Time per atlas
Median wall-clock time of one complete size search. Logarithmic scale; shorter is faster.
Benchmarks
Tedd.RectanglePacker and other .NET packers, measured.
Four comparisons on the same machine and the same inputs: choosing an atlas size, inserting into the same fixed bin, inserting into a search-selected square for each packer, and writing pixels into an atlas. Size-search charts show time alongside packing density.
4 of 5datasets where MaxRects Balanced, the default, produced the densest atlas among the size-searching libraries
0.6 % vs 2.9 %mean unused atlas area with default settings, versus the densest other library on each dataset
12.6×speed of MaxRects BestShortSideFit versus the quickest other implementation of that heuristic, one thread in the same fixed bin; geometric mean over 5 datasets (range 1.8× to 292×)
1 · Atlas size search
Choosing an atlas for a set of rectangles.
Each library receives the rectangles and chooses the atlas size itself. Where a library takes a size limit it is 16,384 px per side; rotation is off. Occupancy is the rectangle area divided by the bounding box of all placements. Tedd.RectanglePacker is shown at several efforts and thread counts.
Sprites, 1,000
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Atlas | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| Tedd.RectanglePacker | Skyline Fast · 1 thread | 343 µsmiddle 50%: 327 µs–353 µs | 2,157 × 2,321 | 92.86 % | 32 B |
| Tedd.RectanglePacker | MaxRects Fast · 1 thread | 384 µsmiddle 50%: 368 µs–406 µs | 2,157 × 2,183 | 98.73 % | 32 B |
| Tedd.RectanglePacker | MaxRects Balanced · 32 threads | 3.65 msmiddle 50%: 3.62 ms–3.67 ms | 1,833 × 2,567 | 98.80 % | 3 B |
| Tedd.RectanglePacker | Skyline Balanced · 32 threads | 6.44 msmiddle 50%: 6.36 ms–6.49 ms | 3,235 × 1,520 | 94.55 % | 8 B |
| RectpackSharp | FindBest · 1.2.0 · 1 thread | 11.9 msmiddle 50%: 11.1 ms–13.5 ms | 2,205 × 2,204 | 95.66 % | 48.8 KiB |
| SimulationTree.BinPacker | RecursivePacker FindBest (16384 cap) · 0.3.9 · 1 thread | 12.0 msmiddle 50%: 11.5 ms–12.8 ms | 2,205 × 2,204 | 95.66 % | 47.8 KiB |
| Tedd.RectanglePacker | MaxRects Balanced · 1 thread | 20.3 msmiddle 50%: 19.7 ms–20.7 ms | 1,833 × 2,567 | 98.80 % | 2 B |
| GoldMeridian.RectPack | rectpack2D best packing · 1.0.0 · 1 thread | 28.7 msmiddle 50%: 28.1 ms–29.8 ms | 2,187 × 2,253 | 94.35 % | 4 B |
| Tedd.RectanglePacker | MaxRects Thorough · 32 threads | 685 msmiddle 50%: 680 ms–692 ms | 1,725 × 2,723 | 98.98 % | 165 B |
| ManaBattery.MB.RectangleBinPack | TextureBinPacker · 0.0.4 · 1 thread | 18.52 s1 sample | 2,188 × 2,187 | 97.16 % | 3.2 MiB |
| RectanglePacker | MapperOptimalEfficiency · 1.1.0 · 1 thread | Failed: IndexOutOfRangeException: Index was outside the bounds of the array. | |||
Glyphs, 2,000
Time per atlas
Median wall-clock time of one complete size search. Logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Atlas | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| Tedd.RectanglePacker | MaxRects Fast · 1 thread | 251 µsmiddle 50%: 229 µs–272 µs | 931 × 937 | 99.18 % | 32 B |
| Tedd.RectanglePacker | Skyline Fast · 1 thread | 258 µsmiddle 50%: 252 µs–261 µs | 931 × 957 | 97.11 % | 32 B |
| Tedd.RectanglePacker | MaxRects Balanced · 32 threads | 8.95 msmiddle 50%: 8.88 ms–9.05 ms | 1,210 × 720 | 99.31 % | 8 B |
| Tedd.RectanglePacker | Skyline Balanced · 32 threads | 11.9 msmiddle 50%: 11.8 ms–12.0 ms | 1,396 × 626 | 99.00 % | 12 B |
| RectpackSharp | FindBest · 1.2.0 · 1 thread | 20.5 msmiddle 50%: 19.7 ms–21.0 ms | 961 × 988 | 91.12 % | 95.4 KiB |
| SimulationTree.BinPacker | RecursivePacker FindBest (16384 cap) · 0.3.9 · 1 thread | 21.8 msmiddle 50%: 21.1 ms–23.1 ms | 961 × 988 | 91.12 % | 95.1 KiB |
| Tedd.RectanglePacker | MaxRects Balanced · 1 thread | 41.9 msmiddle 50%: 40.6 ms–42.6 ms | 1,210 × 720 | 99.31 % | 3 B |
| GoldMeridian.RectPack | rectpack2D best packing · 1.0.0 · 1 thread | 69.3 msmiddle 50%: 65.9 ms–73.9 ms | 993 × 996 | 87.48 % | 4 B |
| Tedd.RectanglePacker | MaxRects Thorough · 32 threads | 104 msmiddle 50%: 102 ms–107 ms | 838 × 1,036 | 99.66 % | 2.8 KiB |
| ManaBattery.MB.RectangleBinPack | TextureBinPacker · 0.0.4 · 1 thread | Skipped: more than 1,000 rectangles (one run on 2,000 rectangles did not finish within five minutes) | |||
| RectanglePacker | MapperOptimalEfficiency · 1.1.0 · 1 thread | Failed: IndexOutOfRangeException: Index was outside the bounds of the array. | |||
UI images, 500
Time per atlas
Median wall-clock time of one complete size search. Logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Atlas | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| Tedd.RectanglePacker | Skyline Fast · 1 thread | 114 µsmiddle 50%: 112 µs–115 µs | 1,974 × 2,272 | 86.84 % | 32 B |
| Tedd.RectanglePacker | MaxRects Fast · 1 thread | 168 µsmiddle 50%: 167 µs–170 µs | 1,974 × 1,997 | 98.80 % | 32 B |
| Tedd.RectanglePacker | MaxRects Balanced · 32 threads | 1.43 msmiddle 50%: 1.30 ms–1.52 ms | 2,270 × 1,736 | 98.84 % | 1 B |
| Tedd.RectanglePacker | Skyline Balanced · 32 threads | 2.41 msmiddle 50%: 2.13 ms–2.56 ms | 2,961 × 1,438 | 91.47 % | 2 B |
| SimulationTree.BinPacker | RecursivePacker FindBest (16384 cap) · 0.3.9 · 1 thread | 3.33 msmiddle 50%: 3.30 ms–3.43 ms | 2,001 × 1,990 | 97.81 % | 23.6 KiB |
| RectpackSharp | FindBest · 1.2.0 · 1 thread | 3.44 msmiddle 50%: 3.37 ms–3.46 ms | 2,001 × 1,990 | 97.81 % | 23.9 KiB |
| Tedd.RectanglePacker | MaxRects Balanced · 1 thread | 5.96 msmiddle 50%: 5.84 ms–6.00 ms | 2,270 × 1,736 | 98.84 % | 0 B |
| GoldMeridian.RectPack | rectpack2D best packing · 1.0.0 · 1 thread | 9.61 msmiddle 50%: 8.05 ms–10.2 ms | 2,090 × 1,916 | 97.26 % | 0 B |
| Tedd.RectanglePacker | MaxRects Thorough · 32 threads | 124 msmiddle 50%: 117 ms–128 ms | 1,382 × 2,840 | 99.24 % | 5.6 KiB |
| ManaBattery.MB.RectangleBinPack | TextureBinPacker · 0.0.4 · 1 thread | 1.05 smiddle 50%: 1.04 s–1.05 s | 1,982 × 1,982 | 99.15 % | 1.1 MiB |
| RectanglePacker | MapperOptimalEfficiency · 1.1.0 · 1 thread | Failed: ArgumentException: Destination array was not long enough. Check the destination index, length, and the array's lower bounds. (Parameter 'destinationArray') | |||
Tiles, 4,096
Time per atlas
Median wall-clock time of one complete size search. Logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Atlas | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| Tedd.RectanglePacker | Skyline Fast · 1 thread | 540 µsmiddle 50%: 534 µs–553 µs | 2,128 × 2,160 | 99.44 % | 32 B |
| Tedd.RectanglePacker | MaxRects Fast · 1 thread | 588 µsmiddle 50%: 578 µs–591 µs | 2,128 × 2,160 | 99.44 % | 32 B |
| Tedd.RectanglePacker | MaxRects Balanced · 32 threads | 1.43 msmiddle 50%: 1.39 ms–1.58 ms | 1,808 × 2,528 | 100.00 % | 1 B |
| Tedd.RectanglePacker | Skyline Balanced · 32 threads | 1.43 msmiddle 50%: 1.38 ms–1.47 ms | 1,808 × 2,528 | 100.00 % | 1 B |
| Tedd.RectanglePacker | MaxRects Balanced · 1 thread | 3.95 msmiddle 50%: 3.84 ms–4.02 ms | 1,808 × 2,528 | 100.00 % | 0 B |
| Tedd.RectanglePacker | MaxRects Thorough · 32 threads | 8.51 msmiddle 50%: 8.47 ms–8.68 ms | 1,808 × 2,528 | 100.00 % | 7 B |
| GoldMeridian.RectPack | rectpack2D best packing · 1.0.0 · 1 thread | 21.3 msmiddle 50%: 19.0 ms–23.1 ms | 2,144 × 2,144 | 99.43 % | 1 B |
| SimulationTree.BinPacker | RecursivePacker FindBest (16384 cap) · 0.3.9 · 1 thread | 43.5 msmiddle 50%: 40.1 ms–47.9 ms | 2,144 × 2,144 | 99.43 % | 196.4 KiB |
| RectpackSharp | FindBest · 1.2.0 · 1 thread | 46.4 msmiddle 50%: 43.5 ms–48.4 ms | 2,144 × 2,144 | 99.43 % | 198.8 KiB |
| ManaBattery.MB.RectangleBinPack | TextureBinPacker · 0.0.4 · 1 thread | Skipped: more than 1,000 rectangles (one run on 2,000 rectangles did not finish within five minutes) | |||
| RectanglePacker | MapperOptimalEfficiency · 1.1.0 · 1 thread | Failed: IndexOutOfRangeException: Index was outside the bounds of the array. | |||
Sprites, 10,000
Time per atlas
Median wall-clock time of one complete size search. Logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Atlas | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| Tedd.RectanglePacker | Skyline Fast · 1 thread | 1.74 msmiddle 50%: 1.66 ms–1.78 ms | 3,421 × 3,448 | 99.18 % | 32 B |
| Tedd.RectanglePacker | MaxRects Fast · 1 thread | 1.79 msmiddle 50%: 1.72 ms–1.90 ms | 3,421 × 3,422 | 99.93 % | 32 B |
| Tedd.RectanglePacker | MaxRects Balanced · 32 threads | 58.7 msmiddle 50%: 58.6 ms–59.7 ms | 3,421 × 3,422 | 99.93 % | 48 B |
| Tedd.RectanglePacker | Skyline Balanced · 32 threads | 177 msmiddle 50%: 168 ms–183 ms | 5,131 × 2,291 | 99.52 % | 96 B |
| Tedd.RectanglePacker | MaxRects Balanced · 1 thread | 348 msmiddle 50%: 343 ms–350 ms | 3,421 × 3,422 | 99.93 % | 9 B |
| RectpackSharp | FindBest · 1.2.0 · 1 thread | 621 msmiddle 50%: 588 ms–629 ms | 3,443 × 3,444 | 98.66 % | 500.4 KiB |
| SimulationTree.BinPacker | RecursivePacker FindBest (16384 cap) · 0.3.9 · 1 thread | 664 msmiddle 50%: 659 ms–684 ms | 3,443 × 3,444 | 98.66 % | 500.1 KiB |
| Tedd.RectanglePacker | MaxRects Thorough · 32 threads | 1.96 smiddle 50%: 1.95 s–1.96 s | 5,986 × 1,955 | 99.97 % | 262.5 KiB |
| GoldMeridian.RectPack | rectpack2D best packing · 1.0.0 · 1 thread | 1.98 smiddle 50%: 1.92 s–2.00 s | 3,499 × 3,500 | 95.53 % | 16 B |
| ManaBattery.MB.RectangleBinPack | TextureBinPacker · 0.0.4 · 1 thread | Skipped: more than 1,000 rectangles (one run on 2,000 rectangles did not finish within five minutes) | |||
| RectanglePacker | MapperOptimalEfficiency · 1.1.0 · 1 thread | Failed: IndexOutOfRangeException: Index was outside the bounds of the array. | |||
2 · Common fixed bin
Same rectangles, same bin.
Every packer receives the same sequence, sorted by descending height
then width, and the same square side: the greater of the largest
input dimension and ceil(sqrt(total area) × 1.25).
One complete insertion pass is timed. Failures remain visible;
occupancy is constant for successful packers on each dataset.
Sprites, 1,000
Time in the same fixed bin
One insertion pass into a 2,696 × 2,696 square for every packer. Median wall-clock time; logarithmic scale; shorter is faster.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 91 µsmiddle 50%: 86 µs–99 µs | 2,696 | 63.96 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 275 µsmiddle 50%: 266 µs–297 µs | 2,696 | 63.96 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 289 µsmiddle 50%: 267 µs–301 µs | 2,696 | 63.96 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 509 µsmiddle 50%: 449 µs–519 µs | 2,696 | 63.96 % | 16.2 KiB |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 573 µsmiddle 50%: 536 µs–677 µs | 2,696 | 63.96 % | 0 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 1.69 msmiddle 50%: 1.46 ms–1.78 ms | 2,696 | 63.96 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 2.36 msmiddle 50%: 2.08 ms–2.46 ms | 2,696 | 63.96 % | 0 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 3.07 msmiddle 50%: 2.78 ms–3.29 ms | 2,696 | 63.96 % | 186.7 KiB |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 4.95 msmiddle 50%: 4.92 ms–5.37 ms | 2,696 | 63.96 % | 112 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 26.7 msmiddle 50%: 24.9 ms–30.2 ms | 2,696 | 63.96 % | 2 B |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 525 msmiddle 50%: 504 ms–532 ms | 2,696 | 63.96 % | 9 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 567 msmiddle 50%: 540 ms–574 ms | 2,696 | 63.96 % | 9 B |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | Failed: PackingFailedException: Input 962 did not fit. | |||
Glyphs, 2,000
Time in the same fixed bin
One insertion pass into a 1,163 × 1,163 square for every packer. Median wall-clock time; logarithmic scale; shorter is faster.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 45 µsmiddle 50%: 40 µs–47 µs | 1,163 | 63.97 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 91 µsmiddle 50%: 89 µs–112 µs | 1,163 | 63.97 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 164 µsmiddle 50%: 151 µs–178 µs | 1,163 | 63.97 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 169 µsmiddle 50%: 158 µs–184 µs | 1,163 | 63.97 % | 2.1 KiB |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 188 µsmiddle 50%: 174 µs–199 µs | 1,163 | 63.97 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 293 µsmiddle 50%: 260 µs–330 µs | 1,163 | 63.97 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 1.32 msmiddle 50%: 1.20 ms–1.44 ms | 1,163 | 63.97 % | 0 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 2.78 msmiddle 50%: 2.54 ms–2.95 ms | 1,163 | 63.97 % | 0 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 3.49 msmiddle 50%: 3.30 ms–3.59 ms | 1,163 | 63.97 % | 0 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 10.9 msmiddle 50%: 10.4 ms–11.2 ms | 1,163 | 63.97 % | 359.0 KiB |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 13.3 msmiddle 50%: 12.6 ms–13.6 ms | 1,163 | 63.97 % | 113 B |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 1.14 smiddle 50%: 1.13 s–1.15 s | 1,163 | 63.97 % | 16 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 1.17 smiddle 50%: 1.16 s–1.19 s | 1,163 | 63.97 % | 16 B |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
UI images, 500
Time in the same fixed bin
One insertion pass into a 2,467 × 2,467 square for every packer. Median wall-clock time; logarithmic scale; shorter is faster.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 43 µsmiddle 50%: 40 µs–44 µs | 2,467 | 64.00 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 111 µsmiddle 50%: 110 µs–113 µs | 2,467 | 64.00 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 130 µsmiddle 50%: 115 µs–137 µs | 2,467 | 64.00 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 152 µsmiddle 50%: 147 µs–156 µs | 2,467 | 64.00 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 181 µsmiddle 50%: 172 µs–184 µs | 2,467 | 64.00 % | 8.1 KiB |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 346 µsmiddle 50%: 323 µs–352 µs | 2,467 | 64.00 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 837 µsmiddle 50%: 835 µs–839 µs | 2,467 | 64.00 % | 0 B |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 1.39 msmiddle 50%: 1.24 ms–1.45 ms | 2,467 | 64.00 % | 112 B |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 38.1 msmiddle 50%: 36.1 ms–40.5 ms | 2,467 | 64.00 % | 3 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 50.6 msmiddle 50%: 48.9 ms–51.2 ms | 2,467 | 64.00 % | 4 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | Failed: PackingFailedException: Insert returned no placement. | |||
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | Failed: PackingFailedException: Input 37 did not fit. | |||
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | Failed: PackingFailedException: Input 263 did not fit. | |||
Tiles, 4,096
Time in the same fixed bin
One insertion pass into a 2,673 × 2,673 square for every packer. Median wall-clock time; logarithmic scale; shorter is faster.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 74 µsmiddle 50%: 74 µs–76 µs | 2,673 | 63.97 % | 0 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 125 µsmiddle 50%: 113 µs–140 µs | 2,673 | 63.97 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 218 µsmiddle 50%: 173 µs–226 µs | 2,673 | 63.97 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 249 µsmiddle 50%: 233 µs–271 µs | 2,673 | 63.97 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 272 µsmiddle 50%: 254 µs–294 µs | 2,673 | 63.97 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 302 µsmiddle 50%: 261 µs–322 µs | 2,673 | 63.97 % | 0 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 477 µsmiddle 50%: 448 µs–549 µs | 2,673 | 63.97 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 581 µsmiddle 50%: 574 µs–610 µs | 2,673 | 63.97 % | 2.1 KiB |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 587 µsmiddle 50%: 558 µs–606 µs | 2,673 | 63.97 % | 0 B |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 639 µsmiddle 50%: 632 µs–644 µs | 2,673 | 63.97 % | 0 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 1.91 msmiddle 50%: 1.89 ms–1.92 ms | 2,673 | 63.97 % | 0 B |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 3.84 msmiddle 50%: 3.80 ms–3.88 ms | 2,673 | 63.97 % | 112 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 43.8 msmiddle 50%: 41.8 ms–46.8 ms | 2,673 | 63.97 % | 830.7 KiB |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
Sprites, 10,000
Time in the same fixed bin
One insertion pass into a 4,276 × 4,276 square for every packer. Median wall-clock time; logarithmic scale; shorter is faster.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 237 µsmiddle 50%: 226 µs–244 µs | 4,276 | 63.98 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 588 µsmiddle 50%: 568 µs–652 µs | 4,276 | 63.98 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 1.07 msmiddle 50%: 1.06 ms–1.08 ms | 4,276 | 63.98 % | 0 B |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 1.70 msmiddle 50%: 1.67 ms–1.76 ms | 4,276 | 63.98 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 2.14 msmiddle 50%: 2.09 ms–2.19 ms | 4,276 | 63.98 % | 4.1 KiB |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 8.66 msmiddle 50%: 8.15 ms–8.86 ms | 4,276 | 63.98 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 19.8 msmiddle 50%: 19.3 ms–19.9 ms | 4,276 | 63.98 % | 1 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 40.3 msmiddle 50%: 38.1 ms–41.3 ms | 4,276 | 63.98 % | 3 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 274 msmiddle 50%: 260 ms–291 ms | 4,276 | 63.98 % | 1.9 MiB |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 352 msmiddle 50%: 344 ms–371 ms | 4,276 | 63.98 % | 121 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 1.89 smiddle 50%: 1.85 s–1.90 s | 4,276 | 63.98 % | 16 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | Skipped: more than 4,096 rectangles (one pass over 2,000 rectangles takes over a second, and the bisection needs a dozen passes) | |||
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | Skipped: more than 4,096 rectangles (one pass over 2,000 rectangles takes over a second, and the bisection needs a dozen passes) | |||
3 · Square search
Heuristic-selected square per packer.
An untimed growth search and bisection select a successful square for each packer. Greedy packing success need not be monotone in bin size, so this does not establish a global minimum. One insertion pass into that selected square is timed. Bin sizes and insertion difficulty differ; compare density alongside time.
Sprites, 1,000
Time per insertion pass
One insertion pass into each packer's search-selected square. Bin sizes differ; logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 20 µsmiddle 50%: 17 µs–23 µs | 2,736 | 62.11 % | 0 B |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 100 µsmiddle 50%: 98 µs–105 µs | 2,304 | 87.58 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 268 µsmiddle 50%: 247 µs–275 µs | 2,223 | 94.08 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 311 µsmiddle 50%: 297 µs–326 µs | 2,169 | 98.82 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 527 µsmiddle 50%: 507 µs–563 µs | 2,310 | 87.12 % | 16.2 KiB |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 765 µsmiddle 50%: 724 µs–784 µs | 2,197 | 96.32 % | 0 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 1.41 msmiddle 50%: 1.36 ms–1.44 ms | 2,183 | 97.56 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 2.21 msmiddle 50%: 2.11 ms–2.27 ms | 2,568 | 70.50 % | 0 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 3.81 msmiddle 50%: 3.63 ms–3.91 ms | 2,235 | 93.07 % | 187.2 KiB |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 5.14 msmiddle 50%: 5.00 ms–5.22 ms | 2,194 | 96.58 % | 112 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 12.6 msmiddle 50%: 12.3 ms–13.0 ms | 2,192 | 96.76 % | 1 B |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 583 msmiddle 50%: 571 ms–596 ms | 2,197 | 96.32 % | 9 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 640 msmiddle 50%: 630 ms–655 ms | 2,197 | 96.32 % | 9 B |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
Glyphs, 2,000
Time per insertion pass
One insertion pass into each packer's search-selected square. Bin sizes differ; logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 52 µsmiddle 50%: 52 µs–52 µs | 975 | 91.01 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 125 µsmiddle 50%: 116 µs–133 µs | 939 | 98.12 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 146 µsmiddle 50%: 145 µs–146 µs | 939 | 98.12 % | 4.1 KiB |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 164 µsmiddle 50%: 164 µs–164 µs | 1,000 | 86.52 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 198 µsmiddle 50%: 191 µs–203 µs | 935 | 98.97 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 397 µsmiddle 50%: 355 µs–411 µs | 942 | 97.50 % | 0 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 1.15 msmiddle 50%: 1.11 ms–1.20 ms | 939 | 98.12 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 1.51 msmiddle 50%: 1.43 ms–1.54 ms | 952 | 95.46 % | 0 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 3.20 msmiddle 50%: 3.10 ms–3.28 ms | 950 | 95.86 % | 0 B |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 9.47 msmiddle 50%: 9.25 ms–9.71 ms | 938 | 98.33 % | 112 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 10.1 msmiddle 50%: 9.65 ms–11.0 ms | 975 | 91.01 % | 359.5 KiB |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 885 msmiddle 50%: 833 ms–1.09 s | 952 | 95.46 % | 9 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 1.20 smiddle 50%: 1.18 s–1.21 s | 952 | 95.46 % | 16 B |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
UI images, 500
Time per insertion pass
One insertion pass into each packer's search-selected square. Bin sizes differ; logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 9 µsmiddle 50%: 9 µs–9 µs | 2,631 | 56.27 % | 0 B |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 37 µsmiddle 50%: 37 µs–37 µs | 2,140 | 85.05 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 51 µsmiddle 50%: 51 µs–52 µs | 2,122 | 86.50 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 94 µsmiddle 50%: 94 µs–94 µs | 1,985 | 98.85 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 98 µsmiddle 50%: 97 µs–98 µs | 1,986 | 98.75 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 124 µsmiddle 50%: 123 µs–124 µs | 2,370 | 69.34 % | 8.1 KiB |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 193 µsmiddle 50%: 192 µs–193 µs | 2,096 | 88.66 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 388 µsmiddle 50%: 387 µs–388 µs | 2,248 | 77.07 % | 0 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 580 µsmiddle 50%: 579 µs–584 µs | 2,870 | 47.29 % | 91.2 KiB |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 1.03 msmiddle 50%: 984 µs–1.05 ms | 2,112 | 87.32 % | 112 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 23.8 msmiddle 50%: 23.7 ms–23.8 ms | 3,390 | 33.89 % | 1 B |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 32.7 msmiddle 50%: 32.6 ms–32.7 ms | 1,986 | 98.75 % | 2 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 35.7 msmiddle 50%: 35.3 ms–38.0 ms | 1,986 | 98.75 % | 3 B |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
Tiles, 4,096
Time per insertion pass
One insertion pass into each packer's search-selected square. Bin sizes differ; logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 68 µsmiddle 50%: 58 µs–73 µs | 2,144 | 99.43 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 91 µsmiddle 50%: 91 µs–91 µs | 2,144 | 99.43 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 94 µsmiddle 50%: 85 µs–96 µs | 2,144 | 99.43 % | 1.1 KiB |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 123 µsmiddle 50%: 113 µs–128 µs | 2,144 | 99.43 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 156 µsmiddle 50%: 148 µs–165 µs | 2,144 | 99.43 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 202 µsmiddle 50%: 201 µs–202 µs | 2,144 | 99.43 % | 0 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 215 µsmiddle 50%: 215 µs–215 µs | 2,144 | 99.43 % | 0 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 320 µsmiddle 50%: 281 µs–360 µs | 2,144 | 99.43 % | 0 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | 322 µsmiddle 50%: 280 µs–337 µs | 2,144 | 99.43 % | 0 B |
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | 334 µsmiddle 50%: 296 µs–349 µs | 2,144 | 99.43 % | 0 B |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 656 µsmiddle 50%: 617 µs–723 µs | 2,160 | 97.96 % | 0 B |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 3.21 msmiddle 50%: 3.01 ms–3.34 ms | 2,144 | 99.43 % | 112 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 47.6 msmiddle 50%: 42.8 ms–55.8 ms | 2,144 | 99.43 % | 830.8 KiB |
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
Sprites, 10,000
Time per insertion pass
One insertion pass into each packer's search-selected square. Bin sizes differ; logarithmic scale; shorter is faster.
Unused atlas area
Share of the atlas not covered by rectangles. Linear scale; shorter is denser. Rows are in the same order as the time chart.
| Library | Configuration | Median time | Square side | Occupancy | Allocated per call |
|---|---|---|---|---|---|
| GoldMeridian.RectPack | EmptySpaces (rectpack2D) · 1.0.0 · 1 thread | 466 µsmiddle 50%: 415 µs–497 µs | 3,641 | 88.25 % | 0 B |
| Tedd.RectanglePacker | Skyline BottomLeft · 1 thread | 733 µsmiddle 50%: 602 µs–774 µs | 3,436 | 99.09 % | 0 B |
| Tedd.RectanglePacker | MaxRects BottomLeft · 1 thread | 925 µsmiddle 50%: 873 µs–960 µs | 3,422 | 99.90 % | 0 B |
| RectPacker2 | Packer · 2.1.0 · 1 thread | 1.88 msmiddle 50%: 1.83 ms–2.02 ms | 3,423 | 99.84 % | 8.1 KiB |
| AtlasAllocator | ShelfAtlasAllocator · 1.0.1 · 1 thread | 1.95 msmiddle 50%: 1.85 ms–2.00 ms | 3,702 | 85.36 % | 0 B |
| Tedd.RectanglePacker | Skyline MinWaste · 1 thread | 17.1 msmiddle 50%: 16.5 ms–18.5 ms | 3,498 | 95.61 % | 1 B |
| Tedd.RectanglePacker | MaxRects BestShortSideFit · 1 thread | 24.4 msmiddle 50%: 22.7 ms–26.2 ms | 3,439 | 98.92 % | 1 B |
| RectangleBinPack.CSharp | Guillotine BestAreaFit, MinimizeArea · 1.0.4 · 1 thread | 45.0 msmiddle 50%: 42.6 ms–50.5 ms | 3,441 | 98.80 % | 3 B |
| AtlasAllocator | GuillotineAtlasAllocator · 1.0.1 · 1 thread | 48.2 msmiddle 50%: 45.0 ms–49.6 ms | 3,450 | 98.29 % | 3 B |
| RectangleBinPacking | MaxRects BestShortSideFit (new bin) · 1.0.2 · 1 thread | 283 msmiddle 50%: 268 ms–294 ms | 3,527 | 94.04 % | 1.9 MiB |
| StbRectPackSharp | stbrp_pack_rects (skyline) · 1.0.4 · 1 thread | 341 msmiddle 50%: 318 ms–389 ms | 3,429 | 99.50 % | 121 B |
| RectangleBinPack.CSharp | MaxRects BestShortSideFit · 1.0.4 · 1 thread | Skipped: more than 4,096 rectangles (one pass over 2,000 rectangles takes over a second, and the bisection needs a dozen passes) | |||
| RectangleBinPack.CSharp | Skyline BottomLeft · 1.0.4 · 1 thread | Skipped: SkylineBinPack always considers rotated placements; no option disables it | |||
| ManaBattery.MB.RectangleBinPack | MaxRects BestShortSideFit · 0.0.4 · 1 thread | Skipped: more than 4,096 rectangles (one pass over 2,000 rectangles takes over a second, and the bisection needs a dozen passes) | |||
4 · Composition
Writing sprite pixels into the atlas.
Every contender writes the same layout: RGBA32 sprites into a reused
RGBA32 atlas, clearing the background first. Every output was compared
byte for byte with the reference. The plain Span.CopyTo
row loop is a single-thread baseline, not a library.
Sprites, 1,000
Atlas throughput
1,833 × 2,567 RGBA32 atlas, megapixels of atlas written per second including the background. Linear scale; longer is faster.
| Library | Configuration | Median time | MP/s | Allocated per call |
|---|---|---|---|---|
| Tedd.RectanglePacker | AtlasBuilder.Compose · 32 threads | 1.94 msmiddle 50%: 1.88 ms–2.12 ms | 2,426 | 2 B |
| Tedd.RectanglePacker | AtlasBuilder.Compose · 1 thread | 3.51 msmiddle 50%: 3.33 ms–4.46 ms | 1,341 | 208 B |
| SkiaSharp | SKCanvas.DrawImage (Src) · 4.153.0 · 1 thread | 5.60 msmiddle 50%: 5.42 ms–6.28 ms | 840 | 0 B |
| Span.CopyTo loop | reference · 1 thread | 5.62 msmiddle 50%: 5.14 ms–6.50 ms | 837 | 0 B |
| SixLabors.ImageSharp | DrawImage (Src) · 3.1.12 · 1 thread | 47.9 msmiddle 50%: 44.5 ms–50.7 ms | 98 | 7.2 MiB |
UI images, 500
Atlas throughput
2,270 × 1,736 RGBA32 atlas, megapixels of atlas written per second including the background. Linear scale; longer is faster.
| Library | Configuration | Median time | MP/s | Allocated per call |
|---|---|---|---|---|
| Tedd.RectanglePacker | AtlasBuilder.Compose · 32 threads | 1.06 msmiddle 50%: 958 µs–1.15 ms | 3,718 | 1 B |
| Tedd.RectanglePacker | AtlasBuilder.Compose · 1 thread | 2.39 msmiddle 50%: 2.27 ms–2.52 ms | 1,651 | 0 B |
| Span.CopyTo loop | reference · 1 thread | 3.64 msmiddle 50%: 3.25 ms–3.79 ms | 1,083 | 0 B |
| SkiaSharp | SKCanvas.DrawImage (Src) · 4.153.0 · 1 thread | 3.92 msmiddle 50%: 3.39 ms–4.21 ms | 1,004 | 0 B |
| SixLabors.ImageSharp | DrawImage (Src) · 3.1.12 · 1 thread | 24.0 msmiddle 50%: 23.1 ms–24.7 ms | 164 | 3.0 MiB |
Method
How the numbers were produced.
AMD Ryzen 9 5950X 16-Core Processor, 32 logical
processors, Microsoft Windows 10.0.26200, .NET 11.0.0-rc.1.26425.128,
X64. Measured 2026-09-30 (UTC),
library commit 37274944e775.
- Inputs. Five synthetic rectangle sets generated from seed
20260929. Every library receives identical sizes. - Sprites, 1,000 (
sprites-1k): 1,000 sprites, width and height uniform in 8-128 px; total area 4,649,054 px². - Glyphs, 2,000 (
glyphs-2k): 2,000 font glyphs, width 4-32 px, height 16-32 px; total area 865,179 px². - UI images, 500 (
ui-500): 500 UI images: 400 icons of 8-64 px and 100 panels of 64-384 x 32-256 px; total area 3,894,874 px². - Tiles, 4,096 (
tiles-4k): 4,096 square tiles of 16, 32 or 64 px (50/35/15 %); total area 4,570,624 px². - Sprites, 10,000 (
sprites-10k): 10,000 small sprites, width and height uniform in 4-64 px; total area 11,698,731 px². - Timing. The process first runs 6 s of packing load. Each configuration is then warmed up and timed in 15 samples of at least 100 ms each. The median per-call time and middle 50% of sample times are reported. Operations slower than one second use three samples; slower than ten seconds, one. Allocation is the managed allocation on all threads, averaged per call.
- Validation. Runs before and after timing are checked independently: every rectangle placed once at its original size, inside the bin, without overlap. Composition outputs must match the reference byte for byte.
- Settings. Libraries use their documented defaults or recommended heuristics, without rotation. Conversion to each library's input types happens before timing; where a library sorts its input array in place, the per-call copy of that array is timed.
- Threads. Only Tedd.RectanglePacker uses more than one thread. Its single-thread results are listed separately. Every single-thread configuration, of every library, runs with the process pinned to logical processor 2; unpinned single-thread timings on this machine switch between speed levels within one process. Multi-thread configurations use all processors.
- Limits. The results describe these inputs on this machine. Sample medians summarize one process run; slow cases may have only one sample. Other size distributions, especially very uniform or very elongated rectangles, can rank the packers differently.
Excluded or failed configurations:
- RectanglePacker 1.1.0 (MapperOptimalEfficiency), all datasets: Failed: IndexOutOfRangeException: Index was outside the bounds of the array. / Failed: ArgumentException: Destination array was not long enough. Check the destination index, length, and the array's lower bounds. (Parameter 'destinationArray')
- ManaBattery.MB.RectangleBinPack 0.0.4 (TextureBinPacker),
glyphs-2k,tiles-4k,sprites-10k: Skipped: more than 1,000 rectangles (one run on 2,000 rectangles did not finish within five minutes) - RectangleBinPack.CSharp 1.0.4 (Skyline BottomLeft), all datasets: Skipped: SkylineBinPack always considers rotated placements; no option disables it
- RectangleBinPack.CSharp 1.0.4 (MaxRects BestShortSideFit),
sprites-10k: Skipped: more than 4,096 rectangles (one pass over 2,000 rectangles takes over a second, and the bisection needs a dozen passes) - ManaBattery.MB.RectangleBinPack 0.0.4 (MaxRects BestShortSideFit),
sprites-10k: Skipped: more than 4,096 rectangles (one pass over 2,000 rectangles takes over a second, and the bisection needs a dozen passes) - GoldMeridian.RectPack 1.0.0 (EmptySpaces (rectpack2D)),
sprites-1k,ui-500: Failed: PackingFailedException: Input 962 did not fit. / Failed: PackingFailedException: Input 263 did not fit. - RectangleBinPacking 1.0.2 (MaxRects BestShortSideFit (new bin)),
ui-500: Failed: PackingFailedException: Insert returned no placement. - RectangleBinPack.CSharp 1.0.4 (Guillotine BestAreaFit, MinimizeArea),
ui-500: Failed: PackingFailedException: Input 37 did not fit.
Raw CSV files, per-sample timing JSON, the environment record and the measured layouts are in benchmarks/results/2026-09-30-3727494-expanded. The harness is src/Tedd.RectanglePacker.Comparison; the command line is in BENCHMARKS.md.