DArray/stencil: Add AntiReflect boundary condition - #741
Open
jpsamaroo wants to merge 1 commit into
Open
Conversation
Reflects indices like Reflect, but negates the reflected values, for quantities that must vanish at the domain edge - most commonly the velocity or momentum component normal to a solid wall, which Reflect turns into an outflow. Usually combined with other BCs via a tuple, since only the wall-normal component flips sign. Co-authored-by: Cursor <cursoragent@cursor.com>
Contributor
Dagger benchmarks:
|
| master | dirty | master / dirty | |
|---|---|---|---|
| array/dagger/N=1024 (block 512)/add (X + X) | 3.59 ± 0.47 ms | 3.47 ± 0.066 ms | 1.03 ± 0.14 |
| array/dagger/N=1024 (block 512)/alloc (rand) | 3.65 ± 0.7 ms | 3.41 ± 0.061 ms | 1.07 ± 0.21 |
| array/dagger/N=1024 (block 512)/broadcast (X .+ 1) | 2.62 ± 0.59 ms | 2.63 ± 0.21 ms | 0.997 ± 0.24 |
| array/dagger/N=1024 (block 512)/map (sin.(X)) | 7.92 ± 0.44 ms | 8.61 ± 0.47 ms | 0.919 ± 0.071 |
| array/dagger/N=1024 (block 512)/norm | 1.16 ± 0.3 ms | 1.17 ± 0.11 ms | 0.993 ± 0.27 |
| array/dagger/N=1024 (block 512)/reduce (sum) | 2.32 ± 0.2 ms | 3.33 ± 1.5 ms | 0.695 ± 0.32 |
| array/dagger/N=1024 (block 512)/transpose (permutedims) | 7.5 ± 0.88 ms | 7.14 ± 0.72 ms | 1.05 ± 0.16 |
| array/dagger/N=256 (block 256)/add (X + X) | 1.03 ± 0.2 ms | 1.01 ± 0.027 ms | 1.02 ± 0.2 |
| array/dagger/N=256 (block 256)/alloc (rand) | 0.915 ± 0.35 ms | 0.921 ± 0.052 ms | 0.994 ± 0.38 |
| array/dagger/N=256 (block 256)/broadcast (X .+ 1) | 0.765 ± 3.1 ms | 0.572 ± 0.11 ms | 1.34 ± 5.4 |
| array/dagger/N=256 (block 256)/map (sin.(X)) | 1.42 ± 0.17 ms | 1.16 ± 0.027 ms | 1.22 ± 0.15 |
| array/dagger/N=256 (block 256)/norm | 0.517 ± 0.043 ms | 0.547 ± 0.11 ms | 0.945 ± 0.21 |
| array/dagger/N=256 (block 256)/reduce (sum) | 1.22 ± 2.9 ms | 1.11 ± 2.5 ms | 1.1 ± 3.6 |
| array/dagger/N=256 (block 256)/transpose (permutedims) | 1.11 ± 0.19 ms | 1.16 ± 0.056 ms | 0.952 ± 0.17 |
| linalg/dagger/N=1024 (block 512)/cholesky | 23.5 ± 2.9 ms | 27.9 ± 5.2 ms | 0.843 ± 0.19 |
| linalg/dagger/N=1024 (block 512)/lu | 0.0472 ± 0.0047 s | 0.0493 ± 0.0033 s | 0.959 ± 0.11 |
| linalg/dagger/N=1024 (block 512)/matmul (A*A) | 0.0529 ± 0.0037 s | 0.052 ± 0.0041 s | 1.02 ± 0.11 |
| linalg/dagger/N=1024 (block 512)/matvec (A*x) | 3.46 ± 1.1 ms | 3.09 ± 0.96 ms | 1.12 ± 0.49 |
| linalg/dagger/N=1024 (block 512)/qr | 0.111 ± 0.0037 s | 0.112 ± 0.0051 s | 0.988 ± 0.056 |
| linalg/dagger/N=1024 (block 512)/solve (A\b via lu) | 0.0561 ± 0.0053 s | 0.0568 ± 0.0073 s | 0.988 ± 0.16 |
| linalg/dagger/N=1024 (block 512)/svd | 0.0407 h | 0.0394 h | 1.03 |
| linalg/dagger/N=1024 (block 512)/syrk (A'*A) | 0.0355 ± 0.0089 s | 0.0368 ± 0.0046 s | 0.967 ± 0.27 |
| linalg/dagger/N=256 (block 256)/cholesky | 2.77 ± 0.082 ms | 2.81 ± 0.2 ms | 0.986 ± 0.076 |
| linalg/dagger/N=256 (block 256)/lu | 5.02 ± 1.1 ms | 4.16 ± 0.33 ms | 1.21 ± 0.29 |
| linalg/dagger/N=256 (block 256)/matmul (A*A) | 2.36 ± 0.47 ms | 2.5 ± 0.19 ms | 0.947 ± 0.2 |
| linalg/dagger/N=256 (block 256)/matvec (A*x) | 1.34 ± 0.47 ms | 1.26 ± 0.55 ms | 1.07 ± 0.6 |
| linalg/dagger/N=256 (block 256)/qr | 5.16 ± 0.68 ms | 5.11 ± 0.22 ms | 1.01 ± 0.14 |
| linalg/dagger/N=256 (block 256)/solve (A\b via lu) | 9.52 ± 0.94 ms | 10.9 ± 4.2 ms | 0.877 ± 0.35 |
| linalg/dagger/N=256 (block 256)/svd | 0.611 ± 0.13 s | 0.541 ± 0.053 s | 1.13 ± 0.27 |
| linalg/dagger/N=256 (block 256)/syrk (A'*A) | 2.95 ± 0.12 ms | 3.49 ± 0.68 ms | 0.844 ± 0.17 |
| stencil/dagger/N=1024 (block 512)/alloc (neighbors Wrap) | 8.81 ± 0.61 ms | 8.7 ± 0.31 ms | 1.01 ± 0.079 |
| stencil/dagger/N=1024 (block 512)/assign (const) | 1.33 ± 0.17 ms | 1.42 ± 0.052 ms | 0.934 ± 0.12 |
| stencil/dagger/N=1024 (block 512)/multi-expr | 4.27 ± 0.82 ms | 4.09 ± 0.44 ms | 1.04 ± 0.23 |
| stencil/dagger/N=1024 (block 512)/neighbors (Clamp) | 7.2 ± 0.84 ms | 7.14 ± 0.31 ms | 1.01 ± 0.12 |
| stencil/dagger/N=1024 (block 512)/neighbors (Pad) | 7.57 ± 1.1 ms | 8.16 ± 0.49 ms | 0.927 ± 0.15 |
| stencil/dagger/N=1024 (block 512)/neighbors (Reflect) | 7.61 ± 0.69 ms | 8.15 ± 0.23 ms | 0.934 ± 0.089 |
| stencil/dagger/N=1024 (block 512)/neighbors (Wrap) | 7.75 ± 0.36 ms | 7.24 ± 1.1 ms | 1.07 ± 0.17 |
| stencil/dagger/N=1024 (block 512)/update (+) | 2.82 ± 0.34 ms | 2.93 ± 0.42 ms | 0.963 ± 0.18 |
| stencil/dagger/N=256 (block 256)/alloc (neighbors Wrap) | 1.92 ± 0.017 ms | 1.99 ± 0.26 ms | 0.961 ± 0.12 |
| stencil/dagger/N=256 (block 256)/assign (const) | 0.685 ± 0.023 ms | 0.795 ± 0.15 ms | 0.862 ± 0.17 |
| stencil/dagger/N=256 (block 256)/multi-expr | 1.36 ± 0.13 ms | 1.34 ± 0.032 ms | 1.02 ± 0.097 |
| stencil/dagger/N=256 (block 256)/neighbors (Clamp) | 1.61 ± 0.26 ms | 1.55 ± 0.019 ms | 1.04 ± 0.17 |
| stencil/dagger/N=256 (block 256)/neighbors (Pad) | 1.43 ± 0.054 ms | 1.89 ± 0.2 ms | 0.76 ± 0.084 |
| stencil/dagger/N=256 (block 256)/neighbors (Reflect) | 1.53 ± 0.16 ms | 1.7 ± 0.099 ms | 0.903 ± 0.11 |
| stencil/dagger/N=256 (block 256)/neighbors (Wrap) | 1.58 ± 0.095 ms | 1.64 ± 0.11 ms | 0.964 ± 0.086 |
| stencil/dagger/N=256 (block 256)/update (+) | 0.901 ± 0.49 ms | 0.92 ± 0.035 ms | 0.979 ± 0.53 |
| time_to_load | 1.06 ± 0.031 s | 1.06 ± 0.018 s | 0.997 ± 0.033 |
Plots
⚠️ Regressions (> 25.0%)
array/dagger/N=1024 (block 512)/reduce (sum): +43.9%stencil/dagger/N=256 (block 256)/neighbors (Pad): +31.6%
Improvements (> 25.0% faster)
array/dagger/N=256 (block 256)/broadcast (X .+ 1): -25.2%
Distributed benchmarks (4 processes)
Dagger benchmarks: dirty vs master
| master | dirty | master / dirty | |
|---|---|---|---|
| array/dagger/N=1024 (block 512)/add (X + X) | 6.27 ± 0.054 ms | 6.1 ± 0.046 ms | 1.03 ± 0.012 |
| array/dagger/N=1024 (block 512)/alloc (rand) | 5.81 ± 0.027 ms | 5.81 ± 0.057 ms | 0.999 ± 0.011 |
| array/dagger/N=1024 (block 512)/broadcast (X .+ 1) | 4.71 ± 0.1 ms | 4.75 ± 0.15 ms | 0.992 ± 0.037 |
| array/dagger/N=1024 (block 512)/map (sin.(X)) | 13.1 ± 0.054 ms | 13.1 ± 0.052 ms | 1 ± 0.0057 |
| array/dagger/N=1024 (block 512)/norm | 1.34 ± 0.065 ms | 1.23 ± 0.03 ms | 1.09 ± 0.06 |
| array/dagger/N=1024 (block 512)/reduce (sum) | 2.27 ± 0.061 ms | 2.31 ± 0.026 ms | 0.983 ± 0.029 |
| array/dagger/N=1024 (block 512)/transpose (permutedims) | 12.5 ± 0.034 ms | 12.4 ± 0.13 ms | 1 ± 0.011 |
| array/dagger/N=256 (block 256)/add (X + X) | 0.997 ± 0.014 ms | 0.993 ± 0.047 ms | 1 ± 0.05 |
| array/dagger/N=256 (block 256)/alloc (rand) | 1.28 ± 0.018 ms | 1.28 ± 0.012 ms | 1 ± 0.017 |
| array/dagger/N=256 (block 256)/broadcast (X .+ 1) | 0.568 ± 0.08 ms | 0.553 ± 0.039 ms | 1.03 ± 0.16 |
| array/dagger/N=256 (block 256)/map (sin.(X)) | 1.13 ± 0.017 ms | 1.09 ± 0.021 ms | 1.03 ± 0.025 |
| array/dagger/N=256 (block 256)/norm | 0.403 ± 0.021 ms | 0.457 ± 0.021 ms | 0.881 ± 0.061 |
| array/dagger/N=256 (block 256)/reduce (sum) | 0.763 ± 0.028 ms | 0.8 ± 0.053 ms | 0.953 ± 0.072 |
| array/dagger/N=256 (block 256)/transpose (permutedims) | 1.05 ± 0.023 ms | 1.16 ± 0.016 ms | 0.905 ± 0.023 |
| linalg/dagger/N=1024 (block 512)/cholesky | 18.8 ± 0.11 ms | 18.2 ± 0.15 ms | 1.03 ± 0.011 |
| linalg/dagger/N=1024 (block 512)/lu | 0.0448 ± 0.0015 s | 0.042 ± 0.0046 s | 1.07 ± 0.12 |
| linalg/dagger/N=1024 (block 512)/matmul (A*A) | 0.0348 ± 0.00021 s | 0.0335 ± 3e-05 s | 1.04 ± 0.0063 |
| linalg/dagger/N=1024 (block 512)/matvec (A*x) | 3.14 ± 0.09 ms | 3.17 ± 0.14 ms | 0.989 ± 0.053 |
| linalg/dagger/N=1024 (block 512)/qr | 0.105 ± 0.0013 s | 0.109 ± 0.00011 s | 0.962 ± 0.012 |
| linalg/dagger/N=1024 (block 512)/solve (A\b via lu) | 0.0452 ± 0.00045 s | 0.0478 ± 0.00017 s | 0.945 ± 0.0099 |
| linalg/dagger/N=1024 (block 512)/svd | 0.0369 h | 0.037 h | 0.997 |
| linalg/dagger/N=1024 (block 512)/syrk (A'*A) | 0.0321 ± 0.0034 s | 0.035 ± 0.0036 s | 0.916 ± 0.13 |
| linalg/dagger/N=256 (block 256)/cholesky | 2.52 ± 0.045 ms | 2.44 ± 0.14 ms | 1.03 ± 0.063 |
| linalg/dagger/N=256 (block 256)/lu | 4.11 ± 0.099 ms | 4.28 ± 0.069 ms | 0.961 ± 0.028 |
| linalg/dagger/N=256 (block 256)/matmul (A*A) | 1.95 ± 0.066 ms | 2.09 ± 0.056 ms | 0.933 ± 0.04 |
| linalg/dagger/N=256 (block 256)/matvec (A*x) | 1.3 ± 0.054 ms | 1.38 ± 0.061 ms | 0.943 ± 0.057 |
| linalg/dagger/N=256 (block 256)/qr | 3.91 ± 0.041 ms | 4.07 ± 0.14 ms | 0.962 ± 0.034 |
| linalg/dagger/N=256 (block 256)/solve (A\b via lu) | 6.41 ± 0.057 ms | 6.84 ± 0.03 ms | 0.938 ± 0.0093 |
| linalg/dagger/N=256 (block 256)/svd | 0.578 ± 0.027 s | 0.531 ± 0.0044 s | 1.09 ± 0.051 |
| linalg/dagger/N=256 (block 256)/syrk (A'*A) | 2.64 ± 0.012 ms | 2.86 ± 0.11 ms | 0.921 ± 0.036 |
| stencil/dagger/N=1024 (block 512)/alloc (neighbors Wrap) | 14.9 ± 0.057 ms | 15.4 ± 0.18 ms | 0.971 ± 0.012 |
| stencil/dagger/N=1024 (block 512)/assign (const) | 1.51 ± 0.066 ms | 1.49 ± 0.063 ms | 1.01 ± 0.062 |
| stencil/dagger/N=1024 (block 512)/multi-expr | 3.25 ± 0.059 ms | 3.94 ± 0.23 ms | 0.825 ± 0.051 |
| stencil/dagger/N=1024 (block 512)/neighbors (Clamp) | 9.81 ± 0.056 ms | 11.1 ± 0.091 ms | 0.887 ± 0.0089 |
| stencil/dagger/N=1024 (block 512)/neighbors (Pad) | 10.6 ± 0.062 ms | 11.2 ± 0.072 ms | 0.949 ± 0.0082 |
| stencil/dagger/N=1024 (block 512)/neighbors (Reflect) | 11.2 ± 0.15 ms | 9.99 ± 0.13 ms | 1.12 ± 0.021 |
| stencil/dagger/N=1024 (block 512)/neighbors (Wrap) | 11.1 ± 0.036 ms | 11.6 ± 0.34 ms | 0.956 ± 0.028 |
| stencil/dagger/N=1024 (block 512)/update (+) | 2.26 ± 0.07 ms | 2.56 ± 0.022 ms | 0.885 ± 0.029 |
| stencil/dagger/N=256 (block 256)/alloc (neighbors Wrap) | 1.92 ± 0.053 ms | 1.93 ± 0.11 ms | 0.996 ± 0.061 |
| stencil/dagger/N=256 (block 256)/assign (const) | 0.608 ± 0.057 ms | 0.698 ± 0.043 ms | 0.871 ± 0.098 |
| stencil/dagger/N=256 (block 256)/multi-expr | 1.37 ± 0.024 ms | 1.42 ± 0.17 ms | 0.964 ± 0.11 |
| stencil/dagger/N=256 (block 256)/neighbors (Clamp) | 1.69 ± 0.064 ms | 1.68 ± 0.08 ms | 1.01 ± 0.061 |
| stencil/dagger/N=256 (block 256)/neighbors (Pad) | 1.7 ± 0.031 ms | 1.74 ± 0.029 ms | 0.979 ± 0.024 |
| stencil/dagger/N=256 (block 256)/neighbors (Reflect) | 1.72 ± 0.014 ms | 1.63 ± 0.064 ms | 1.05 ± 0.042 |
| stencil/dagger/N=256 (block 256)/neighbors (Wrap) | 1.54 ± 0.026 ms | 1.6 ± 0.064 ms | 0.965 ± 0.042 |
| stencil/dagger/N=256 (block 256)/update (+) | 0.825 ± 0.11 ms | 0.875 ± 0.024 ms | 0.943 ± 0.13 |
| time_to_load | 1.11 ± 0.0096 s | 1.13 ± 0.028 s | 0.983 ± 0.026 |
Plots
No regressions beyond 35.0% 🎉
MPI benchmarks (4 ranks)
Dagger benchmarks: dirty vs master
| master | dirty | master / dirty | |
|---|---|---|---|
| array/dagger/N=1024 (block 512)/add (X + X) | 0.0398 ± 0.0018 s | 0.0388 ± 0.0017 s | 1.03 ± 0.063 |
| array/dagger/N=1024 (block 512)/alloc (rand) | 11.1 ± 9.3 ms | 0.0408 ± 0.0035 s | 0.271 ± 0.23 |
| array/dagger/N=1024 (block 512)/broadcast (X .+ 1) | 4.33 ± 0.49 ms | 3.83 ± 0.42 ms | 1.13 ± 0.18 |
| array/dagger/N=1024 (block 512)/map (sin.(X)) | 6.59 ± 0.49 ms | 6.69 ± 0.11 ms | 0.984 ± 0.075 |
| array/dagger/N=1024 (block 512)/norm | 4.9 ± 0.024 ms | 5.04 ± 0.096 ms | 0.971 ± 0.019 |
| array/dagger/N=1024 (block 512)/reduce (sum) | 7.51 ± 0.006 ms | 7.77 ± 0.25 ms | 0.966 ± 0.031 |
| array/dagger/N=1024 (block 512)/transpose (permutedims) | 27.4 ± 0.26 ms | 29.1 ± 0.062 ms | 0.94 ± 0.0092 |
| array/dagger/N=256 (block 256)/add (X + X) | 11.2 ± 0.14 ms | 11 ± 0.11 ms | 1.01 ± 0.016 |
| array/dagger/N=256 (block 256)/alloc (rand) | 21.5 ± 19 ms | 2.62 ± 0.02 ms | 8.21 ± 7.2 |
| array/dagger/N=256 (block 256)/broadcast (X .+ 1) | 1.29 ± 0.079 ms | 1.32 ± 0.012 ms | 0.981 ± 0.06 |
| array/dagger/N=256 (block 256)/map (sin.(X)) | 1.95 ± 0.0077 ms | 2.03 ± 0.0035 ms | 0.96 ± 0.0041 |
| array/dagger/N=256 (block 256)/norm | 1.87 ± 0.019 ms | 1.89 ± 0.069 ms | 0.992 ± 0.037 |
| array/dagger/N=256 (block 256)/reduce (sum) | 2 ± 0.0065 ms | 2.13 ± 0.064 ms | 0.939 ± 0.028 |
| array/dagger/N=256 (block 256)/transpose (permutedims) | 6.14 ± 0.69 ms | 5.56 ± 0.1 ms | 1.1 ± 0.13 |
| linalg/dagger/N=1024 (block 512)/cholesky | 0.516 ± 0.026 s | 0.52 ± 0.021 s | 0.992 ± 0.064 |
| linalg/dagger/N=1024 (block 512)/lu | 0.0403 h | 0.0487 h | 0.828 |
| linalg/dagger/N=1024 (block 512)/matmul (A*A) | 1.01 ± 0.0025 s | 0.986 ± 0.0018 s | 1.02 ± 0.0031 |
| linalg/dagger/N=1024 (block 512)/matvec (A*x) | 0.0713 ± 0.0035 s | 0.0742 ± 0.0067 s | 0.961 ± 0.099 |
| linalg/dagger/N=1024 (block 512)/qr | 1.12 ± 0.035 s | 1.04 ± 0.027 s | 1.07 ± 0.043 |
| linalg/dagger/N=1024 (block 512)/solve (A\b via lu) | 0.042 h | 0.0481 h | 0.874 |
| linalg/dagger/N=1024 (block 512)/syrk (A'*A) | 0.83 ± 0.01 s | 0.833 ± 0.0026 s | 0.997 ± 0.012 |
| linalg/dagger/N=256 (block 256)/cholesky | 19.2 ± 0.36 ms | 17.5 ± 0.72 ms | 1.1 ± 0.05 |
| linalg/dagger/N=256 (block 256)/lu | 25.8 s | 22.7 s | 1.14 |
| linalg/dagger/N=256 (block 256)/matmul (A*A) | 13.8 ± 0.078 ms | 24.8 ± 1.3 ms | 0.558 ± 0.03 |
| linalg/dagger/N=256 (block 256)/matvec (A*x) | 19.4 ± 3.4 ms | 19.4 ± 3 ms | 1 ± 0.24 |
| linalg/dagger/N=256 (block 256)/qr | 0.05 ± 0.0021 s | 0.189 ± 0.077 s | 0.265 ± 0.11 |
| linalg/dagger/N=256 (block 256)/solve (A\b via lu) | 29.7 s | 28.1 s | 1.06 |
| linalg/dagger/N=256 (block 256)/syrk (A'*A) | 0.136 ± 0.0015 s | 0.138 ± 0.0014 s | 0.987 ± 0.015 |
| stencil/dagger/N=1024 (block 512)/alloc (neighbors Wrap) | 0.179 ± 0.0015 s | 0.26 ± 0.0016 s | 0.687 ± 0.0072 |
| stencil/dagger/N=1024 (block 512)/assign (const) | 12 ± 0.15 ms | 11.7 ± 0.26 ms | 1.03 ± 0.026 |
| stencil/dagger/N=1024 (block 512)/multi-expr | 29.9 ± 0.26 ms | 30.6 ± 0.56 ms | 0.976 ± 0.02 |
| stencil/dagger/N=1024 (block 512)/neighbors (Clamp) | 0.167 ± 0.00034 s | 0.201 ± 0.0045 s | 0.83 ± 0.019 |
| stencil/dagger/N=1024 (block 512)/neighbors (Pad) | 0.162 ± 0.0018 s | 0.143 ± 0.0031 s | 1.13 ± 0.028 |
| stencil/dagger/N=1024 (block 512)/neighbors (Reflect) | 0.166 ± 0.01 s | 0.196 ± 3.5e-05 s | 0.848 ± 0.051 |
| stencil/dagger/N=1024 (block 512)/neighbors (Wrap) | 0.177 ± 0.0026 s | 0.245 ± 0.0094 s | 0.722 ± 0.03 |
| stencil/dagger/N=1024 (block 512)/update (+) | 19.5 ± 0.7 ms | 20.4 ± 0.16 ms | 0.956 ± 0.035 |
| stencil/dagger/N=256 (block 256)/alloc (neighbors Wrap) | 26.2 ± 5.4 ms | 26 ± 5.2 ms | 1.01 ± 0.29 |
| stencil/dagger/N=256 (block 256)/assign (const) | 3.52 ± 0.081 ms | 3.67 ± 0.11 ms | 0.96 ± 0.035 |
| stencil/dagger/N=256 (block 256)/multi-expr | 8.81 ± 0.11 ms | 8.62 ± 0.049 ms | 1.02 ± 0.014 |
| stencil/dagger/N=256 (block 256)/neighbors (Clamp) | 28.3 ± 0.6 ms | 23.4 ± 5.4 ms | 1.21 ± 0.28 |
| stencil/dagger/N=256 (block 256)/neighbors (Pad) | 28.8 ± 0.018 ms | 19 ± 0.2 ms | 1.51 ± 0.016 |
| stencil/dagger/N=256 (block 256)/neighbors (Reflect) | 23.7 ± 5.5 ms | 27 ± 0.79 ms | 0.878 ± 0.2 |
| stencil/dagger/N=256 (block 256)/neighbors (Wrap) | 28.6 ± 0.51 ms | 28.5 ± 0.36 ms | 1 ± 0.022 |
| stencil/dagger/N=256 (block 256)/update (+) | 5.27 ± 0.34 ms | 5.6 ± 0.31 ms | 0.941 ± 0.081 |
| time_to_load | 1.1 ± 0.0093 s | 1.11 ± 0.0092 s | 0.995 ± 0.012 |
Plots
⚠️ Regressions (> 35.0%)
linalg/dagger/N=256 (block 256)/qr: +277.4%array/dagger/N=1024 (block 512)/alloc (rand): +268.6%linalg/dagger/N=256 (block 256)/matmul (A*A): +79.3%stencil/dagger/N=1024 (block 512)/alloc (neighbors Wrap): +45.6%stencil/dagger/N=1024 (block 512)/neighbors (Wrap): +38.5%
Improvements (> 35.0% faster)
array/dagger/N=256 (block 256)/alloc (rand): -87.8%
Full results and plots (download the benchmark-results-* artifacts).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reflects indices like Reflect, but negates the reflected values, for quantities that must vanish at the domain edge - most commonly the velocity or momentum component normal to a solid wall, which Reflect turns into an outflow.
Written by Claude