renderedMidpoint (2 calls) — gpu

1.10 µs

dimensions: source/targetEndpoint (2 calls) — gpu

363 ns

dimensions: rendered source/targetEndpoint (2 calls) — gpu

384 ns

dimensions: shift + silentShift (round-trip)

60×

dimensions: silentPosition

16×

dimensions: positions( fn ) over 100

51×

dimensions: padding + paddedWidth + paddedHeight (3 calls)

12×

curves: isBundledBezier — gpu

20.7 ns

curves: controlPoints + renderedControlPoints (2 calls) — gpu

525 ns

curves: segmentPoints (undefined for bezier) — gpu

33.3 ns

compound: parent + parents + ancestors (3 calls)

3.6×

compound: children + descendants + siblings (3 calls)

4.8×

compound: orphans + nonorphans (2 calls)

7.1×

compound: commonAncestors

121×

compound: isParent/isChild/isChildless/isOrphan (4 calls)

1.48×

flags: select + unselect + selected (band)

16×

flags: selectify + unselectify + selectable

94×

flags: lock + unlock + locked

93×

flags: grabify + ungrabify + grabbable

94×

flags: show + hide + visible

228×

flags: active/activate/unactivate (3 calls)

16×

flags: panify + unpanify + pannable

94×

flags: takesUpSpace + interactive + transparent (3 calls)

3.0×

flags: effectiveOpacity

2.3×

element data: scratch get + set

7.5×

element data: json() one element

2.0×

element data: jsons() over 100

2.00×

element data: label() — gpu

15.2 ns

core: batch + per-id data() (was batchData, round 90)

9.5×

core: animated()

4.8×

core: zoomRange get — gpu

2.16 ns

core: resize() / invalidateSize()

2.5×

core: on('render') + off('render') (was onRender/offRender, round 90)

2.4×

core: removeScratch()

9890×

core: selectionType get

1.99×

core: isReady()

2.2×

core: window()

2.8×

core: container()

2.7×

core: destroyed()

2.2×

core: removeAllListeners()

8.7×

collection: cy()

3.5×

collection: indexOfId()

1.75×

collection: removed() + inside() (2 calls)

1.17×

collection: silentPositions( fn ) over 100

30×

collection: animated()

2.9×

collection: delayAnimation()

2.8×

collection: removeScratch()

8.8×

collection: component()

43×

collection: layoutDimensions()

3.3×

degree: maxIndegree + minOutdegree (2 calls)

25×

flags: grabbed + hidden + inactive (3 calls)

2.2×

curves: renderedSegmentPoints — gpu

36.1 ns

layout contract: edgeSlots + endpoints + degreeOf — gpu

34.5 µs

formats: isColumnarElements + isSerializedElements + isPackedIds — gpu

4.52 ns

0.100 ns1 ns10 ns100 ns1 µs10 µs100 µs1 ms10 ms100 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
viewport: zoom() get	v3	5.68 ns	5.88 ns	12.7 ns	5.32 ns	28692
viewport: zoom() get	gpu	1.21 ns	1.23 ns	1.91 ns	1.18 ns	128843
viewport: zoom( n ) set	v3	606 ns	612 ns	701 ns	579 ns	271
viewport: zoom( n ) set	gpu	326 ns	333 ns	389 ns	316 ns	504
viewport: pan() get	v3	6.94 ns	7.33 ns	14.8 ns	6.69 ns	23011
viewport: pan() get	gpu	4.92 ns	5.08 ns	10.0 ns	4.76 ns	33203
viewport: pan( p ) set	v3	594 ns	627 ns	1.05 µs	580 ns	267
viewport: pan( p ) set	gpu	331 ns	345 ns	615 ns	309 ns	490
viewport: panBy()	v3	590 ns	604 ns	801 ns	581 ns	279
viewport: panBy()	gpu	345 ns	353 ns	591 ns	316 ns	477
viewport: extent()	v3	30.0 ns	44.9 ns	193 ns	28.2 ns	3757
viewport: extent()	gpu	19.3 ns	26.1 ns	115 ns	18.2 ns	6478
viewport: renderedExtent()	v3	27.1 ns	28.4 ns	50.8 ns	25.8 ns	5956
viewport: renderedExtent()	gpu	16.4 ns	23.4 ns	100 ns	15.1 ns	7199
viewport: width() + height() + size() (3 calls)	v3	21.4 ns	23.0 ns	46.4 ns	20.5 ns	7393
viewport: width() + height() + size() (3 calls)	gpu	5.55 ns	5.80 ns	11.5 ns	5.31 ns	29048
viewport: minZoom/maxZoom get (2 calls)	v3	14.8 ns	15.3 ns	25.2 ns	13.9 ns	11126
viewport: minZoom/maxZoom get (2 calls)	gpu	0.999 ns	1.04 ns	1.02 ns	0.977 ns	152116
viewport: center( eles )	v3	119 µs	123 µs	186 µs	112 µs	5648
viewport: center( eles )	gpu	6.20 µs	6.23 µs	6.47 µs	6.09 µs	23
viewport: reset()	v3	939 ns	1.03 µs	1.67 µs	910 ns	162
viewport: reset()	gpu	8.81 ns	9.23 ns	18.6 ns	8.39 ns	18285
viewport: viewport({ zoom, pan })	v3	933 ns	1.03 µs	1.69 µs	911 ns	162
viewport: viewport({ zoom, pan })	gpu	16.2 ns	17.1 ns	41.0 ns	15.4 ns	9953
viewport: getCenterPan( eles )	gpu	6.10 µs	6.13 µs	6.26 µs	6.07 µs	24
core: instanceString + headless + styleEnabled (3 calls)	v3	32.1 ns	32.8 ns	45.7 ns	31.4 ns	5195
core: instanceString + headless + styleEnabled (3 calls)	gpu	25.7 ns	26.6 ns	39.6 ns	24.8 ns	6410
core: hasElementWithId + hasCompoundNodes (2 calls)	v3	37.6 ns	38.9 ns	54.7 ns	36.6 ns	4410
core: hasElementWithId + hasCompoundNodes (2 calls)	gpu	31.8 ns	33.1 ns	49.7 ns	30.4 ns	5158
core: options()	v3	91.1 ns	93.6 ns	128 ns	89.1 ns	1826
core: options()	gpu	1.07 ns	1.24 ns	6.21 ns	1.03 ns	128909
core: elements() memo hit (was mutableElements, round 90)	v3	2.41 ns	2.50 ns	2.99 ns	2.35 ns	66268
core: elements() memo hit (was mutableElements, round 90)	gpu	2.35 ns	2.40 ns	2.85 ns	2.25 ns	69209
core: gating getters (5 calls)	v3	15.6 ns	16.2 ns	24.6 ns	14.9 ns	10552
core: gating getters (5 calls)	gpu	13.5 ns	14.0 ns	27.6 ns	12.4 ns	12165
core: gating setters round-trip (autolock/autoungrabify/autounselectify)	v3	11.2 ns	11.5 ns	19.3 ns	10.5 ns	14766
core: gating setters round-trip (autolock/autoungrabify/autounselectify)	gpu	6.01 ns	6.21 ns	12.5 ns	5.73 ns	27333
core: interaction tuning getters (5 calls)	gpu	7.30 ns	7.53 ns	13.2 ns	7.14 ns	22548
core: graph data() set + get (v3 restyles the graph)	v3	2.84 ms	2.91 ms	3.70 ms	2.72 ms	237
core: graph data() set + get (v3 restyles the graph)	gpu	328 ns	336 ns	390 ns	320 ns	505
core: graph scratch() set + get (v3 restyles the graph)	v3	2.74 ms	2.78 ms	3.49 ms	2.69 ms	243
core: graph scratch() set + get (v3 restyles the graph)	gpu	169 ns	171 ns	229 ns	166 ns	997
core: batch( fn ) (empty)	v3	524 ns	625 ns	1.51 µs	501 ns	270
core: batch( fn ) (empty)	gpu	83.7 ns	86.6 ns	113 ns	81.1 ns	1956
core: startBatch + endBatch + batching	v3	526 ns	625 ns	1.67 µs	513 ns	265
core: startBatch + endBatch + batching	gpu	89.8 ns	164 ns	2.88 µs	86.9 ns	1010
core: collection()	v3	451 ns	525 ns	1.49 µs	442 ns	320
core: collection()	gpu	62.4 ns	66.9 ns	175 ns	60.8 ns	2520
core: compact() (no dead slots — the early-out)	gpu	16.9 µs	21.0 µs	97.7 µs	16.4 µs	32603
collection: element accessors (id/group/isNode/isEdge, 4 calls)	v3	60.5 ns	64.6 ns	193 ns	56.0 ns	2636
collection: element accessors (id/group/isNode/isEdge, 4 calls)	gpu	34.3 ns	36.6 ns	70.9 ns	32.9 ns	4653
collection: length + empty + nonempty (3 calls)	v3	6.66 ns	6.87 ns	12.2 ns	6.34 ns	24644
collection: length + empty + nonempty (3 calls)	gpu	1.47 ns	1.52 ns	1.94 ns	1.43 ns	107403
collection: eq/first/last (3 calls)	v3	15.2 ns	15.7 ns	25.2 ns	14.9 ns	10870
collection: eq/first/last (3 calls)	gpu	8.30 ns	8.53 ns	14.8 ns	7.99 ns	19967
collection: slice()	v3	3.83 µs	3.90 µs	4.24 µs	3.64 µs	40
collection: slice()	gpu	1.70 µs	1.70 µs	1.79 µs	1.63 µs	96
collection: toArray()	v3	8.66 µs	8.76 µs	8.75 µs	8.54 µs	16
collection: toArray()	gpu	734 ns	749 ns	872 ns	720 ns	218
collection: sort( fn )	v3	14.1 µs	14.6 µs	14.8 µs	13.9 µs	12
collection: sort( fn )	gpu	9.02 µs	9.04 µs	9.16 µs	8.92 µs	15
collection: reduce( fn )	v3	157 ns	158 ns	177 ns	115 ns	1073
collection: reduce( fn )	gpu	114 ns	120 ns	223 ns	110 ns	1436
collection: min + max ( fn ) (2 calls)	v3	11.1 µs	11.2 µs	11.4 µs	11.0 µs	12
collection: min + max ( fn ) (2 calls)	gpu	5.49 µs	5.51 µs	5.63 µs	5.42 µs	26
collection: some + every (2 calls)	v3	878 ns	892 ns	1.11 µs	852 ns	186
collection: some + every (2 calls)	gpu	554 ns	561 ns	598 ns	553 ns	302
collection: same + anySame + contains (3 calls)	v3	7.20 µs	7.22 µs	7.37 µs	7.09 µs	20
collection: same + anySame + contains (3 calls)	gpu	101 ns	103 ns	124 ns	97.4 ns	1670
collection: allAre (all match: the full walk)	v3	3.65 µs	3.73 µs	4.21 µs	3.41 µs	41
collection: allAre (all match: the full walk)	gpu	2.46 µs	2.51 µs	3.35 µs	2.40 µs	62
collection: is (none match: the full walk)	v3	5.53 µs	5.54 µs	5.76 µs	5.33 µs	27
collection: is (none match: the full walk)	gpu	2.36 µs	2.41 µs	2.84 µs	2.31 µs	66
collection: union	v3	93.8 µs	103 µs	139 µs	91.9 µs	6729
collection: union	gpu	16.4 µs	16.6 µs	16.7 µs	15.9 µs	12
collection: intersection	v3	92.8 µs	102 µs	140 µs	91.1 µs	6790
collection: intersection	gpu	3.21 µs	3.26 µs	3.62 µs	3.19 µs	48
collection: difference	v3	91.6 µs	101 µs	129 µs	90.2 µs	6871
collection: difference	gpu	3.18 µs	3.25 µs	3.55 µs	3.12 µs	48
collection: symmetricDifference	v3	24.7 µs	26.2 µs	28.2 µs	24.1 µs	12
collection: symmetricDifference	gpu	6.50 µs	6.53 µs	6.64 µs	6.41 µs	22
collection: absoluteComplement	v3	6.01 ms	6.56 ms	9.88 ms	5.79 ms	103
collection: absoluteComplement	gpu	408 µs	419 µs	696 µs	404 µs	1670
collection: byGroup	v3	1.00 ms	1.01 ms	1.16 ms	976 µs	691
collection: byGroup	gpu	326 µs	332 µs	535 µs	239 µs	2107
collection: diff	v3	35.5 µs	35.7 µs	37.5 µs	33.4 µs	12
collection: diff	gpu	11.5 µs	11.6 µs	12.3 µs	11.2 µs	11
collection: indexOf	v3	31.6 ns	32.7 ns	52.6 ns	30.5 ns	5246
collection: indexOf	gpu	23.4 ns	24.1 ns	36.3 ns	22.5 ns	7100
traversal: openNeighborhood + closedNeighborhood (2 calls)	v3	9.30 µs	9.45 µs	9.97 µs	8.97 µs	14
traversal: openNeighborhood + closedNeighborhood (2 calls)	gpu	3.21 µs	3.41 µs	4.06 µs	3.05 µs	45
traversal: connectedNodes	v3	1.01 µs	1.09 µs	1.80 µs	977 ns	152
traversal: connectedNodes	gpu	300 ns	311 ns	525 ns	293 ns	546
traversal: sources + targets (2 calls)	v3	1.72 µs	1.92 µs	2.84 µs	1.67 µs	85
traversal: sources + targets (2 calls)	gpu	508 ns	715 ns	2.44 µs	492 ns	235
traversal: source() + target() (2 calls)	v3	71.9 ns	72.9 ns	89.4 ns	59.9 ns	2344
traversal: source() + target() (2 calls)	gpu	56.0 ns	57.3 ns	71.8 ns	53.3 ns	2985
traversal: edgesWith	v3	67.6 µs	69.0 µs	107 µs	66.1 µs	10103
traversal: edgesWith	gpu	54.0 µs	54.8 µs	57.0 µs	53.5 µs	12
traversal: edgesTo	v3	68.8 µs	69.7 µs	106 µs	65.7 µs	10002
traversal: edgesTo	gpu	53.8 µs	54.1 µs	55.9 µs	52.7 µs	12
traversal: parallelEdges + codirectedEdges (2 calls)	v3	3.52 µs	3.78 µs	4.48 µs	3.40 µs	41
traversal: parallelEdges + codirectedEdges (2 calls)	gpu	1.05 µs	1.27 µs	2.08 µs	1.04 µs	128
traversal: roots + leaves (2 calls)	v3	5.17 ms	5.62 ms	9.34 ms	4.94 ms	120
traversal: roots + leaves (2 calls)	gpu	386 µs	393 µs	520 µs	379 µs	1782
traversal: allAreNeighbors	v3	194 µs	207 µs	279 µs	189 µs	3360
traversal: allAreNeighbors	gpu	37.2 µs	37.6 µs	38.8 µs	36.8 µs	12
traversal: isLoop + isSimple (2 calls)	v3	128 ns	130 ns	172 ns	125 ns	1316
traversal: isLoop + isSimple (2 calls)	gpu	49.4 ns	51.1 ns	77.1 ns	48.2 ns	3352
degree: indegree + outdegree (2 calls)	v3	3.43 µs	3.75 µs	4.71 µs	3.31 µs	42
degree: indegree + outdegree (2 calls)	gpu	46.0 ns	47.7 ns	81.8 ns	45.5 ns	3580
degree: minDegree + maxDegree (2 calls)	v3	15.9 ms	17.0 ms	20.6 ms	15.4 ms	37
degree: minDegree + maxDegree (2 calls)	gpu	364 µs	372 µs	530 µs	357 µs	1882
degree: minIndegree + maxOutdegree (2 calls)	v3	8.95 ms	9.57 ms	13.2 ms	8.75 ms	68
degree: minIndegree + maxOutdegree (2 calls)	gpu	355 µs	361 µs	465 µs	344 µs	1938
dimensions: renderedPosition get	v3	69.4 ns	72.8 ns	130 ns	67.5 ns	2332
dimensions: renderedPosition get	gpu	30.3 ns	35.3 ns	107 ns	29.2 ns	4763
dimensions: relativePosition get	v3	842 ns	889 ns	1.21 µs	797 ns	188
dimensions: relativePosition get	gpu	73.2 ns	77.4 ns	141 ns	70.1 ns	2189
dimensions: width + height + outerWidth + outerHeight (4 calls)	v3	602 ns	609 ns	696 ns	585 ns	279
dimensions: width + height + outerWidth + outerHeight (4 calls)	gpu	241 ns	244 ns	359 ns	233 ns	696
dimensions: rendered dims (4 calls)	v3	687 ns	697 ns	832 ns	667 ns	242
dimensions: rendered dims (4 calls)	gpu	249 ns	255 ns	365 ns	248 ns	671
dimensions: renderedBoundingBox	v3	1.76 µs	1.79 µs	2.18 µs	1.69 µs	90
dimensions: renderedBoundingBox	gpu	918 ns	934 ns	1.24 µs	882 ns	178
dimensions: midpoint + renderedMidpoint (2 calls)	gpu	1.10 µs	1.10 µs	1.19 µs	1.06 µs	147
dimensions: source/targetEndpoint (2 calls)	gpu	363 ns	369 ns	435 ns	355 ns	460
dimensions: rendered source/targetEndpoint (2 calls)	gpu	384 ns	389 ns	452 ns	369 ns	437
dimensions: shift + silentShift (round-trip)	v3	293 µs	303 µs	622 µs	286 µs	2310
dimensions: shift + silentShift (round-trip)	gpu	4.91 µs	4.91 µs	5.03 µs	4.83 µs	31
dimensions: silentPosition	v3	873 ns	877 ns	953 ns	843 ns	191
dimensions: silentPosition	gpu	53.2 ns	55.3 ns	80.4 ns	51.5 ns	3068
dimensions: positions( fn ) over 100	v3	180 µs	196 µs	342 µs	171 µs	3554
dimensions: positions( fn ) over 100	gpu	3.54 µs	3.60 µs	4.29 µs	3.44 µs	43
dimensions: padding + paddedWidth + paddedHeight (3 calls)	v3	2.01 µs	2.02 µs	2.24 µs	1.97 µs	81
dimensions: padding + paddedWidth + paddedHeight (3 calls)	gpu	162 ns	165 ns	193 ns	157 ns	1041
curves: isBundledBezier	gpu	20.7 ns	21.4 ns	33.5 ns	19.6 ns	7927
curves: controlPoints + renderedControlPoints (2 calls)	gpu	525 ns	540 ns	739 ns	513 ns	304
curves: segmentPoints (undefined for bezier)	gpu	33.3 ns	34.3 ns	49.9 ns	32.1 ns	4978
compound: parent + parents + ancestors (3 calls)	v3	2.92 µs	3.20 µs	4.00 µs	2.85 µs	49
compound: parent + parents + ancestors (3 calls)	gpu	802 ns	1.12 µs	2.37 µs	770 ns	147
compound: children + descendants + siblings (3 calls)	v3	8.52 µs	8.56 µs	9.32 µs	7.60 µs	16
compound: children + descendants + siblings (3 calls)	gpu	1.78 µs	2.25 µs	3.57 µs	1.70 µs	70
compound: orphans + nonorphans (2 calls)	v3	3.64 µs	3.89 µs	4.70 µs	3.52 µs	40
compound: orphans + nonorphans (2 calls)	gpu	517 ns	744 ns	3.00 µs	506 ns	220
compound: commonAncestors	v3	9.40 µs	9.62 µs	9.98 µs	9.25 µs	14
compound: commonAncestors	gpu	77.7 ns	80.8 ns	151 ns	73.6 ns	2108
compound: isParent/isChild/isChildless/isOrphan (4 calls)	v3	67.4 ns	68.1 ns	79.0 ns	60.6 ns	2522
compound: isParent/isChild/isChildless/isOrphan (4 calls)	gpu	45.4 ns	46.3 ns	57.0 ns	44.2 ns	3694
flags: select + unselect + selected (band)	v3	193 µs	201 µs	397 µs	186 µs	3480
flags: select + unselect + selected (band)	gpu	11.8 µs	11.8 µs	11.9 µs	11.5 µs	11
flags: selectify + unselectify + selectable	v3	190 µs	196 µs	283 µs	184 µs	3573
flags: selectify + unselectify + selectable	gpu	2.02 µs	2.04 µs	2.23 µs	1.99 µs	80
flags: lock + unlock + locked	v3	189 µs	196 µs	314 µs	182 µs	3566
flags: lock + unlock + locked	gpu	2.04 µs	2.05 µs	2.24 µs	1.97 µs	80
flags: grabify + ungrabify + grabbable	v3	190 µs	198 µs	386 µs	183 µs	3533
flags: grabify + ungrabify + grabbable	gpu	2.02 µs	2.03 µs	2.15 µs	2.00 µs	81
flags: show + hide + visible	v3	2.57 ms	2.67 ms	3.91 ms	2.49 ms	257
flags: show + hide + visible	gpu	11.3 µs	11.3 µs	11.5 µs	11.1 µs	11
flags: active/activate/unactivate (3 calls)	v3	191 µs	197 µs	282 µs	184 µs	3555
flags: active/activate/unactivate (3 calls)	gpu	12.0 µs	12.0 µs	12.1 µs	12.0 µs	10
flags: panify + unpanify + pannable	v3	190 µs	196 µs	311 µs	185 µs	3563
flags: panify + unpanify + pannable	gpu	2.01 µs	2.03 µs	2.14 µs	1.96 µs	81
flags: takesUpSpace + interactive + transparent (3 calls)	v3	194 ns	199 ns	266 ns	187 ns	856
flags: takesUpSpace + interactive + transparent (3 calls)	gpu	64.7 ns	67.4 ns	132 ns	61.7 ns	2536
flags: effectiveOpacity	v3	49.0 ns	50.4 ns	64.8 ns	47.0 ns	3409
flags: effectiveOpacity	gpu	21.2 ns	21.6 ns	35.5 ns	20.2 ns	7961
element data: scratch get + set	v3	1.29 µs	1.29 µs	1.40 µs	1.24 µs	129
element data: scratch get + set	gpu	172 ns	177 ns	258 ns	171 ns	964
element data: json() one element	v3	299 ns	385 ns	3.45 µs	291 ns	425
element data: json() one element	gpu	149 ns	190 ns	539 ns	145 ns	849
element data: jsons() over 100	v3	39.6 µs	40.5 µs	43.1 µs	38.7 µs	12
element data: jsons() over 100	gpu	19.8 µs	21.0 µs	24.0 µs	15.8 µs	12
element data: label()	gpu	15.2 ns	15.8 ns	28.9 ns	14.2 ns	10862
core: batch + per-id data() (was batchData, round 90)	v3	2.49 µs	2.57 µs	2.94 µs	2.42 µs	61
core: batch + per-id data() (was batchData, round 90)	gpu	263 ns	351 ns	2.11 µs	259 ns	471
core: animated()	v3	7.58 ns	7.81 ns	14.0 ns	7.22 ns	21661
core: animated()	gpu	1.56 ns	1.60 ns	1.99 ns	1.50 ns	103571
core: zoomRange get	gpu	2.16 ns	2.30 ns	6.28 ns	2.05 ns	72156
core: resize() / invalidateSize()	v3	314 ns	317 ns	385 ns	301 ns	533
core: resize() / invalidateSize()	gpu	127 ns	130 ns	161 ns	123 ns	1309
core: on('render') + off('render') (was onRender/offRender, round 90)	v3	462 ns	522 ns	1.09 µs	454 ns	322
core: on('render') + off('render') (was onRender/offRender, round 90)	gpu	195 ns	202 ns	337 ns	188 ns	818
core: removeScratch()	v3	2.75 ms	2.79 ms	3.30 ms	2.69 ms	246
core: removeScratch()	gpu	278 ns	282 ns	330 ns	268 ns	603
core: selectionType get	v3	4.51 ns	4.73 ns	8.58 ns	4.26 ns	35909
core: selectionType get	gpu	2.26 ns	2.40 ns	6.21 ns	2.10 ns	69527
core: isReady()	v3	3.26 ns	3.34 ns	6.51 ns	3.18 ns	50430
core: isReady()	gpu	1.48 ns	1.53 ns	2.01 ns	1.45 ns	107570
core: window()	v3	4.10 ns	4.24 ns	7.79 ns	4.01 ns	39847
core: window()	gpu	1.48 ns	1.53 ns	1.91 ns	1.45 ns	107941
core: container()	v3	3.93 ns	4.04 ns	8.25 ns	3.75 ns	41929
core: container()	gpu	1.48 ns	1.53 ns	1.90 ns	1.45 ns	107666
core: destroyed()	v3	3.32 ns	3.44 ns	6.44 ns	3.18 ns	48890
core: destroyed()	gpu	1.48 ns	1.53 ns	2.31 ns	1.45 ns	106940
core: removeAllListeners()	v3	13.8 ns	14.3 ns	25.0 ns	13.3 ns	11901
core: removeAllListeners()	gpu	1.59 ns	1.65 ns	2.67 ns	1.55 ns	99426
collection: cy()	v3	7.58 ns	8.13 ns	15.2 ns	7.17 ns	20890
collection: cy()	gpu	2.16 ns	2.37 ns	6.09 ns	2.06 ns	70095
collection: indexOfId()	v3	66.1 ns	67.8 ns	88.4 ns	64.7 ns	2524
collection: indexOfId()	gpu	37.8 ns	38.4 ns	52.0 ns	36.9 ns	4462
collection: removed() + inside() (2 calls)	v3	17.6 ns	17.9 ns	24.7 ns	17.3 ns	9523
collection: removed() + inside() (2 calls)	gpu	15.1 ns	15.4 ns	24.5 ns	14.6 ns	11139
collection: silentPositions( fn ) over 100	v3	112 µs	116 µs	198 µs	108 µs	6034
collection: silentPositions( fn ) over 100	gpu	3.67 µs	3.65 µs	3.79 µs	3.47 µs	43
collection: animated()	v3	12.2 ns	12.4 ns	21.4 ns	11.7 ns	13848
collection: animated()	gpu	4.19 ns	4.33 ns	8.24 ns	4.00 ns	38951
collection: delayAnimation()	v3	458 ns	583 ns	1.96 µs	449 ns	287
collection: delayAnimation()	gpu	165 ns	170 ns	242 ns	159 ns	990
collection: removeScratch()	v3	1.79 µs	1.80 µs	2.01 µs	1.72 µs	91
collection: removeScratch()	gpu	202 ns	206 ns	241 ns	197 ns	831
collection: component()	v3	78.8 ms	80.9 ms	87.2 ms	74.8 ms	12
collection: component()	gpu	1.84 ms	2.04 ms	5.31 ms	1.79 ms	338
collection: layoutDimensions()	v3	616 ns	632 ns	817 ns	601 ns	267
collection: layoutDimensions()	gpu	187 ns	194 ns	313 ns	185 ns	877
degree: maxIndegree + minOutdegree (2 calls)	v3	8.89 ms	9.59 ms	13.3 ms	8.70 ms	69
degree: maxIndegree + minOutdegree (2 calls)	gpu	355 µs	362 µs	464 µs	342 µs	1937
flags: grabbed + hidden + inactive (3 calls)	v3	84.5 ns	85.8 ns	104 ns	78.8 ns	1994
flags: grabbed + hidden + inactive (3 calls)	gpu	38.4 ns	39.6 ns	61.3 ns	36.2 ns	4327
curves: renderedSegmentPoints	gpu	36.1 ns	37.4 ns	64.0 ns	34.3 ns	4578
layout contract: edgeSlots + endpoints + degreeOf	gpu	34.5 µs	36.3 µs	64.4 µs	29.5 µs	18950
formats: isColumnarElements + isSerializedElements + isPackedIds	gpu	4.52 ns	4.73 ns	10.5 ns	4.51 ns	35620

compaction N=20,000 · 53.9 s

cpu pick: background miss (full highWater walk) — peak

207 µs

cpu pick: background miss (full highWater walk) — compacted

48.7 µs

cpu pick: node hit — peak

218 µs

cpu pick: node hit — compacted

48.9 µs

isCurrent: current ref (forwards present vs absent) — no forwards (peak)

5.95 ns

isCurrent: current ref (forwards present vs absent) — forwards present (compacted)

5.90 ns

isCurrent: stale-ref repair (chase + rewrite, alloc included) — current (baseline alloc)

8.89 ns

isCurrent: stale-ref repair (chase + rewrite, alloc included) — stale → repair

49.1 ns

control: nodes({ selected }) scan (order-list, expected parity) — peak

25.2 µs

control: nodes({ selected }) scan (order-list, expected parity) — compacted

21.6 µs

control: whole-graph boundingBox (expected parity) — peak

112 µs

control: whole-graph boundingBox (expected parity) — compacted

112 µs

1 ns10 ns100 ns1 µs10 µs100 µs1 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
cpu pick: background miss (full highWater walk)	peak	207 µs	212 µs	314 µs	201 µs	3246
cpu pick: background miss (full highWater walk)	compacted	48.7 µs	49.3 µs	50.4 µs	48.4 µs	12
cpu pick: node hit	peak	218 µs	225 µs	406 µs	212 µs	3115
cpu pick: node hit	compacted	48.9 µs	49.3 µs	49.8 µs	48.5 µs	12
isCurrent: current ref (forwards present vs absent)	no forwards (peak)	5.95 ns	6.13 ns	10.4 ns	5.69 ns	27553
isCurrent: current ref (forwards present vs absent)	forwards present (compacted)	5.90 ns	6.06 ns	10.6 ns	5.63 ns	27978
isCurrent: stale-ref repair (chase + rewrite, alloc included)	current (baseline alloc)	8.89 ns	9.51 ns	21.7 ns	8.36 ns	17761
isCurrent: stale-ref repair (chase + rewrite, alloc included)	stale → repair	49.1 ns	50.5 ns	68.7 ns	47.8 ns	3388
control: nodes({ selected }) scan (order-list, expected parity)	peak	25.2 µs	26.7 µs	33.1 µs	21.2 µs	12
control: nodes({ selected }) scan (order-list, expected parity)	compacted	21.6 µs	22.6 µs	24.6 µs	21.3 µs	12
control: whole-graph boundingBox (expected parity)	peak	112 µs	117 µs	186 µs	109 µs	5958
control: whole-graph boundingBox (expected parity)	compacted	112 µs	115 µs	177 µs	109 µs	6089

compound N=2,000 · 27.7 sv4 (gpu) v3

parent drag (shift subtree + bb settle)

183×

child drag (move + parent re-derive)

50×

reparent round-trip

77×

1 µs10 µs100 µs1 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
parent drag (shift subtree + bb settle)	v3	252 µs	276 µs	618 µs	239 µs	2524
parent drag (shift subtree + bb settle)	gpu	1.38 µs	1.44 µs	2.17 µs	1.31 µs	110
child drag (move + parent re-derive)	v3	80.1 µs	85.3 µs	124 µs	75.5 µs	7963
child drag (move + parent re-derive)	gpu	1.62 µs	1.65 µs	1.99 µs	1.57 µs	98
reparent round-trip	v3	83.0 µs	88.2 µs	128 µs	79.8 µs	7705
reparent round-trip	gpu	1.08 µs	1.08 µs	1.18 µs	1.04 µs	148

curves N=2,000 · 1.6 s

curve premium: build (style apply + derivation) — curved

12.0 ms

curve premium: build (style apply + derivation) — straight

6.64 ms

curve premium: node.position x 1000 (drag) — curved

377 µs

curve premium: node.position x 1000 (drag) — straight

326 µs

curve premium: nodes().positions() (bulk move) — curved

309 µs

curve premium: nodes().positions() (bulk move) — straight

296 µs

curve premium: first read after a bulk move (cold) — curved

5.59 ms

curve premium: first read after a bulk move (cold) — straight

2.71 ms

curve premium: same read again (warm) — curved

1.43 ms

curve premium: same read again (warm) — straight

1.42 ms

curve premium: edge.midpoint x 4000 — curved

2.37 ms

curve premium: edge.midpoint x 4000 — straight

4.21 ms

curve premium: edge.controlPoints x 4000 — curved

2.39 ms

curve premium: edge.controlPoints x 4000 — straight

1.43 ms

curve premium: cy.elements().boundingBox() x 20 — curved

17.7 ms

curve premium: cy.elements().boundingBox() x 20 — straight

21.0 ms

curve premium: cy.fit() x 20 (store scan) — curved

3.69 ms

curve premium: cy.fit() x 20 (store scan) — straight

3.38 ms

curve premium: elementsInBox x 20 (half the graph) — curved

24.3 ms

curve premium: elementsInBox x 20 (half the graph) — straight

6.03 ms

curve premium: edge.hide/show x 1000 (bundle re-fan) — curved

3.72 ms

curve premium: edge.hide/show x 1000 (bundle re-fan) — straight

2.29 ms

100 µs1 ms10 ms100 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
curve premium: build (style apply + derivation)	curved	12.0 ms	12.0 ms	12.0 ms	12.0 ms	1
curve premium: build (style apply + derivation)	straight	6.64 ms	6.64 ms	6.64 ms	6.64 ms	1
curve premium: node.position x 1000 (drag)	curved	377 µs	377 µs	377 µs	377 µs	1
curve premium: node.position x 1000 (drag)	straight	326 µs	326 µs	326 µs	326 µs	1
curve premium: nodes().positions() (bulk move)	curved	309 µs	309 µs	309 µs	309 µs	1
curve premium: nodes().positions() (bulk move)	straight	296 µs	296 µs	296 µs	296 µs	1
curve premium: first read after a bulk move (cold)	curved	5.59 ms	5.59 ms	5.59 ms	5.59 ms	1
curve premium: first read after a bulk move (cold)	straight	2.71 ms	2.71 ms	2.71 ms	2.71 ms	1
curve premium: same read again (warm)	curved	1.43 ms	1.43 ms	1.43 ms	1.43 ms	1
curve premium: same read again (warm)	straight	1.42 ms	1.42 ms	1.42 ms	1.42 ms	1
curve premium: edge.midpoint x 4000	curved	2.37 ms	2.37 ms	2.37 ms	2.37 ms	1
curve premium: edge.midpoint x 4000	straight	4.21 ms	4.21 ms	4.21 ms	4.21 ms	1
curve premium: edge.controlPoints x 4000	curved	2.39 ms	2.39 ms	2.39 ms	2.39 ms	1
curve premium: edge.controlPoints x 4000	straight	1.43 ms	1.43 ms	1.43 ms	1.43 ms	1
curve premium: cy.elements().boundingBox() x 20	curved	17.7 ms	17.7 ms	17.7 ms	17.7 ms	1
curve premium: cy.elements().boundingBox() x 20	straight	21.0 ms	21.0 ms	21.0 ms	21.0 ms	1
curve premium: cy.fit() x 20 (store scan)	curved	3.69 ms	3.69 ms	3.69 ms	3.69 ms	1
curve premium: cy.fit() x 20 (store scan)	straight	3.38 ms	3.38 ms	3.38 ms	3.38 ms	1
curve premium: elementsInBox x 20 (half the graph)	curved	24.3 ms	24.3 ms	24.3 ms	24.3 ms	1
curve premium: elementsInBox x 20 (half the graph)	straight	6.03 ms	6.03 ms	6.03 ms	6.03 ms	1
curve premium: edge.hide/show x 1000 (bundle re-fan)	curved	3.72 ms	3.72 ms	3.72 ms	3.72 ms	1
curve premium: edge.hide/show x 1000 (bundle re-fan)	straight	2.29 ms	2.29 ms	2.29 ms	2.29 ms	1

arrows N=2,000 · 8.7 s

arrow premium: build (style apply + arrow fold) — arrowed

6.68 ms

arrow premium: build (style apply + arrow fold) — none

6.34 ms

arrow premium: cy.style() re-apply x 4 — arrowed

7.95 ms

arrow premium: cy.style() re-apply x 4 — none

7.99 ms

arrow premium: node.position x 200000 (drag) — arrowed

21.7 ms

arrow premium: node.position x 200000 (drag) — none

21.8 ms

arrow premium: edge.sourceEndpoint x 200000 — arrowed

62.1 ms

arrow premium: edge.sourceEndpoint x 200000 — none

59.1 ms

arrow premium: edge.targetEndpoint x 200000 — arrowed

59.9 ms

arrow premium: edge.targetEndpoint x 200000 — none

59.7 ms

arrow premium: edge.midpoint x 200000 — arrowed

141 ms

arrow premium: edge.midpoint x 200000 — none

137 ms

arrow premium: edge.style('target-arrow-shape') x 200000 — arrowed

33.3 ms

arrow premium: edge.style('target-arrow-shape') x 200000 — none

27.3 ms

arrow premium: cy.elements().boundingBox() x 20 — arrowed

18.4 ms

arrow premium: cy.elements().boundingBox() x 20 — none

17.7 ms

arrow premium: elementsInBox x 20 (half the graph) — arrowed

5.45 ms

arrow premium: elementsInBox x 20 (half the graph) — none

5.38 ms

arrow premium: control: sourceEndpoint, bezier vs straight — arrowed

74.1 ms

arrow premium: control: sourceEndpoint, bezier vs straight — none

58.7 ms

arrow premium: edge column bytes — B/edge

156 ms

1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
arrow premium: build (style apply + arrow fold)	arrowed	6.68 ms	6.68 ms	6.68 ms	6.68 ms	1
arrow premium: build (style apply + arrow fold)	none	6.34 ms	6.34 ms	6.34 ms	6.34 ms	1
arrow premium: cy.style() re-apply x 4	arrowed	7.95 ms	7.95 ms	7.95 ms	7.95 ms	1
arrow premium: cy.style() re-apply x 4	none	7.99 ms	7.99 ms	7.99 ms	7.99 ms	1
arrow premium: node.position x 200000 (drag)	arrowed	21.7 ms	21.7 ms	21.7 ms	21.7 ms	1
arrow premium: node.position x 200000 (drag)	none	21.8 ms	21.8 ms	21.8 ms	21.8 ms	1
arrow premium: edge.sourceEndpoint x 200000	arrowed	62.1 ms	62.1 ms	62.1 ms	62.1 ms	1
arrow premium: edge.sourceEndpoint x 200000	none	59.1 ms	59.1 ms	59.1 ms	59.1 ms	1
arrow premium: edge.targetEndpoint x 200000	arrowed	59.9 ms	59.9 ms	59.9 ms	59.9 ms	1
arrow premium: edge.targetEndpoint x 200000	none	59.7 ms	59.7 ms	59.7 ms	59.7 ms	1
arrow premium: edge.midpoint x 200000	arrowed	141 ms	141 ms	141 ms	141 ms	1
arrow premium: edge.midpoint x 200000	none	137 ms	137 ms	137 ms	137 ms	1
arrow premium: edge.style('target-arrow-shape') x 200000	arrowed	33.3 ms	33.3 ms	33.3 ms	33.3 ms	1
arrow premium: edge.style('target-arrow-shape') x 200000	none	27.3 ms	27.3 ms	27.3 ms	27.3 ms	1
arrow premium: cy.elements().boundingBox() x 20	arrowed	18.4 ms	18.4 ms	18.4 ms	18.4 ms	1
arrow premium: cy.elements().boundingBox() x 20	none	17.7 ms	17.7 ms	17.7 ms	17.7 ms	1
arrow premium: elementsInBox x 20 (half the graph)	arrowed	5.45 ms	5.45 ms	5.45 ms	5.45 ms	1
arrow premium: elementsInBox x 20 (half the graph)	none	5.38 ms	5.38 ms	5.38 ms	5.38 ms	1
arrow premium: control: sourceEndpoint, bezier vs straight	arrowed	74.1 ms	74.1 ms	74.1 ms	74.1 ms	1
arrow premium: control: sourceEndpoint, bezier vs straight	none	58.7 ms	58.7 ms	58.7 ms	58.7 ms	1
arrow premium: edge column bytes	B/edge	156 ms	156 ms	156 ms	156 ms	1

labels N=20,000 · 1.8 s

labels: breakLines x 20000 — gpu

85.8 ms

labels: estimateBlock x 20000 — gpu

89.0 ms

labels: addNode x 20000 — gpu

28.4 ms

labels: setLabel (build) x 20000 — gpu

97.1 ms

labels: setLabel (rewrite) x 20000 — gpu

101 ms

labels: boundingBox with 20000 label terms — gpu

3.05 ms

labels: computeSdf tier 1 (30x42) x 96 glyphs — gpu

14.3 ms

labels: computeSdf tier 2 (59x85) x 96 glyphs — gpu

38.6 ms

1 ms10 ms100 ms1 s

All stats (table view)

group	bench	p50	avg	p99	min	samples
labels: breakLines x 20000	gpu	85.8 ms	85.8 ms	85.8 ms	85.8 ms	1
labels: estimateBlock x 20000	gpu	89.0 ms	89.0 ms	89.0 ms	89.0 ms	1
labels: addNode x 20000	gpu	28.4 ms	28.4 ms	28.4 ms	28.4 ms	1
labels: setLabel (build) x 20000	gpu	97.1 ms	97.1 ms	97.1 ms	97.1 ms	1
labels: setLabel (rewrite) x 20000	gpu	101 ms	101 ms	101 ms	101 ms	1
labels: boundingBox with 20000 label terms	gpu	3.05 ms	3.05 ms	3.05 ms	3.05 ms	1
labels: computeSdf tier 1 (30x42) x 96 glyphs	gpu	14.3 ms	14.3 ms	14.3 ms	14.3 ms	1
labels: computeSdf tier 2 (59x85) x 96 glyphs	gpu	38.6 ms	38.6 ms	38.6 ms	38.6 ms	1

transitions N=2,000 · 28.8 s

transition: auto-extent shift, whole-channel re-derive (2000 slots) — transitions off

1.16 ms

transition: auto-extent shift, whole-channel re-derive (2000 slots) — transitions on

3.52 ms

transition: bulk tween tick (2000 slots, one manager tick) — tick

128 µs

transition: explicit-domain write (O(changed) — one element) — transitions off

14.4 µs

transition: explicit-domain write (O(changed) — one element) — transitions on

18.2 µs

transition: sheet swap, whole-graph apply (2000 slots diff + tween) — transitions off

2.05 ms

transition: sheet swap, whole-graph apply (2000 slots diff + tween) — transitions on

4.75 ms

10 µs100 µs1 ms10 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
transition: auto-extent shift, whole-channel re-derive (2000 slots)	transitions off	1.16 ms	1.25 ms	2.77 ms	1.10 ms	558
transition: auto-extent shift, whole-channel re-derive (2000 slots)	transitions on	3.52 ms	3.72 ms	5.46 ms	3.30 ms	182
transition: bulk tween tick (2000 slots, one manager tick)	tick	128 µs	132 µs	222 µs	123 µs	5300
transition: explicit-domain write (O(changed) — one element)	transitions off	14.4 µs	14.9 µs	16.2 µs	13.9 µs	12
transition: explicit-domain write (O(changed) — one element)	transitions on	18.2 µs	18.5 µs	19.4 µs	17.9 µs	12
transition: sheet swap, whole-graph apply (2000 slots diff + tween)	transitions off	2.05 ms	2.08 ms	2.45 ms	2.00 ms	334
transition: sheet swap, whole-graph apply (2000 slots diff + tween)	transitions on	4.75 ms	4.41 ms	6.57 ms	3.46 ms	154

geometry-tween N=2,000 · 32.6 s

size tick: unlabelled (2000 slots) — tick

567 µs

size tick: labelled, center/center (re-anchor early-out) (2000 slots) — tick

588 µs

size tick: labelled, hanging (sidecar rewrite per tick) (2000 slots) — tick

1.24 ms

edge width tick (4000 slots) — bare

276 µs

edge width tick (4000 slots) — with rides (casing + match-line arrows)

712 µs

padding tick + auto-bounds flush (250 parents × 8 children) — tick + flush

177 µs

font-size tick (2000 labels) — wrap none (dims scale-patch)

369 µs

font-size tick (2000 labels) — wrapped (re-break per tick)

5.54 ms

baseline: paint (opacity) tick (2000 slots) — tick

146 µs

100 µs1 ms10 ms

All stats (table view)

group	bench	p50	avg	p99	min	samples
size tick: unlabelled (2000 slots)	tick	567 µs	601 µs	1.18 ms	549 µs	1144
size tick: labelled, center/center (re-anchor early-out) (2000 slots)	tick	588 µs	619 µs	1.05 ms	570 µs	1112
size tick: labelled, hanging (sidecar rewrite per tick) (2000 slots)	tick	1.24 ms	1.31 ms	2.21 ms	1.10 ms	516
edge width tick (4000 slots)	bare	276 µs	284 µs	397 µs	267 µs	2442
edge width tick (4000 slots)	with rides (casing + match-line arrows)	712 µs	721 µs	960 µs	680 µs	964
padding tick + auto-bounds flush (250 parents × 8 children)	tick + flush	177 µs	183 µs	358 µs	170 µs	3800
font-size tick (2000 labels)	wrap none (dims scale-patch)	369 µs	385 µs	891 µs	320 µs	1796
font-size tick (2000 labels)	wrapped (re-break per tick)	5.54 ms	5.58 ms	6.08 ms	5.16 ms	119
baseline: paint (opacity) tick (2000 slots)	tick	146 µs	151 µs	248 µs	141 µs	4612