oneapi-agent-toolkit

GPU run summary (real tool output)

fir

step status summary
reference_build rc=0 g++ -O2 -std=c++17 -I/workspace/repo/examples/fir/../common -I/workspace/repo/examples/fir /workspace/repo/examples/fir/fir_ref.cpp -o build/reference
cuda_build rc=0 /usr/local/cuda/bin/nvcc -O2 -std=c++17 -arch=native -I/workspace/repo/examples/common -I/workspace/repo/examples/fir /workspace/repo/examples/fir/fir.cu -o build/cuda
cuda_verify_vs_cpp ok PASS: 2/2 case(s) match (worst residual peak -135.0 dBFS, threshold -96.0 dBFS)
cuda_benchmark ok 10 run(s) each; none: median 336.73 ms, p95 404.30 ms
sycl_gpu_compile ok compiled 1 source(s) -> /workspace/gpu-run/work/fir/build/sycl_gpu (0 warning(s))
sycl_verify_vs_cpp ok PASS: 2/2 case(s) match (worst residual peak -135.0 dBFS, threshold -96.0 dBFS)
sycl_verify_vs_cuda ok PASS: 2/2 case(s) match (bit-exact, threshold -96.0 dBFS)
sycl_benchmark ok 10 run(s) each; cuda:0: median 375.23 ms, p95 399.08 ms
cpp_benchmark ok 10 run(s) each; none: median 62.31 ms, p95 71.09 ms

fuzz

step status summary
reference_build rc=0 g++ -O2 -std=c++17 -I/workspace/repo/examples/fuzz/../common /workspace/repo/examples/fuzz/fuzz_ref.cpp -o build/reference
cuda_build rc=0 /usr/local/cuda/bin/nvcc -O2 -std=c++17 -arch=native -I/workspace/repo/examples/common /workspace/repo/examples/fuzz/fuzz.cu -o build/cuda
cuda_verify_vs_cpp ok PASS: 2/2 case(s) match (worst residual peak -138.5 dBFS, threshold -96.0 dBFS)
cuda_benchmark ok 10 run(s) each; none: median 347.76 ms, p95 493.95 ms
sycl_gpu_compile ok compiled 1 source(s) -> /workspace/gpu-run/work/fuzz/build/sycl_gpu (0 warning(s))
sycl_verify_vs_cpp ok PASS: 2/2 case(s) match (worst residual peak -138.5 dBFS, threshold -96.0 dBFS)
sycl_verify_vs_cuda ok PASS: 2/2 case(s) match (bit-exact, threshold -96.0 dBFS)
sycl_benchmark ok 10 run(s) each; cuda:0: median 365.17 ms, p95 395.68 ms
cpp_benchmark ok 10 run(s) each; none: median 59.49 ms, p95 66.86 ms

saxpy

step status summary
reference_build rc=0 g++ -O2 -std=c++17 /workspace/repo/examples/saxpy/saxpy_ref.cpp -o build/reference
cuda_build rc=0 /usr/local/cuda/bin/nvcc -O2 -std=c++17 -arch=native /workspace/repo/examples/saxpy/saxpy.cu -o build/cuda
cuda_verify_vs_cpp ok PASS: 1/1 case(s) match (worst max abs err 4.77e-07, atol 1e-06)
cuda_benchmark ok 10 run(s) each; none: median 357.74 ms, p95 402.51 ms
sycl_gpu_compile ok compiled 1 source(s) -> /workspace/gpu-run/work/saxpy/build/sycl_gpu (0 warning(s))
sycl_verify_vs_cpp ok PASS: 1/1 case(s) match (worst max abs err 4.77e-07, atol 1e-06)
sycl_verify_vs_cuda ok PASS: 1/1 case(s) match (worst max abs err 0, atol 1e-06)
sycl_benchmark ok 10 run(s) each; cuda:0: median 363.79 ms, p95 550.12 ms
cpp_benchmark ok 10 run(s) each; none: median 58.66 ms, p95 79.98 ms

softmax_hecbench

step status summary
reference_build rc=0 g++ -O2 -std=c++17 -I/workspace/repo/examples/softmax_hecbench /workspace/repo/examples/softmax_hecbench/softmax_ref.cpp -o build/reference
cuda_build rc=0 /usr/local/cuda/bin/nvcc -O2 -std=c++17 -arch=native -I/workspace/repo/examples/softmax_hecbench /workspace/repo/examples/softmax_hecbench/softmax.cu -o build/cuda
cuda_verify_vs_cpp ok PASS: 1/1 case(s) match (worst max abs err 9.69e-08, atol 1e-06)
cuda_benchmark ok 10 run(s) each; none: median 329.19 ms, p95 383.48 ms
sycl_gpu_compile ok compiled 1 source(s) -> /workspace/gpu-run/work/softmax_hecbench/build/sycl_gpu (0 warning(s))
sycl_verify_vs_cpp ok PASS: 1/1 case(s) match (worst max abs err 9.69e-08, atol 1e-06)
sycl_verify_vs_cuda ok PASS: 1/1 case(s) match (worst max abs err 7.45e-09, atol 1e-06)
sycl_benchmark ok 10 run(s) each; cuda:0: median 329.86 ms, p95 371.23 ms
cpp_benchmark ok 10 run(s) each; none: median 63.05 ms, p95 80.22 ms
This site is open source. Improve this page.