====== Silicon AppleとGPU処理 ====== MacBook Air (M1)でCLIJを使ったベンチマークをメモ。速度はCPUに比較して20倍から35倍の処理速度になる。 ベンチマークは開発のレポジトリにあるマクロをそのまま使った。三次元の平均フィルタ処理。自分で畳み込みのカーネルを作った場合には、差はあまりないが、二倍程度、早くなる。 https://github.com/clij/clij2-docs/blob/master/src/main/macro/benchmarking.ijm CPU mean filter no 1 took 2687 msec CPU mean filter no 2 took 1759 msec CPU mean filter no 3 took 1959 msec CPU mean filter no 4 took 2253 msec CPU mean filter no 5 took 2557 msec CPU mean filter no 6 took 3406 msec CPU mean filter no 7 took 3526 msec CPU mean filter no 8 took 3722 msec CPU mean filter no 9 took 3520 msec CPU mean filter no 10 took 3419 msec Pushing one image to the GPU took 56 msec CLIJ2 GPU mean filter no 1 took 740 msec CLIJ2 GPU mean filter no 2 took 92 msec CLIJ2 GPU mean filter no 3 took 90 msec CLIJ2 GPU mean filter no 4 took 90 msec CLIJ2 GPU mean filter no 5 took 92 msec CLIJ2 GPU mean filter no 6 took 89 msec CLIJ2 GPU mean filter no 7 took 91 msec CLIJ2 GPU mean filter no 8 took 93 msec CLIJ2 GPU mean filter no 9 took 99 msec CLIJ2 GPU mean filter no 10 took 92 msec Preparing the convolution kernel in GPU memory took 43 msec CLIJ2 GPU mean filter using convolution no 1 took 1500 msec CLIJ2 GPU mean filter using convolution no 2 took 1471 msec CLIJ2 GPU mean filter using convolution no 3 took 1620 msec CLIJ2 GPU mean filter using convolution no 4 took 1551 msec CLIJ2 GPU mean filter using convolution no 5 took 1546 msec CLIJ2 GPU mean filter using convolution no 6 took 1588 msec CLIJ2 GPU mean filter using convolution no 7 took 1475 msec CLIJ2 GPU mean filter using convolution no 8 took 1489 msec CLIJ2 GPU mean filter using convolution no 9 took 1446 msec CLIJ2 GPU mean filter using convolution no 10 took 1551 msec CLIJ GPU mean filter no 1 took 1308 msec CLIJ GPU mean filter no 2 took 106 msec CLIJ GPU mean filter no 3 took 99 msec CLIJ GPU mean filter no 4 took 97 msec CLIJ GPU mean filter no 5 took 100 msec CLIJ GPU mean filter no 6 took 96 msec CLIJ GPU mean filter no 7 took 140 msec CLIJ GPU mean filter no 8 took 116 msec CLIJ GPU mean filter no 9 took 96 msec CLIJ GPU mean filter no 10 took 99 msec Pulling one image from the GPU took 2148 msec GPU: Apple M1 Memory in GB: 16 OpenCL version: 1.2