ManifoldsGPU

Benchmarks

Device: NVIDIA GeForce RTX 5070 Ti, eltype: Float32/ComplexF32

ManifoldOperationCPU median [ms]GPU median [ms]Speedup CPU/GPUError
Euclidean(32, 16, 2048)exp0.350.172.060.0
Euclidean(32, 16, 2048)log!0.350.172.050.0
Euclidean(32, 16, 2048)inner0.190.141.349.357e-8
Euclidean(32, 16, 2048)norm0.130.160.838.423e-8
Euclidean(32, 16, 2048)project!0.230.121.890.0
PowerManifold(Sphere(31), 2048)exp0.050.140.357.092e-8
PowerManifold(Sphere(31), 2048)log!0.080.370.235.125e-8
PowerManifold(Sphere(31), 2048)inner0.020.130.145.86e-7
PowerManifold(Sphere(31), 2048)norm0.020.130.151.064e-7
PowerManifold(Sphere(31), 2048)project!0.030.150.182.819e-8
PowerManifold(Rotations(32), 2048)exp36.192.3515.382.594e-6
PowerManifold(Rotations(32), 2048)log!565.7274.447.69.157e-5
PowerManifold(Rotations(32), 2048)inner0.410.251.654.708e-6
PowerManifold(Rotations(32), 2048)norm1.360.149.531.109e-6
PowerManifold(Rotations(32), 2048)project!20.460.2291.223.644e-7
PowerManifold(Rotations(32), 2048)retract_fused!(PolarRetraction)115.514.8923.622.555e-6
PowerManifold(Rotations(32), 2048)retract_fused!(QRRetraction)90.110.83108.953.204e-7
PowerManifold(UnitaryMatrices(32), 2048)exp85.898.0710.641.957e-6
PowerManifold(UnitaryMatrices(32), 2048)log!729.5769.8910.440.0001844
PowerManifold(UnitaryMatrices(32), 2048)inner0.8356.260.015.979e-5
PowerManifold(UnitaryMatrices(32), 2048)norm1.7444.170.041.516e-6
PowerManifold(UnitaryMatrices(32), 2048)project!31.420.3590.365.512e-7
PowerManifold(Grassmann(32, 16), 2048)exp69.725.2713.227.023e-5
PowerManifold(Grassmann(32, 16), 2048)log!57.993.3617.272.332e-5
PowerManifold(Grassmann(32, 16), 2048)inner0.20.131.548.056e-7
PowerManifold(Grassmann(32, 16), 2048)norm0.810.145.953.696e-7
PowerManifold(Grassmann(32, 16), 2048)project!1.00.234.451.303e-7
PowerManifold(Grassmann(32, 16), 2048)retract_fused!(PolarRetraction)40.672.9913.590.0001873
PowerManifold(Grassmann(32, 16), 2048)retract_fused!(QRRetraction)17.840.7224.634.623e-5
PowerManifold(Stiefel(32, 16), 2048)exp(ExponentialRetraction)70.993.5619.911.164e-6
PowerManifold(Stiefel(32, 16), 2048)retract_fused!(PolarRetraction)43.662.8715.211.37e-6
PowerManifold(Stiefel(32, 16), 2048)retract_fused!(QRRetraction)18.280.7125.891.885e-7