# Chapter 15 experiment summary

- Measurement rows: 1060
- Out-of-place correctness: PASS
- In-place correctness: PASS
- Nsight geometry profiles: 7

## Layered sweeps

| sweep | ch | buffer | threads | bytes | median us | CV | busbw | vs reference |
|---|---:|---:|---:|---:|---:|---:|---:|---:|
| buffer | 12 | 65536 | 512 | 524288 | 57.500 | 0.78% | 13.68 | +9.90% |
| buffer | 12 | 65536 | 512 | 67108864 | 2037.065 | 0.51% | 49.42 | +138.89% |
| buffer | 12 | 262144 | 512 | 524288 | 57.385 | 0.71% | 13.71 | +9.68% |
| buffer | 12 | 262144 | 512 | 67108864 | 1131.525 | 0.22% | 88.96 | +32.70% |
| buffer | 12 | 1048576 | 512 | 524288 | 57.350 | 0.54% | 13.71 | +9.61% |
| buffer | 12 | 1048576 | 512 | 67108864 | 918.710 | 0.16% | 109.57 | +7.74% |
| buffer | 12 | 4194304 | 512 | 524288 | 52.320 | 3.82% | 15.03 | +0.00% |
| buffer | 12 | 4194304 | 512 | 67108864 | 852.715 | 0.56% | 118.05 | +0.00% |
| buffer | 12 | 16777216 | 512 | 524288 | 51.660 | 0.70% | 15.23 | -1.26% |
| buffer | 12 | 16777216 | 512 | 67108864 | 850.545 | 0.09% | 118.35 | -0.25% |
| channels | 1 | 4194304 | 512 | 4096 | 34.675 | 2.01% | 0.18 | +0.00% |
| channels | 1 | 4194304 | 512 | 524288 | 79.030 | 3.05% | 9.95 | +53.08% |
| channels | 1 | 4194304 | 512 | 67108864 | 4437.350 | 2.72% | 22.69 | +421.85% |
| channels | 2 | 4194304 | 512 | 4096 | 34.515 | 0.54% | 0.18 | -0.46% |
| channels | 2 | 4194304 | 512 | 524288 | 64.585 | 0.45% | 12.18 | +25.10% |
| channels | 2 | 4194304 | 512 | 67108864 | 2538.365 | 0.12% | 39.66 | +198.52% |
| channels | 4 | 4194304 | 512 | 4096 | 34.585 | 12.72% | 0.18 | -0.26% |
| channels | 4 | 4194304 | 512 | 524288 | 57.610 | 0.45% | 13.65 | +11.59% |
| channels | 4 | 4194304 | 512 | 67108864 | 1328.300 | 0.06% | 75.78 | +56.21% |
| channels | 8 | 4194304 | 512 | 4096 | 34.570 | 0.96% | 0.18 | -0.30% |
| channels | 8 | 4194304 | 512 | 524288 | 52.445 | 0.56% | 15.00 | +1.59% |
| channels | 8 | 4194304 | 512 | 67108864 | 1145.555 | 0.10% | 87.88 | +34.72% |
| channels | 12 | 4194304 | 512 | 4096 | 34.675 | 1.73% | 0.18 | +0.00% |
| channels | 12 | 4194304 | 512 | 524288 | 51.625 | 1.92% | 15.23 | +0.00% |
| channels | 12 | 4194304 | 512 | 67108864 | 850.310 | 0.27% | 118.38 | +0.00% |
| interaction | 2 | 262144 | 128 | 524288 | 65.445 | 0.33% | 12.02 | +26.87% |
| interaction | 2 | 262144 | 128 | 67108864 | 5277.510 | 0.12% | 19.07 | +520.78% |
| interaction | 2 | 262144 | 512 | 524288 | 64.855 | 0.59% | 12.13 | +25.72% |
| interaction | 2 | 262144 | 512 | 67108864 | 4061.520 | 0.14% | 24.79 | +377.75% |
| interaction | 2 | 4194304 | 128 | 524288 | 65.335 | 0.49% | 12.04 | +26.66% |
| interaction | 2 | 4194304 | 128 | 67108864 | 4526.305 | 0.13% | 22.24 | +432.42% |
| interaction | 2 | 4194304 | 512 | 524288 | 64.780 | 0.40% | 12.14 | +25.58% |
| interaction | 2 | 4194304 | 512 | 67108864 | 2539.165 | 0.09% | 39.64 | +198.68% |
| interaction | 12 | 262144 | 128 | 524288 | 53.265 | 0.67% | 14.77 | +3.26% |
| interaction | 12 | 262144 | 128 | 67108864 | 1188.540 | 0.14% | 84.69 | +39.81% |
| interaction | 12 | 262144 | 512 | 524288 | 57.470 | 0.51% | 13.68 | +11.41% |
| interaction | 12 | 262144 | 512 | 67108864 | 1130.910 | 0.25% | 89.01 | +33.03% |
| interaction | 12 | 4194304 | 128 | 524288 | 48.410 | 0.76% | 16.25 | -6.15% |
| interaction | 12 | 4194304 | 128 | 67108864 | 1028.755 | 0.09% | 97.85 | +21.01% |
| interaction | 12 | 4194304 | 512 | 524288 | 51.585 | 0.41% | 15.25 | +0.00% |
| interaction | 12 | 4194304 | 512 | 67108864 | 850.140 | 0.08% | 118.41 | +0.00% |
| threads | 12 | 4194304 | 64 | 4096 | 34.655 | 0.56% | 0.18 | +0.17% |
| threads | 12 | 4194304 | 64 | 524288 | 50.175 | 0.40% | 15.67 | -2.95% |
| threads | 12 | 4194304 | 64 | 67108864 | 1607.250 | 0.04% | 62.63 | +88.92% |
| threads | 12 | 4194304 | 128 | 4096 | 34.675 | 1.03% | 0.18 | +0.23% |
| threads | 12 | 4194304 | 128 | 524288 | 48.125 | 0.55% | 16.34 | -6.91% |
| threads | 12 | 4194304 | 128 | 67108864 | 1028.420 | 0.07% | 97.88 | +20.88% |
| threads | 12 | 4194304 | 256 | 4096 | 34.635 | 0.51% | 0.18 | +0.12% |
| threads | 12 | 4194304 | 256 | 524288 | 49.830 | 0.48% | 15.79 | -3.62% |
| threads | 12 | 4194304 | 256 | 67108864 | 874.825 | 0.09% | 115.06 | +2.83% |
| threads | 12 | 4194304 | 512 | 4096 | 34.595 | 3.54% | 0.18 | +0.00% |
| threads | 12 | 4194304 | 512 | 524288 | 51.700 | 2.76% | 15.21 | +0.00% |
| threads | 12 | 4194304 | 512 | 67108864 | 850.765 | 0.27% | 118.32 | +0.00% |

## Interaction residuals

| ch | buffer | threads | bytes | actual us | independent prediction | residual |
|---:|---:|---:|---:|---:|---:|---:|
| 2 | 262144 | 128 | 524288 | 65.445 | 65.888 | -0.67% |
| 2 | 262144 | 128 | 67108864 | 5277.510 | 4070.881 | +29.64% |
| 2 | 262144 | 512 | 524288 | 64.855 | 70.782 | -8.37% |
| 2 | 262144 | 512 | 67108864 | 4061.520 | 3367.654 | +20.60% |
| 2 | 4194304 | 128 | 524288 | 65.335 | 60.072 | +8.76% |
| 2 | 4194304 | 128 | 67108864 | 4526.305 | 3067.808 | +47.54% |
| 2 | 4194304 | 512 | 524288 | 64.780 | 64.535 | +0.38% |
| 2 | 4194304 | 512 | 67108864 | 2539.165 | 2537.858 | +0.05% |
| 12 | 262144 | 128 | 524288 | 53.265 | 52.666 | +1.14% |
| 12 | 262144 | 128 | 67108864 | 1188.540 | 1363.677 | -12.84% |
| 12 | 262144 | 512 | 524288 | 57.470 | 56.579 | +1.58% |
| 12 | 262144 | 512 | 67108864 | 1130.910 | 1128.108 | +0.25% |
| 12 | 4194304 | 128 | 524288 | 48.410 | 48.018 | +0.82% |
| 12 | 4194304 | 128 | 67108864 | 1028.755 | 1027.664 | +0.11% |
| 12 | 4194304 | 512 | 524288 | 51.585 | 51.585 | +0.00% |
| 12 | 4194304 | 512 | 67108864 | 850.140 | 850.140 | +0.00% |

## Kernel geometry

| label | bytes | requested ch | requested threads | gridX | blockX | matches channels | median kernel us |
|---|---:|---:|---:|---|---|---:|---:|
| channels_1 | 1048576 | 1 | 512 | 1 | 544 | 1 | 124.784 |
| channels_4 | 1048576 | 4 | 512 | 4 | 544 | 1 | 73.024 |
| channels_12 | 1048576 | 12 | 512 | 12 | 544 | 1 | 67.103 |
| threads_64 | 1048576 | 12 | 64 | 12 | 128 | 1 | 61.583 |
| threads_128 | 1048576 | 12 | 128 | 12 | 160 | 1 | 57.903 |
| threads_256 | 1048576 | 12 | 256 | 12 | 288 | 1 | 66.144 |
| tiny_4k_requested_ch12_t512 | 4096 | 12 | 512 | 1 | 128 | 0 | 42.017 |
