Task07 Denis Sokolov ITMO

[DenChika]

Локальный вывод

$ ./main_sparse_matrix_multiply
Found 2 GPUs in 0.211759 sec (OpenCL: 0.138175 sec, Vulkan: 0.0727802 sec)    
Available devices:                                                            
  Device #0: API: OpenCL. GPU. AMD Radeon(TM) Graphics (gfx902). Free memory: 3069/3137 Mb.                                 
  Device #1: API: OpenCL. CPU. AMD Ryzen 5 5500U with Radeon Graphics         . Intel(R) Corporation. Total memory: 7514 Mb.
Using device #0: API: OpenCL. GPU. AMD Radeon(TM) Graphics (gfx902). Free memory: 3069/3137 Mb.                             
Using OpenCL API...                                                                                                         
Evaluating CSR matrix nrows x ncols=1000000x1000000 with values in range [0; 1000]          
____________________________________________________________________________________________
Evaluating with NNZ per row in range [32; 32], median NNZ per row=32, total NNZ=32000000...
CPU (multi-threaded via OpenMP) finished in 0.0572832 sec      
CPU effective bandwidth: 2.19343 GB/s (551.586 uint millions/s)
Kernels compilation done in 0.0779748 seconds
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.222917 10%=0.223147 median=0.224041 90%=0.306784 max=0.306784)
GPU SpMV median effective VRAM bandwidth: 0.565343 GB/s (142.831 uint millions/s)                                                                  
____________________________________________________________________________________________
Evaluating with NNZ per row in range [128; 128], median NNZ per row=128, total NNZ=128000000...
CPU (multi-threaded via OpenMP) finished in 0.506888 sec
CPU effective bandwidth: 0.874346 GB/s (230.939 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.420493 10%=0.422547 median=0.432062 90%=0.523749 max=0.523749)
GPU SpMV median effective VRAM bandwidth: 1.12088 GB/s (296.254 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [1; 32], median NNZ per row=17, total NNZ=16499998...
CPU (multi-threaded via OpenMP) finished in 0.036304 sec
CPU effective bandwidth: 1.85481 GB/s (424.948 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.222726 10%=0.222782 median=0.223535 90%=0.224793 max=0.224793)
GPU SpMV median effective VRAM bandwidth: 0.308309 GB/s (71.5771 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [1; 128], median NNZ per row=64, total NNZ=64499934...
CPU (multi-threaded via OpenMP) finished in 0.142101 sec
CPU effective bandwidth: 1.73679 GB/s (445.96 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.229239 10%=0.229381 median=0.234751 90%=0.267188 max=0.267188)
GPU SpMV median effective VRAM bandwidth: 1.0553 GB/s (272.63 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [32; 128], median NNZ per row=80, total NNZ=79933808...
CPU (multi-threaded via OpenMP) finished in 0.159288 sec
CPU effective bandwidth: 1.9101 GB/s (493.638 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.271417 10%=0.274579 median=0.315918 90%=0.34568 max=0.34568)
GPU SpMV median effective VRAM bandwidth: 0.96616 GB/s (250.065 uint millions/s)

Вывод Github CI

$ ./main_sparse_matrix_multiply
Found 2 GPUs in 0.0578978 sec (CUDA: 8.7642e-05 sec, OpenCL: 0.0282925 sec, Vulkan: 0.0294639 sec)
Available devices:
  Device #0: API: OpenCL. CPU. AMD EPYC 7763 64-Core Processor                . Intel(R) Corporation. Total memory: 15994 Mb.
  Device #1: API: Vulkan. CPU. llvmpipe (LLVM 20.1.2, 256 bits). Free memory: 15994/15994 Mb.
Using device #0: API: OpenCL. CPU. AMD EPYC 7763 64-Core Processor                . Intel(R) Corporation. Total memory: 15994 Mb.
Using OpenCL API...
Evaluating CSR matrix nrows x ncols=1000000x1000000 with values in range [0; 1000]
____________________________________________________________________________________________
Evaluating with NNZ per row in range [32; 32], median NNZ per row=32, total NNZ=32000000...
CPU (multi-threaded via OpenMP) finished in 0.029[8](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:9)864 sec
CPU effective bandwidth: 4.23377 GB/s (106[9](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:10).6 uint millions/s)
Kernels compilation done in 0.129737 seconds
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - [10](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:11) values (min=0.289951 10%=0.290186 median=0.291467 90%=0.467056 max=0.467056)
GPU SpMV median effective VRAM bandwidth: 0.43456 GB/s (109.789 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [128; 128], median NNZ per row=128, total NNZ=128000000...
CPU (multi-threaded via OpenMP) finished in 0.0684338 sec
CPU effective bandwidth: 7.07314 GB/s (1869.45 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.373149 10%=0.374616 median=0.380655 90%=0.438381 max=0.438381)
GPU SpMV median effective VRAM bandwidth: 1.27225 GB/s (336.263 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [1; 32], median NNZ per row=17, total NNZ=16499998...
CPU (multi-threaded via OpenMP) finished in 0.0143547 sec
CPU effective bandwidth: 4.79126 GB/s ([11](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:12)12.27 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.282293 10%=0.282771 median=0.28363 90%=0.298394 max=0.298394)
GPU SpMV median effective VRAM bandwidth: 0.242985 GB/s (56.4116 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [1; [12](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:13)8], median NNZ per row=64, total NNZ=64499934...
CPU (multi-threaded via OpenMP) finished in 0.0375956 sec
CPU effective bandwidth: 6.58347 GB/s (1700.76 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.328957 10%=0.329019 median=0.330702 90%=0.333874 max=0.333874)
GPU SpMV median effective VRAM bandwidth: 0.749107 GB/s (193.528 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [32; 128], median NNZ per row=80, total NNZ=8001[14](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:15)95...
CPU (multi-threaded via OpenMP) finished in 0.0470027 sec
CPU effective bandwidth: 6.49465 GB/s (1700.6 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.34371 10%=0.344303 median=0.345997 90%=0.349866 max=0.349866)
GPU SpMV median effective VRAM bandwidth: 0.883003 GB/s (231.2[16](https://github.com/GPGPUCourse/GPGPUTasks2025/actions/runs/22972272745/job/66691828238#step:16:17) uint millions/s)

[GPUcourseBOT]

✅ Результаты тестирования PR #1070

Логи тестирования (нажмите чтобы развернуть)

=== СТАТУС: Успешно выполнены программы: main_sparse_matrix_multiply ===
=== main_sparse_matrix_multiply stdout (exit code: -11 (segfault после выполнения)) ===
Found 1 GPUs in 8.57898 sec (CUDA: 0.115557 sec, OpenCL: 0.706491 sec, Vulkan: 7.75687 sec)
Available devices:
Device #0: API: CUDA+OpenCL+Vulkan. GPU. Tesla T4 (CUDA 12020). Free memory: 14822/14930 Mb.
Using device #0: API: CUDA+OpenCL+Vulkan. GPU. Tesla T4 (CUDA 12020). Free memory: 14822/14930 Mb.
Using OpenCL API...
Evaluating CSR matrix nrows x ncols=1000000x1000000 with values in range [0; 1000]
____________________________________________________________________________________________
Evaluating with NNZ per row in range [32; 32], median NNZ per row=32, total NNZ=32000000...
CPU (multi-threaded via OpenMP) finished in 0.0432619 sec
CPU effective bandwidth: 2.92585 GB/s (739.182 uint millions/s)
Kernels compilation done in 3.49236 seconds
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.0130587 10%=0.0130666 median=0.0292415 90%=3.52165 max=3.52165)
GPU SpMV median effective VRAM bandwidth: 4.33151 GB/s (1094.34 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [128; 128], median NNZ per row=128, total NNZ=128000000...
CPU (multi-threaded via OpenMP) finished in 0.167461 sec
CPU effective bandwidth: 2.89139 GB/s (764.205 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.0189664 10%=0.0189897 median=0.0192241 90%=0.0325111 max=0.0325111)
GPU SpMV median effective VRAM bandwidth: 25.1916 GB/s (6658.3 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [1; 32], median NNZ per row=17, total NNZ=16499998...
CPU (multi-threaded via OpenMP) finished in 0.0237571 sec
CPU effective bandwidth: 2.89746 GB/s (672.649 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.01152 10%=0.0115225 median=0.0115239 90%=0.0115559 max=0.0115559)
GPU SpMV median effective VRAM bandwidth: 5.98045 GB/s (1388.42 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [1; 128], median NNZ per row=64, total NNZ=64499934...
CPU (multi-threaded via OpenMP) finished in 0.0830694 sec
CPU effective bandwidth: 2.98108 GB/s (770.134 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.0144483 10%=0.0145255 median=0.030362 90%=0.030388 max=0.030388)
GPU SpMV median effective VRAM bandwidth: 8.15927 GB/s (2107.9 uint millions/s)
____________________________________________________________________________________________
Evaluating with NNZ per row in range [32; 128], median NNZ per row=80, total NNZ=80011495...
CPU (multi-threaded via OpenMP) finished in 0.104959 sec
CPU effective bandwidth: 2.90984 GB/s (761.936 uint millions/s)
GPU SpMV (sparse matrix-vector multiplication) times (in seconds) - 10 values (min=0.0154801 10%=0.0156685 median=0.0308816 90%=0.0309165 max=0.0309165)
GPU SpMV median effective VRAM bandwidth: 9.89317 GB/s (2590.54 uint millions/s)

Посмотреть полные логи

[PolarNick239]

4/5 баллов 👍