Comparing AnswerDotAI:main...openfluke:main · AnswerDotAI/gpu.cpp
Commits on May 17, 2025
-
(base) samuel@Steamy:~/git/gpu.cpp/examples/hello_world$ ./build/benchmark_forward_layer CPU forwardLayer: 5.04969 ms GPU forwardLayer: 2.32466 ms Mean abs diff: 1.63608e-06 Max abs diff: 2.67029e-05
-
(base) samuel@Steamy:~/git/gpu.cpp/examples/hello_world$ clang++ -std=c++17 -I../.. -I../../third_party/headers -L../../third_party/lib benchmark_forward_layer_big.cpp -ldl -lwebgpu_dawn -DNO_LOG -o build/benchmark_forward_layer_big ./build/benchmark_forward_layer_big --- Benchmarking inSize: 256, outSize: 128 --- CPU forwardLayer avg: 0.0801554 ms (best: 0.079726 ms) GPU forwardLayer avg: 0.253004 ms (best: 0.200489 ms) Mean abs diff: 2.88826e-07, Max abs diff: 1.90735e-06 --- Benchmarking inSize: 1024, outSize: 512 --- CPU forwardLayer avg: 1.26819 ms (best: 1.24063 ms) GPU forwardLayer avg: 0.559954 ms (best: 0.482283 ms) Mean abs diff: 8.38525e-07, Max abs diff: 7.62939e-06 --- Benchmarking inSize: 4096, outSize: 2048 --- CPU forwardLayer avg: 20.5462 ms (best: 20.2136 ms) GPU forwardLayer avg: 2.14954 ms (best: 1.33978 ms) Mean abs diff: 2.47097e-06, Max abs diff: 2.67029e-05 --- Benchmarking inSize: 8192, outSize: 4096 --- CPU forwardLayer avg: 82.5799 ms (best: 82.1086 ms) GPU forwardLayer avg: 27.9018 ms (best: 25.2048 ms) Mean abs diff: 4.28932e-06, Max abs diff: 5.34058e-05