Benchmark on RX480, Ubuntu 22 #20

dagelf · 2022-09-11T07:03:12Z

dagelf
Sep 11, 2022

$ ./dlprim_flops 1:0 4
Testing on Ellesmere on AMD Accelerated Parallel Processing
Testing memory speed
- Vector size 1
-- Warming 
-- Running   181.258 GB/s
- Vector size 2
-- Warming 
-- Running   171.371 GB/s
- Vector size 4
-- Warming 
-- Running   176.268 GB/s
- Vector size 8
-- Warming 
-- Running   178.933 GB/s
- Vector size 16
-- Warming 
-- Running   160.408 GB/s
Testing flops float
- Vector size 1
-- Warming 
-- Running   5130.13 GFlops
- Vector size 2
-- Warming 
-- Running   4568.01 GFlops
- Vector size 4
-- Warming 
-- Running   4261.33 GFlops
- Vector size 8
-- Warming 
-- Running   5144.67 GFlops
- Vector size 16
-- Warming 
-- Running   2873.35 GFlops
Testing flops half
- Vector size 1
-- Warming 
-- Running   2486.5 GFlops
- Vector size 2
-- Warming 
-- Running   2693.7 GFlops
- Vector size 4
-- Warming 
-- Running   2602.51 GFlops
- Vector size 8
-- Warming 
-- Running   1690.09 GFlops
- Vector size 16
-- Warming 
-- Running   1490.37 GFlops
Summray for Ellesmere on AMD Accelerated Parallel Processing
Peak GFlops for float 5144.67
Peak GFlops for half 2693.7
Peak memory 181.258 GB/s
GEMM
  NN  0:  512,  512,  512      803.2 GFlops (15.61%)      9.4 GB/s ( 5.87%) limited by gflops 15.61%
  NN  1: 1024, 1024, 1024     1644.4 GFlops (31.96%)      9.6 GB/s ( 6.01%) limited by gflops 31.96%
  NN  2: 1025, 1025, 1025     1364.4 GFlops (26.52%)      8.0 GB/s ( 4.98%) limited by gflops 26.52%
  NN  3: 2048, 2048, 2048     1513.7 GFlops (29.42%)      4.4 GB/s ( 2.77%) limited by gflops 29.42%
  NN  4: 2049, 2049, 2049     1856.7 GFlops (36.09%)      5.4 GB/s ( 3.39%) limited by gflops 36.09%
  NN  5:   64, 2048,   64      553.7 GFlops (10.76%)     35.4 GB/s (22.08%) limited by memory 22.08%
  NN  6: 2048,   64, 2048      646.5 GFlops (12.57%)     21.5 GB/s (13.38%) limited by memory 13.38%
  NN  7: 2048, 2048,   64      933.2 GFlops (18.14%)     31.2 GB/s (19.47%) limited by memory 19.47%
  NN  8: 2048,   64,   64      431.7 GFlops ( 8.39%)     27.6 GB/s (17.22%) limited by memory 17.22%
  NN  9:   64, 2048, 2048     1369.6 GFlops (26.62%)     45.5 GB/s (28.36%) limited by memory 28.36%
  NN 10:   64,   64, 2048      232.7 GFlops ( 4.52%)     14.8 GB/s ( 9.21%) limited by memory  9.21%
  NT  0:  512,  512,  512      728.8 GFlops (14.17%)      8.5 GB/s ( 5.33%) limited by gflops 14.17%
  NT  1: 1024, 1024, 1024     2087.0 GFlops (40.57%)     12.2 GB/s ( 7.63%) limited by gflops 40.57%
  NT  2: 1025, 1025, 1025     1855.5 GFlops (36.07%)     10.9 GB/s ( 6.77%) limited by gflops 36.07%
  NT  3: 2048, 2048, 2048     1607.0 GFlops (31.24%)      4.7 GB/s ( 2.94%) limited by gflops 31.24%
  NT  4: 2049, 2049, 2049     2653.3 GFlops (51.57%)      7.8 GB/s ( 4.84%) limited by gflops 51.57%
  NT  5:   64, 2048,   64      454.1 GFlops ( 8.83%)     29.1 GB/s (18.11%) limited by memory 18.11%
  NT  6: 2048,   64, 2048      354.6 GFlops ( 6.89%)     11.8 GB/s ( 7.34%) limited by memory  7.34%
  NT  7: 2048, 2048,   64     1066.2 GFlops (20.72%)     35.7 GB/s (22.24%) limited by memory 22.24%
  NT  8: 2048,   64,   64      453.2 GFlops ( 8.81%)     29.0 GB/s (18.08%) limited by memory 18.08%
  NT  9:   64, 2048, 2048      898.5 GFlops (17.46%)     29.8 GB/s (18.60%) limited by memory 18.60%
  NT 10:   64,   64, 2048      232.9 GFlops ( 4.53%)     14.8 GB/s ( 9.22%) limited by memory  9.22%
  TN  0:  512,  512,  512      803.7 GFlops (15.62%)      9.4 GB/s ( 5.88%) limited by gflops 15.62%
  TN  1: 1024, 1024, 1024     1643.7 GFlops (31.95%)      9.6 GB/s ( 6.01%) limited by gflops 31.95%
  TN  2: 1025, 1025, 1025     1292.2 GFlops (25.12%)      7.6 GB/s ( 4.72%) limited by gflops 25.12%
  TN  3: 2048, 2048, 2048     1740.6 GFlops (33.83%)      5.1 GB/s ( 3.18%) limited by gflops 33.83%
  TN  4: 2049, 2049, 2049     1762.8 GFlops (34.26%)      5.2 GB/s ( 3.22%) limited by gflops 34.26%
  TN  5:   64, 2048,   64      533.3 GFlops (10.37%)     34.1 GB/s (21.27%) limited by memory 21.27%
  TN  6: 2048,   64, 2048     1363.2 GFlops (26.50%)     45.3 GB/s (28.22%) limited by memory 28.22%
  TN  7: 2048, 2048,   64      935.0 GFlops (18.17%)     31.3 GB/s (19.51%) limited by memory 19.51%
  TN  8: 2048,   64,   64      442.5 GFlops ( 8.60%)     28.3 GB/s (17.65%) limited by memory 17.65%
  TN  9:   64, 2048, 2048     1368.2 GFlops (26.59%)     45.4 GB/s (28.33%) limited by memory 28.33%
  TN 10:   64,   64, 2048      230.0 GFlops ( 4.47%)     14.6 GB/s ( 9.10%) limited by memory  9.10%
  TT  0:  512,  512,  512      824.6 GFlops (16.03%)      9.7 GB/s ( 6.03%) limited by gflops 16.03%
  TT  1: 1024, 1024, 1024     2327.3 GFlops (45.24%)     13.6 GB/s ( 8.51%) limited by gflops 45.24%
  TT  2: 1025, 1025, 1025     1734.0 GFlops (33.71%)     10.2 GB/s ( 6.33%) limited by gflops 33.71%
  TT  3: 2048, 2048, 2048     2378.0 GFlops (46.22%)      7.0 GB/s ( 4.34%) limited by gflops 46.22%
  TT  4: 2049, 2049, 2049     2516.0 GFlops (48.91%)      7.4 GB/s ( 4.59%) limited by gflops 48.91%
  TT  5:   64, 2048,   64      475.6 GFlops ( 9.24%)     30.4 GB/s (18.97%) limited by memory 18.97%
  TT  6: 2048,   64, 2048     1372.8 GFlops (26.68%)     45.6 GB/s (28.42%) limited by memory 28.42%
  TT  7: 2048, 2048,   64     1121.9 GFlops (21.81%)     37.5 GB/s (23.41%) limited by memory 23.41%
  TT  8: 2048,   64,   64      459.4 GFlops ( 8.93%)     29.4 GB/s (18.32%) limited by memory 18.32%
  TT  9:   64, 2048, 2048      779.9 GFlops (15.16%)     25.9 GB/s (16.15%) limited by memory 16.15%
  TT 10:   64,   64, 2048      235.5 GFlops ( 4.58%)     15.0 GB/s ( 9.32%) limited by memory  9.32%
Convolution
   0     effnet  forward b=64 k=3  p=1 s=1 in=480  out=480  g=480 D=14      120.0 GFlops ( 2.33%)     53.4 GB/s (33.27%) limited by memory 33.27% algo=depthwise_separable
   0     effnet bwd-data b=64 k=3  p=1 s=1 in=480  out=480  g=480 D=14       35.3 GFlops ( 0.69%)     15.7 GB/s ( 9.77%) limited by memory  9.77% algo=depthwise_separable
   0     effnet bwd-filt b=64 k=3  p=1 s=1 in=480  out=480  g=480 D=14       11.2 GFlops ( 0.22%)      5.0 GB/s ( 3.11%) limited by memory  3.11% algo=depthwise_separable
   1    alexnet  forward b=64 k=11 p=2 s=4 in=3    out=64   g=1   D=224    2241.7 GFlops (43.57%)     22.0 GB/s (13.70%) limited by gflops 43.57% algo=gemm
   1    alexnet bwd-data b=64 k=11 p=2 s=4 in=3    out=64   g=1   D=224     485.3 GFlops ( 9.43%)      4.8 GB/s ( 2.97%) limited by gflops  9.43% algo=gemm
   1    alexnet bwd-filt b=64 k=11 p=2 s=4 in=3    out=64   g=1   D=224    1031.4 GFlops (20.05%)     10.1 GB/s ( 6.31%) limited by gflops 20.05% algo=gemm
   2    alexnet  forward b=64 k=5  p=2 s=1 in=96   out=192  g=2   D=27     1711.4 GFlops (33.27%)      4.4 GB/s ( 2.71%) limited by gflops 33.27% algo=gemm
   2    alexnet bwd-data b=64 k=5  p=2 s=1 in=96   out=192  g=2   D=27     1168.5 GFlops (22.71%)      3.0 GB/s ( 1.85%) limited by gflops 22.71% algo=gemm
   2    alexnet bwd-filt b=64 k=5  p=2 s=1 in=96   out=192  g=2   D=27     1382.3 GFlops (26.87%)      3.6 GB/s ( 2.23%) limited by gflops 26.87% algo=gemm
   3    alexnet  forward b=64 k=5  p=2 s=1 in=64   out=192  g=1   D=27     2224.1 GFlops (43.23%)      3.8 GB/s ( 2.37%) limited by gflops 43.23% algo=gemm
   3    alexnet bwd-data b=64 k=5  p=2 s=1 in=64   out=192  g=1   D=27     1690.9 GFlops (32.87%)      2.9 GB/s ( 1.80%) limited by gflops 32.87% algo=gemm
   3    alexnet bwd-filt b=64 k=5  p=2 s=1 in=64   out=192  g=1   D=27     1272.7 GFlops (24.74%)      2.2 GB/s ( 1.39%) limited by gflops 24.74% algo=gemm
   4    alexnet  forward b=64 k=3  p=1 s=1 in=384  out=256  g=1   D=13     2575.2 GFlops (50.06%)      4.2 GB/s ( 2.62%) limited by gflops 50.06% algo=winograd
   4    alexnet bwd-data b=64 k=3  p=1 s=1 in=384  out=256  g=1   D=13     2269.3 GFlops (44.11%)      3.7 GB/s ( 2.31%) limited by gflops 44.11% algo=winograd
   4    alexnet bwd-filt b=64 k=3  p=1 s=1 in=384  out=256  g=1   D=13     2838.8 GFlops (55.18%)      5.2 GB/s ( 3.21%) limited by gflops 55.18% algo=winograd
   5     resnet  forward b=64 k=7  p=3 s=2 in=3    out=64   g=1   D=224    2044.8 GFlops (39.75%)     33.0 GB/s (20.60%) limited by gflops 39.75% algo=gemm
   5     resnet bwd-data b=64 k=7  p=3 s=2 in=3    out=64   g=1   D=224     602.5 GFlops (11.71%)      9.7 GB/s ( 6.07%) limited by gflops 11.71% algo=gemm
   5     resnet bwd-filt b=64 k=7  p=3 s=2 in=3    out=64   g=1   D=224     569.8 GFlops (11.08%)      9.2 GB/s ( 5.74%) limited by gflops 11.08% algo=gemm
   6     resnet  forward b=64 k=1  p=0 s=1 in=64   out=256  g=1   D=56     1437.7 GFlops (27.94%)     56.2 GB/s (35.02%) limited by memory 35.02% algo=gemm
   6     resnet bwd-data b=64 k=1  p=0 s=1 in=64   out=256  g=1   D=56     1909.3 GFlops (37.11%)     74.6 GB/s (46.51%) limited by memory 46.51% algo=gemm
   6     resnet bwd-filt b=64 k=1  p=0 s=1 in=64   out=256  g=1   D=56     1023.2 GFlops (19.89%)     40.0 GB/s (24.93%) limited by memory 24.93% algo=gemm
   7     resnet  forward b=64 k=1  p=0 s=1 in=64   out=64   g=1   D=56     1759.3 GFlops (34.20%)    110.0 GB/s (68.56%) limited by memory 68.56% algo=gemm
   7     resnet bwd-data b=64 k=1  p=0 s=1 in=64   out=64   g=1   D=56     1003.4 GFlops (19.50%)     62.7 GB/s (39.10%) limited by memory 39.10% algo=gemm
   7     resnet bwd-filt b=64 k=1  p=0 s=1 in=64   out=64   g=1   D=56      293.0 GFlops ( 5.70%)     18.3 GB/s (11.42%) limited by memory 11.42% algo=gemm
   8     resnet  forward b=64 k=3  p=1 s=1 in=64   out=64   g=1   D=56     2946.1 GFlops (57.26%)     20.5 GB/s (12.77%) limited by gflops 57.26% algo=winograd
   8     resnet bwd-data b=64 k=3  p=1 s=1 in=64   out=64   g=1   D=56     1019.2 GFlops (19.81%)      7.1 GB/s ( 4.42%) limited by gflops 19.81% algo=winograd
   8     resnet bwd-filt b=64 k=3  p=1 s=1 in=64   out=64   g=1   D=56     2982.6 GFlops (57.97%)     20.8 GB/s (12.95%) limited by gflops 57.97% algo=winograd
   9     resnet  forward b=64 k=1  p=0 s=2 in=1024 out=2048 g=1   D=14     1544.6 GFlops (30.02%)     10.0 GB/s ( 6.26%) limited by gflops 30.02% algo=gemm
   9     resnet bwd-data b=64 k=1  p=0 s=2 in=1024 out=2048 g=1   D=14     1537.8 GFlops (29.89%)     10.0 GB/s ( 6.23%) limited by gflops 29.89% algo=gemm
   9     resnet bwd-filt b=64 k=1  p=0 s=2 in=1024 out=2048 g=1   D=14     1485.4 GFlops (28.87%)     10.6 GB/s ( 6.61%) limited by gflops 28.87% algo=gemm
  10     resnet  forward b=64 k=1  p=0 s=1 in=1024 out=256  g=1   D=14     2117.9 GFlops (41.17%)     21.0 GB/s (13.10%) limited by gflops 41.17% algo=gemm
  10     resnet bwd-data b=64 k=1  p=0 s=1 in=1024 out=256  g=1   D=14     1695.8 GFlops (32.96%)     16.8 GB/s (10.49%) limited by gflops 32.96% algo=gemm
  10     resnet bwd-filt b=64 k=1  p=0 s=1 in=1024 out=256  g=1   D=14     1062.0 GFlops (20.64%)     10.7 GB/s ( 6.68%) limited by gflops 20.64% algo=gemm
  11     resnet  forward b=64 k=3  p=1 s=1 in=256  out=256  g=1   D=14     3249.9 GFlops (63.17%)      6.2 GB/s ( 3.84%) limited by gflops 63.17% algo=winograd
  11     resnet bwd-data b=64 k=3  p=1 s=1 in=256  out=256  g=1   D=14     2442.8 GFlops (47.48%)      4.6 GB/s ( 2.89%) limited by gflops 47.48% algo=winograd
  11     resnet bwd-filt b=64 k=3  p=1 s=1 in=256  out=256  g=1   D=14     3089.7 GFlops (60.06%)      6.3 GB/s ( 3.96%) limited by gflops 60.06% algo=winograd
  12        vgg  forward b=64 k=3  p=1 s=1 in=3    out=64   g=1   D=224     846.4 GFlops (16.45%)     65.6 GB/s (40.92%) limited by memory 40.92% algo=gemm
  12        vgg bwd-data b=64 k=3  p=1 s=1 in=3    out=64   g=1   D=224     278.9 GFlops ( 5.42%)     21.6 GB/s (13.48%) limited by memory 13.48% algo=gemm
  12        vgg bwd-filt b=64 k=3  p=1 s=1 in=3    out=64   g=1   D=224     144.7 GFlops ( 2.81%)     11.2 GB/s ( 6.99%) limited by memory  6.99% algo=winograd
  13        vgg  forward b=64 k=3  p=1 s=1 in=64   out=64   g=1   D=224    2864.1 GFlops (55.67%)     19.9 GB/s (12.40%) limited by gflops 55.67% algo=winograd
  13        vgg bwd-data b=64 k=3  p=1 s=1 in=64   out=64   g=1   D=224     848.0 GFlops (16.48%)      5.9 GB/s ( 3.67%) limited by gflops 16.48% algo=winograd
  13        vgg bwd-filt b=64 k=3  p=1 s=1 in=64   out=64   g=1   D=224    2754.5 GFlops (53.54%)     19.1 GB/s (11.93%) limited by gflops 53.54% algo=winograd
  14        vgg  forward b=64 k=3  p=1 s=1 in=512  out=512  g=1   D=28     2704.0 GFlops (52.56%)      2.5 GB/s ( 1.53%) limited by gflops 52.56% algo=winograd
  14        vgg bwd-data b=64 k=3  p=1 s=1 in=512  out=512  g=1   D=28     3092.4 GFlops (60.11%)      2.8 GB/s ( 1.75%) limited by gflops 60.11% algo=winograd
  14        vgg bwd-filt b=64 k=3  p=1 s=1 in=512  out=512  g=1   D=28     3103.1 GFlops (60.32%)      2.9 GB/s ( 1.83%) limited by gflops 60.32% algo=winograd
  15     mobile  forward b=64 k=3  p=1 s=2 in=3    out=32   g=1   D=224     699.0 GFlops (13.59%)     71.2 GB/s (44.38%) limited by memory 44.38% algo=gemm
  15     mobile bwd-data b=64 k=3  p=1 s=2 in=3    out=32   g=1   D=224     247.4 GFlops ( 4.81%)     25.2 GB/s (15.71%) limited by memory 15.71% algo=gemm
  15     mobile bwd-filt b=64 k=3  p=1 s=2 in=3    out=32   g=1   D=224      90.2 GFlops ( 1.75%)      9.2 GB/s ( 5.73%) limited by memory  5.73% algo=gemm
  16     mobile  forward b=64 k=3  p=1 s=1 in=144  out=144  g=144 D=56      125.9 GFlops ( 2.45%)     56.0 GB/s (34.89%) limited by memory 34.89% algo=depthwise_separable
  16     mobile bwd-data b=64 k=3  p=1 s=1 in=144  out=144  g=144 D=56       46.2 GFlops ( 0.90%)     20.5 GB/s (12.80%) limited by memory 12.80% algo=depthwise_separable
  16     mobile bwd-filt b=64 k=3  p=1 s=1 in=144  out=144  g=144 D=56       38.7 GFlops ( 0.75%)     17.2 GB/s (10.73%) limited by memory 10.73% algo=depthwise_separable
  17     mobile  forward b=64 k=3  p=1 s=2 in=144  out=144  g=144 D=56        9.5 GFlops ( 0.18%)     10.6 GB/s ( 6.59%) limited by memory  6.59% algo=gemm
  17     mobile bwd-data b=64 k=3  p=1 s=2 in=144  out=144  g=144 D=56        8.0 GFlops ( 0.16%)      8.9 GB/s ( 5.56%) limited by memory  5.56% algo=gemm
  17     mobile bwd-filt b=64 k=3  p=1 s=2 in=144  out=144  g=144 D=56       26.4 GFlops ( 0.51%)     29.4 GB/s (18.32%) limited by memory 18.32% algo=gemm
  18     mobile  forward b=64 k=1  p=0 s=1 in=144  out=24   g=1   D=56      707.9 GFlops (13.76%)     68.8 GB/s (42.91%) limited by memory 42.91% algo=gemm
  18     mobile bwd-data b=64 k=1  p=0 s=1 in=144  out=24   g=1   D=56      317.2 GFlops ( 6.16%)     30.8 GB/s (19.23%) limited by memory 19.23% algo=gemm
  18     mobile bwd-filt b=64 k=1  p=0 s=1 in=144  out=24   g=1   D=56      542.5 GFlops (10.54%)     52.8 GB/s (32.89%) limited by memory 32.89% algo=gemm
  19     mobile  forward b=64 k=1  p=0 s=1 in=24   out=144  g=1   D=56     1015.0 GFlops (19.73%)     98.7 GB/s (61.53%) limited by memory 61.53% algo=gemm
  19     mobile bwd-data b=64 k=1  p=0 s=1 in=24   out=144  g=1   D=56      488.9 GFlops ( 9.50%)     47.5 GB/s (29.63%) limited by memory 29.63% algo=gemm
  19     mobile bwd-filt b=64 k=1  p=0 s=1 in=24   out=144  g=1   D=56      530.4 GFlops (10.31%)     51.6 GB/s (32.15%) limited by memory 32.15% algo=gemm
  20     mobile  forward b=64 k=1  p=0 s=1 in=960  out=160  g=1   D=7      1887.2 GFlops (36.68%)     28.7 GB/s (17.91%) limited by gflops 36.68% algo=gemm
  20     mobile bwd-data b=64 k=1  p=0 s=1 in=960  out=160  g=1   D=7       877.2 GFlops (17.05%)     13.4 GB/s ( 8.32%) limited by gflops 17.05% algo=gemm
  20     mobile bwd-filt b=64 k=1  p=0 s=1 in=960  out=160  g=1   D=7       977.5 GFlops (19.00%)     15.5 GB/s ( 9.66%) limited by gflops 19.00% algo=gemm
  21     mobile  forward b=64 k=1  p=0 s=1 in=960  out=320  g=1   D=7      1805.6 GFlops (35.10%)     16.2 GB/s (10.10%) limited by gflops 35.10% algo=gemm
  21     mobile bwd-data b=64 k=1  p=0 s=1 in=960  out=320  g=1   D=7      1562.7 GFlops (30.38%)     14.0 GB/s ( 8.74%) limited by gflops 30.38% algo=gemm
  21     mobile bwd-filt b=64 k=1  p=0 s=1 in=960  out=320  g=1   D=7      1025.1 GFlops (19.93%)      9.9 GB/s ( 6.14%) limited by gflops 19.93% algo=gemm
  22     mobile  forward b=64 k=3  p=1 s=1 in=960  out=960  g=960 D=7        65.1 GFlops ( 1.27%)     29.0 GB/s (18.06%) limited by memory 18.06% algo=depthwise_separable
  22     mobile bwd-data b=64 k=3  p=1 s=1 in=960  out=960  g=960 D=7        23.7 GFlops ( 0.46%)     10.5 GB/s ( 6.56%) limited by memory  6.56% algo=depthwise_separable
  22     mobile bwd-filt b=64 k=3  p=1 s=1 in=960  out=960  g=960 D=7        14.2 GFlops ( 0.28%)      6.3 GB/s ( 3.95%) limited by memory  3.95% algo=depthwise_separable
  23      scale  forward b=64 k=1  p=0 s=1 in=256  out=256  g=256 D=56       29.4 GFlops ( 0.57%)    117.6 GB/s (73.31%) limited by memory 73.31% algo=depthwise_separable
  23      scale bwd-data b=64 k=1  p=0 s=1 in=256  out=256  g=256 D=56       15.5 GFlops ( 0.30%)     62.2 GB/s (38.76%) limited by memory 38.76% algo=depthwise_separable
  23      scale bwd-filt b=64 k=1  p=0 s=1 in=256  out=256  g=256 D=56       32.5 GFlops ( 0.63%)    129.9 GB/s (80.98%) limited by memory 80.98% algo=depthwise_separable
  24      scale  forward b=64 k=1  p=0 s=1 in=1024 out=1024 g=1024 D=7        17.3 GFlops ( 0.34%)     69.3 GB/s (43.18%) limited by memory 43.18% algo=depthwise_separable
  24      scale bwd-data b=64 k=1  p=0 s=1 in=1024 out=1024 g=1024 D=7        10.3 GFlops ( 0.20%)     41.2 GB/s (25.69%) limited by memory 25.69% algo=depthwise_separable
  24      scale bwd-filt b=64 k=1  p=0 s=1 in=1024 out=1024 g=1024 D=7         2.1 GFlops ( 0.04%)      8.5 GB/s ( 5.27%) limited by memory  5.27% algo=depthwise_separable
Broadcast/Reduce
     float (64,512,24,24)  (64,512,24,24)  (64,512,24,24)        30.6 GFlops ( 0.59%)    183.6 GB/s (114.43%) limited by memory 114.43%
     float (64,512,24,24)  (512,1,1)       (64,512,24,24)        23.1 GFlops ( 0.45%)     92.4 GB/s (57.59%) limited by memory 57.59%
     float (64,512,24,24)  (1,512,1,1)     (1,512,1,1)           20.1 GFlops ( 0.39%)     26.8 GB/s (16.72%) limited by memory 16.72%
     float (64,512,24,24)  (64,512,24,24)  (1,512,1,1)           17.6 GFlops ( 0.34%)     46.9 GB/s (29.25%) limited by memory 29.25%
     float (64,512,24,24)  (64,512,24,24)  (64,1,1,1)            49.1 GFlops ( 0.95%)    130.8 GB/s (81.55%) limited by memory 81.55%
     float (256,1000)      (256,1)         (1)                   23.7 GFlops ( 0.46%)     31.7 GB/s (19.74%) limited by memory 19.74%
      long (64,512,24,24)  (64,512,24,24)  (64,512,24,24)        15.7 GFlops ( 0.30%)    188.3 GB/s (117.37%) limited by memory 117.37%
      long (64,512,24,24)  (512,1,1)       (64,512,24,24)        20.3 GFlops ( 0.39%)    162.5 GB/s (101.32%) limited by memory 101.32%
      long (64,512,24,24)  (1,512,1,1)     (1,512,1,1)           19.1 GFlops ( 0.37%)     51.0 GB/s (31.81%) limited by memory 31.81%
      long (64,512,24,24)  (64,512,24,24)  (1,512,1,1)           12.3 GFlops ( 0.24%)     65.5 GB/s (40.83%) limited by memory 40.83%
      long (64,512,24,24)  (64,512,24,24)  (64,1,1,1)            12.0 GFlops ( 0.23%)     63.9 GB/s (39.81%) limited by memory 39.81%
      long (256,1000)      (256,1)         (1)                   22.5 GFlops ( 0.44%)     60.1 GB/s (37.49%) limited by memory 37.49%
     short (64,512,24,24)  (64,512,24,24)  (64,512,24,24)        50.2 GFlops ( 0.98%)    150.6 GB/s (93.86%) limited by memory 93.86%
     short (64,512,24,24)  (512,1,1)       (64,512,24,24)        22.7 GFlops ( 0.44%)     45.4 GB/s (28.32%) limited by memory 28.32%
     short (64,512,24,24)  (1,512,1,1)     (1,512,1,1)           20.0 GFlops ( 0.39%)     13.3 GB/s ( 8.32%) limited by memory  8.32%
     short (64,512,24,24)  (64,512,24,24)  (1,512,1,1)           17.5 GFlops ( 0.34%)     23.3 GB/s (14.55%) limited by memory 14.55%
     short (64,512,24,24)  (64,512,24,24)  (64,1,1,1)            88.1 GFlops ( 1.71%)    117.5 GB/s (73.25%) limited by memory 73.25%
     short (256,1000)      (256,1)         (1)                   23.8 GFlops ( 0.46%)     15.9 GB/s ( 9.88%) limited by memory  9.88%


$ ./dlprim_benchmark 1:0 ../docs/nets_for_benchmark/resnet18-b16.js 
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5     80.330
Step -4     36.077
Step -3     35.337
Step -2     35.630
Step -1     35.008
Step  0     34.952
Step  1     35.214
Step  2     34.761
Step  3     35.090
Step  4     35.260
Step  5     35.024
Step  6     34.899
Step  7     35.728
Step  8     35.166
Step  9     35.458
Step 10     34.814
Step 11     35.465
Step 12     35.240
Step 13     35.293
Step 14     34.999
Step 15     34.989
Step 16     34.958
Step 17     35.108
Step 18     35.097
Step 19     35.087
Time per sample: 2.196 ms
TOT time per batch:  35.130 ms

$ ./dlprim_benchmark 1:0 ../docs/nets_for_benchmark/resnet50-b16.js 
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5    132.289
Step -4     91.184
Step -3     90.216
Step -2     90.509
Step -1     90.597
Step  0     90.742
Step  1     90.965
Step  2     90.320
Step  3     91.756
Step  4     90.960
Step  5     90.557
Step  6     90.805
Step  7     90.848
Step  8     92.255
Step  9     91.454
Step 10     91.886
Step 11     91.015
Step 12     90.576
Step 13     90.880
Step 14     91.256
Step 15     90.494
Step 16     91.185
Step 17     90.390
Step 18     90.666
Step 19     90.921
Time per sample: 5.687 ms
TOT time per batch:  90.996 ms

$ ./dlprim_benchmark 1:0 ../docs/nets_for_benchmark/vgg16-b16.js 
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5    185.414
Step -4    185.442
Step -3    188.930
Step -2    186.092
Step -1    185.907
Step  0    187.530
Step  1    186.507
Step  2    185.399
Step  3    185.453
Step  4    185.883
Step  5    187.993
Step  6    186.317
Step  7    187.125
Step  8    187.122
Step  9    187.363
Step 10    186.331
Step 11    184.917
Step 12    188.269
Step 13    187.451
Step 14    185.449
Step 15    187.041
Step 16    186.303
Step 17    185.764
Step 18    187.314
Step 19    188.277
Time per sample: 11.668 ms
TOT time per batch:  186.690 ms

$ ./dlprim_benchmark -b 1:0 ../docs/nets_for_benchmark/vgg16-b16.js 
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5    770.122   187.484   582.638
Step -4    743.399   186.710   556.690
Step -3    761.797   188.238   573.559
Step -2    743.536   187.428   556.108
Step -1    742.265   185.859   556.406
Step  0    746.577   186.576   560.001
Step  1    743.734   186.139   557.595
Step  2    743.463   186.844   556.619
Step  3    744.668   187.335   557.332
Step  4    741.201   187.092   554.109
Step  5    743.327   186.306   557.021
Step  6    744.942   187.064   557.878
Step  7    740.202   186.793   553.409
Step  8    743.599   188.014   555.585
Step  9    739.598   186.159   553.439
Step 10    743.211   185.697   557.514
Step 11    743.652   186.177   557.475
Step 12    744.369   186.880   557.490
Step 13    741.782   186.042   555.741
Step 14    740.910   186.574   554.336
Step 15    750.555   189.514   561.041
Step 16    744.013   188.112   555.901
Step 17    744.750   188.831   555.919
Step 18    738.204   185.355   552.849
Step 19    757.115   189.433   567.681
Time per sample: 46.500 ms
FWD time per batch:  187.047 ms
BWD time per batch:  556.947 ms
TOT time per batch:  743.994 ms

$ ./dlprim_benchmark -b 1:0 ../docs/nets_for_benchmark/resnet50-b16.js 
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5    447.884   161.900   285.984
Step -4    407.349   120.718   286.630
Step -3    405.071   120.815   284.256
Step -2    403.187   119.322   283.865
Step -1    406.504   122.619   283.886
Step  0    407.171   120.855   286.316
Step  1    407.500   121.306   286.194
Step  2    405.939   120.074   285.865
Step  3    405.650   120.879   284.770
Step  4    404.071   119.365   284.706
Step  5    405.826   120.930   284.895
Step  6    406.268   121.494   284.774
Step  7    408.355   120.590   287.766
Step  8    409.165   122.810   286.354
Step  9    405.334   119.881   285.453
Step 10    403.831   120.196   283.635
Step 11    405.789   119.233   286.556
Step 12    405.945   121.241   284.704
Step 13    405.756   120.543   285.212
Step 14    404.247   120.621   283.627
Step 15    404.761   120.222   284.539
Step 16    405.180   120.644   284.536
Step 17    405.355   120.172   285.183
Step 18    407.071   121.578   285.492
Step 19    404.739   120.551   284.188
Time per sample: 25.369 ms
FWD time per batch:  120.659 ms
BWD time per batch:  285.238 ms
TOT time per batch:  405.898 ms

$ ./dlprim_benchmark -b 1:0 ../docs/nets_for_benchmark/resnet18-b16.js
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5    181.940    85.001    96.940
Step -4    137.857    41.113    96.743
Step -3    137.407    40.562    96.846
Step -2    137.063    40.797    96.266
Step -1    137.592    40.736    96.856
Step  0    137.711    41.241    96.469
Step  1    137.795    40.820    96.976
Step  2    137.608    40.523    97.085
Step  3    136.721    40.684    96.037
Step  4    136.594    40.909    95.684
Step  5    137.399    40.675    96.724
Step  6    138.939    41.656    97.283
Step  7    139.907    42.870    97.037
Step  8    138.866    41.203    97.664
Step  9    137.604    40.593    97.012
Step 10    137.740    40.678    97.062
Step 11    137.307    40.457    96.850
Step 12    138.597    41.025    97.573
Step 13    141.730    41.928    99.801
Step 14    138.254    40.460    97.794
Step 15    137.741    41.115    96.626
Step 16    136.908    40.174    96.734
Step 17    136.797    40.137    96.660
Step 18    136.989    40.397    96.592
Step 19    137.402    40.729    96.673
Time per sample: 8.621 ms
FWD time per batch:  40.914 ms
BWD time per batch:  97.017 ms
TOT time per batch:  137.930 ms

$ ./dlprim_benchmark -b 1:0 ../docs/nets_for_benchmark/mobilenet_v2-b16.js
Using: Ellesmere on AMD Accelerated Parallel Processing
Inputs
- data: (16,3,224,224)
Outputs
- loss: (16,1000)
Step -5    256.687   104.412   152.275
Step -4    206.847    56.856   149.990
Step -3    206.469    57.477   148.992
Step -2    207.143    57.319   149.824
Step -1    205.645    56.432   149.214
Step  0    206.860    57.350   149.510
Step  1    206.530    56.938   149.592
Step  2    206.432    56.736   149.696
Step  3    207.289    57.767   149.522
Step  4    206.740    57.231   149.509
Step  5    207.204    57.766   149.438
Step  6    210.465    59.797   150.667
Step  7    207.892    57.249   150.643
Step  8    206.792    57.428   149.365
Step  9    206.696    57.349   149.346
Step 10    206.496    57.284   149.211
Step 11    206.400    57.229   149.171
Step 12    206.679    57.180   149.499
Step 13    206.885    57.544   149.341
Step 14    207.031    57.509   149.522
Step 15    206.900    57.606   149.294
Step 16    206.717    57.345   149.373
Step 17    207.377    57.390   149.986
Step 18    206.459    57.476   148.984
Step 19    206.174    57.283   148.892
Time per sample: 12.938 ms
FWD time per batch:  57.473 ms
BWD time per batch:  149.528 ms
TOT time per batch:  207.001 ms

$ clinfo 
Number of platforms                               3
  Platform Name                                   Clover
  Platform Vendor                                 Mesa
  Platform Version                                OpenCL 1.1 Mesa 22.0.5
  Platform Profile                                FULL_PROFILE
  Platform Extensions                             cl_khr_icd
  Platform Extensions function suffix             MESA

  Platform Name                                   AMD Accelerated Parallel Processing
  Platform Vendor                                 Advanced Micro Devices, Inc.
  Platform Version                                OpenCL 2.1 AMD-APP (3380.4)
  Platform Profile                                FULL_PROFILE
  Platform Extensions                             cl_khr_icd cl_amd_event_callback cl_amd_offline_devices 
  Platform Extensions function suffix             AMD
  Platform Host timer resolution                  1ns

  Platform Name                                   Portable Computing Language
  Platform Vendor                                 The pocl project
  Platform Version                                OpenCL 2.0 pocl 1.8  Linux, None+Asserts, RELOC, LLVM 11.1.0, SLEEF, DISTRO, POCL_DEBUG
  Platform Profile                                FULL_PROFILE
  Platform Extensions                             cl_khr_icd cl_pocl_content_size
  Platform Extensions function suffix             POCL

  Platform Name                                   Clover
Number of devices                                 1
  Device Name                                     AMD Radeon (TM) RX 480 Graphics (polaris10, LLVM 13.0.1, DRM 3.46, 5.15.0-47-generic)
  Device Vendor                                   AMD
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 1.1 Mesa 22.0.5
  Device Numeric Version                          0x401000 (1.1.0)
  Driver Version                                  22.0.5
  Device OpenCL C Version                         OpenCL C 1.1 
  Device Type                                     GPU
  Device Profile                                  FULL_PROFILE
  Device Available                                Yes
  Compiler Available                              Yes
  Max compute units                               36
  Max clock frequency                             1303MHz
  Max work item dimensions                        3
  Max work item sizes                             256x256x256
  Max work group size                             256
  Preferred work group size multiple (kernel)     64
  Preferred / native vector sizes                 
    char                                                16 / 16      
    short                                                8 / 8       
    int                                                  4 / 4       
    long                                                 2 / 2       
    half                                                 0 / 0        (n/a)
    float                                                4 / 4       
    double                                               2 / 2        (cl_khr_fp64)
  Half-precision Floating-point support           (n/a)
  Single-precision Floating-point support         (core)
    Denormals                                     No
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 No
    Round to infinity                             No
    IEEE754-2008 fused multiply-add               No
    Support is emulated in software               No
    Correctly-rounded divide and sqrt operations  No
  Double-precision Floating-point support         (cl_khr_fp64)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
  Address bits                                    64, Little-Endian
  Global memory size                              27487790692 (25.6GiB)
  Error Correction support                        No
  Max memory allocation                           6871947673 (6.4GiB)
  Unified memory for Host and Device              No
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       32768 bits (4096 bytes)
  Global Memory cache type                        None
  Image support                                   No
  Local memory type                               Local
  Local memory size                               32768 (32KiB)
  Max number of constant args                     16
  Max constant buffer size                        67108864 (64MiB)
  Max size of kernel argument                     1024
  Queue properties                                
    Out-of-order execution                        No
    Profiling                                     Yes
  Profiling timer resolution                      0ns
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            No
    ILs with version                              (n/a)
  Built-in kernels with version                   (n/a)
  Device Extensions                               cl_khr_byte_addressable_store cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp64 cl_khr_extended_versioning
  Device Extensions with Version                  cl_khr_byte_addressable_store                                    0x400000 (1.0.0)
                                                  cl_khr_global_int32_base_atomics                                 0x400000 (1.0.0)
                                                  cl_khr_global_int32_extended_atomics                             0x400000 (1.0.0)
                                                  cl_khr_local_int32_base_atomics                                  0x400000 (1.0.0)
                                                  cl_khr_local_int32_extended_atomics                              0x400000 (1.0.0)
                                                  cl_khr_int64_base_atomics                                        0x400000 (1.0.0)
                                                  cl_khr_int64_extended_atomics                                    0x400000 (1.0.0)
                                                  cl_khr_fp64                                                      0x400000 (1.0.0)
                                                  cl_khr_extended_versioning                                       0x400000 (1.0.0)

  Platform Name                                   AMD Accelerated Parallel Processing
Number of devices                                 1
  Device Name                                     Ellesmere
  Device Vendor                                   Advanced Micro Devices, Inc.
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 2.0 AMD-APP (3380.4)
  Driver Version                                  3380.4 (PAL,HSAIL)
  Device OpenCL C Version                         OpenCL C 2.0 
  Device Type                                     GPU
  Device Board Name (AMD)                         AMD Radeon (TM) RX 480 Graphics
  Device PCI-e ID (AMD)                           0x67df
  Device Topology (AMD)                           PCI-E, 0000:01:00.0
  Device Profile                                  FULL_PROFILE
  Device Available                                Yes
  Compiler Available                              Yes
  Linker Available                                Yes
  Max compute units                               36
  SIMD per compute unit (AMD)                     4
  SIMD width (AMD)                                16
  SIMD instruction width (AMD)                    1
  Max clock frequency                             1303MHz
  Graphics IP (AMD)                               8.0
  Device Partition                                (core)
    Max number of sub-devices                     36
    Supported partition types                     None
    Supported affinity domains                    (n/a)
  Max work item dimensions                        3
  Max work item sizes                             1024x1024x1024
  Max work group size                             256
  Preferred work group size (AMD)                 256
  Max work group size (AMD)                       1024
  Preferred work group size multiple (kernel)     64
  Wavefront width (AMD)                           64
  Preferred / native vector sizes                 
    char                                                 4 / 4       
    short                                                2 / 2       
    int                                                  1 / 1       
    long                                                 1 / 1       
    half                                                 1 / 1        (cl_khr_fp16)
    float                                                1 / 1       
    double                                               1 / 1        (cl_khr_fp64)
  Half-precision Floating-point support           (cl_khr_fp16)
    Denormals                                     No
    Infinity and NANs                             No
    Round to nearest                              No
    Round to zero                                 No
    Round to infinity                             No
    IEEE754-2008 fused multiply-add               No
    Support is emulated in software               No
  Single-precision Floating-point support         (core)
    Denormals                                     No
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
    Correctly-rounded divide and sqrt operations  Yes
  Double-precision Floating-point support         (cl_khr_fp64)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
  Address bits                                    64, Little-Endian
  Global memory size                              8589934592 (8GiB)
  Global free memory (AMD)                        8323072 (7.938GiB) 8060928 (7.688GiB)
  Global memory channels (AMD)                    8
  Global memory banks per channel (AMD)           4
  Global memory bank width (AMD)                  256 bytes
  Error Correction support                        No
  Max memory allocation                           7073274265 (6.587GiB)
  Unified memory for Host and Device              No
  Shared Virtual Memory (SVM) capabilities        (core)
    Coarse-grained buffer sharing                 Yes
    Fine-grained buffer sharing                   Yes
    Fine-grained system sharing                   No
    Atomics                                       No
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       2048 bits (256 bytes)
  Preferred alignment for atomics                 
    SVM                                           0 bytes
    Global                                        0 bytes
    Local                                         0 bytes
  Max size for global variable                    6365946624 (5.929GiB)
  Preferred total size of global vars             8589934592 (8GiB)
  Global Memory cache type                        Read/Write
  Global Memory cache size                        16384 (16KiB)
  Global Memory cache line size                   64 bytes
  Image support                                   Yes
    Max number of samplers per kernel             16
    Max size for 1D images from buffer            442079641 pixels
    Max 1D or 2D image array size                 2048 images
    Base address alignment for 2D image buffers   256 bytes
    Pitch alignment for 2D image buffers          256 pixels
    Max 2D image size                             16384x16384 pixels
    Max 3D image size                             2048x2048x2048 pixels
    Max number of read image args                 128
    Max number of write image args                64
    Max number of read/write image args           64
  Max number of pipe args                         16
  Max active pipe reservations                    16
  Max pipe packet size                            2778306969 (2.587GiB)
  Local memory type                               Local
  Local memory size                               65536 (64KiB)
  Local memory size per CU (AMD)                  65536 (64KiB)
  Local memory banks (AMD)                        32
  Max number of constant args                     8
  Max constant buffer size                        7073274265 (6.587GiB)
  Preferred constant buffer size (AMD)            16384 (16KiB)
  Max size of kernel argument                     1024
  Queue properties (on host)                      
    Out-of-order execution                        No
    Profiling                                     Yes
  Queue properties (on device)                    
    Out-of-order execution                        Yes
    Profiling                                     Yes
    Preferred size                                262144 (256KiB)
    Max size                                      8388608 (8MiB)
  Max queues on device                            1
  Max events on device                            1024
  Prefer user sync for interop                    Yes
  Number of P2P devices (AMD)                     0
  Profiling timer resolution                      1ns
  Profiling timer offset since Epoch (AMD)        1662878599093686350ns (Sun Sep 11 08:43:19 2022)
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            No
    Thread trace supported (AMD)                  Yes
    Number of async queues (AMD)                  4
    Max real-time compute queues (AMD)            1
    Max real-time compute units (AMD)             0
  printf() buffer size                            4194304 (4MiB)
  Built-in kernels                                (n/a)
  Device Extensions                               cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_fp16 cl_khr_gl_sharing cl_khr_gl_depth_images cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_image2d_from_buffer cl_khr_subgroups cl_khr_gl_event cl_khr_depth_images cl_khr_mipmap_image cl_khr_mipmap_image_writes cl_amd_copy_buffer_p2p 

  Platform Name                                   Portable Computing Language
Number of devices                                 1
  Device Name                                     pthread-Intel(R) Core(TM) i3-9100F CPU @ 3.60GHz
  Device Vendor                                   GenuineIntel
  Device Vendor ID                                0x8086
  Device Version                                  OpenCL 1.2 pocl HSTR: pthread-x86_64-pc-linux-gnu-skylake
  Driver Version                                  1.8
  Device OpenCL C Version                         OpenCL C 1.2 pocl
  Device Type                                     CPU
  Device Profile                                  FULL_PROFILE
  Device Available                                Yes
  Compiler Available                              Yes
  Linker Available                                Yes
  Max compute units                               4
  Max clock frequency                             4200MHz
  Device Partition                                (core)
    Max number of sub-devices                     4
    Supported partition types                     equally, by counts
    Supported affinity domains                    (n/a)
  Max work item dimensions                        3
  Max work item sizes                             4096x4096x4096
  Max work group size                             4096
  Preferred work group size multiple (kernel)     8
  Preferred / native vector sizes                 
    char                                                16 / 16      
    short                                               16 / 16      
    int                                                  8 / 8       
    long                                                 4 / 4       
    half                                                 0 / 0        (n/a)
    float                                                8 / 8       
    double                                               4 / 4        (cl_khr_fp64)
  Half-precision Floating-point support           (n/a)
  Single-precision Floating-point support         (core)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
    Correctly-rounded divide and sqrt operations  Yes
  Double-precision Floating-point support         (cl_khr_fp64)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
  Address bits                                    64, Little-Endian
  Global memory size                              31450091520 (29.29GiB)
  Error Correction support                        No
  Max memory allocation                           8589934592 (8GiB)
  Unified memory for Host and Device              Yes
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       1024 bits (128 bytes)
  Global Memory cache type                        Read/Write
  Global Memory cache size                        6291456 (6MiB)
  Global Memory cache line size                   64 bytes
  Image support                                   Yes
    Max number of samplers per kernel             16
    Max size for 1D images from buffer            536870912 pixels
    Max 1D or 2D image array size                 2048 images
    Max 2D image size                             16384x16384 pixels
    Max 3D image size                             2048x2048x2048 pixels
    Max number of read image args                 128
    Max number of write image args                128
  Local memory type                               Global
  Local memory size                               262144 (256KiB)
  Max number of constant args                     8
  Max constant buffer size                        262144 (256KiB)
  Max size of kernel argument                     1024
  Queue properties                                
    Out-of-order execution                        Yes
    Profiling                                     Yes
  Prefer user sync for interop                    Yes
  Profiling timer resolution                      1ns
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            Yes
  printf() buffer size                            16777216 (16MiB)
  Built-in kernels                                (n/a)
  Device Extensions                               cl_khr_byte_addressable_store cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_fp64 cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp64


NULL platform behavior
  clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...)  No platform
  clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...)   No platform
  clCreateContext(NULL, ...) [default]            No platform
  clCreateContext(NULL, ...) [other]              Success [MESA]
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_DEFAULT)  Success (1)
    Platform Name                                 Clover
    Device Name                                   AMD Radeon (TM) RX 480 Graphics (polaris10, LLVM 13.0.1, DRM 3.46, 5.15.0-47-generic)
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU)  Success (1)
    Platform Name                                 Clover
    Device Name                                   AMD Radeon (TM) RX 480 Graphics (polaris10, LLVM 13.0.1, DRM 3.46, 5.15.0-47-generic)
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL)  Success (1)
    Platform Name                                 Clover
    Device Name                                   AMD Radeon (TM) RX 480 Graphics (polaris10, LLVM 13.0.1, DRM 3.46, 5.15.0-47-generic)

artyom-beilis · 2022-09-11T10:29:18Z

artyom-beilis
Sep 11, 2022
Maintainer

Thanks for the benchmarks

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Benchmark on RX480, Ubuntu 22 #20

{{title}}

{{editor}}'s edit

{{editor}}'s edit

Replies: 1 comment

{{title}}

Select a reply

Benchmark on RX480, Ubuntu 22 #20

dagelf Sep 11, 2022

Replies: 1 comment

artyom-beilis Sep 11, 2022 Maintainer

dagelf
Sep 11, 2022

artyom-beilis
Sep 11, 2022
Maintainer