block_recur_perf_27-03-2022.out

Julia Version 1.7.2
Commit bf53498635 (2022-02-06 15:21 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM) i7-8700K CPU @ 3.70GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-12.0.1 (ORCJIT, skylake)
Environment:
  JULIA_VERSION = 1.7
  JULIA_LOAD_PATH = /home/matt/Documents/Flux.jl/test/:/home/matt/Documents/Flux.jl/experiment/:
RNN Vec CPU n=2, ts=1
  forward
  264.172 ns (4 allocations: 288 bytes)
  backward
  9.584 μs (117 allocations: 6.41 KiB)
  forw and back
  19.113 μs (212 allocations: 31.94 KiB)
RNN Vec CUDA n=2, ts=1
  forward
  91.317 μs (93 allocations: 6.45 KiB)
  backward
  219.462 μs (377 allocations: 18.89 KiB)
  forw and back
  335.586 μs (617 allocations: 39.53 KiB)
RNN Vec CPU n=2, ts=4
  forward
  700.552 ns (13 allocations: 1.00 KiB)
  backward
  27.689 μs (393 allocations: 22.86 KiB)
  forw and back
  46.552 μs (584 allocations: 92.36 KiB)
RNN Vec CUDA n=2, ts=4
  forward
  171.984 μs (369 allocations: 25.64 KiB)
  backward
  496.415 μs (1478 allocations: 78.22 KiB)
  forw and back
  725.018 μs (2195 allocations: 141.38 KiB)
RNN Vec CPU n=2, ts=16
  forward
  2.415 μs (49 allocations: 3.91 KiB)
  backward
  88.227 μs (1498 allocations: 88.41 KiB)
  forw and back
  136.871 μs (2074 allocations: 333.61 KiB)
RNN Vec CUDA n=2, ts=16
  forward
  461.223 μs (1473 allocations: 102.42 KiB)
  backward
  1.469 ms (5882 allocations: 315.09 KiB)
  forw and back
  2.131 ms (8507 allocations: 547.91 KiB)
RNN Vec CPU n=2, ts=64
  forward
  9.167 μs (193 allocations: 15.55 KiB)
  backward
  326.686 μs (5914 allocations: 350.95 KiB)
  forw and back
  494.513 μs (8026 allocations: 1.27 MiB)
RNN Vec CUDA n=2, ts=64
  forward
  1.599 ms (5889 allocations: 409.56 KiB)
  backward
  5.257 ms (23499 allocations: 1.23 MiB)
  forw and back
  7.637 ms (33757 allocations: 2.12 MiB)
RNN Vec CPU n=20, ts=1
  forward
  2.141 μs (4 allocations: 3.73 KiB)
  backward
  13.331 μs (117 allocations: 16.67 KiB)
  forw and back
  25.025 μs (212 allocations: 49.03 KiB)
RNN Vec CUDA n=20, ts=1
  forward
  94.467 μs (93 allocations: 6.45 KiB)
  backward
  218.825 μs (377 allocations: 18.89 KiB)
  forw and back
  337.910 μs (617 allocations: 39.53 KiB)
RNN Vec CPU n=20, ts=4
  forward
  9.432 μs (13 allocations: 19.64 KiB)
  backward
  46.884 μs (393 allocations: 83.42 KiB)
  forw and back
  74.057 μs (584 allocations: 190.12 KiB)
RNN Vec CUDA n=20, ts=4
  forward
  175.496 μs (369 allocations: 25.64 KiB)
  backward
  497.927 μs (1481 allocations: 78.27 KiB)
  forw and back
  732.290 μs (2198 allocations: 141.42 KiB)
RNN Vec CPU n=20, ts=16
  forward
  39.080 μs (49 allocations: 83.30 KiB)
  backward
  172.115 μs (1498 allocations: 350.16 KiB)
  forw and back
  259.511 μs (2074 allocations: 754.06 KiB)
RNN Vec CUDA n=20, ts=16
  forward
  470.589 μs (1473 allocations: 102.42 KiB)
  backward
  1.477 ms (5897 allocations: 315.33 KiB)
  forw and back
  2.144 ms (8522 allocations: 548.14 KiB)
RNN Vec CPU n=20, ts=64
  forward
  157.590 μs (193 allocations: 337.94 KiB)
  backward
  672.133 μs (5914 allocations: 1.38 MiB)
  forw and back
  999.957 μs (8026 allocations: 2.94 MiB)
RNN Vec CUDA n=20, ts=64
  forward
  1.639 ms (5889 allocations: 409.56 KiB)
  backward
  5.262 ms (23562 allocations: 1.23 MiB)
  forw and back
  7.663 ms (33820 allocations: 2.12 MiB)
RNN Vec CPU n=200, ts=1
  forward
  396.093 μs (6 allocations: 313.53 KiB)
  backward
  431.399 μs (122 allocations: 946.98 KiB)
  forw and back
  657.308 μs (221 allocations: 1.56 MiB)
RNN Vec CUDA n=200, ts=1
  forward
  103.224 μs (94 allocations: 6.47 KiB)
  backward
  236.049 μs (380 allocations: 18.94 KiB)
  forw and back
  372.350 μs (664 allocations: 41.84 KiB)
RNN Vec CPU n=200, ts=4
  forward
  899.520 μs (24 allocations: 1.68 MiB)
  backward
  1.427 ms (422 allocations: 5.52 MiB)
  forw and back
  2.166 ms (635 allocations: 8.95 MiB)
RNN Vec CUDA n=200, ts=4
  forward
  203.731 μs (373 allocations: 25.70 KiB)
  backward
  542.370 μs (1496 allocations: 78.50 KiB)
  forw and back
  809.731 μs (2266 allocations: 144.06 KiB)
RNN Vec CPU n=200, ts=16
  forward
  3.709 ms (96 allocations: 7.17 MiB)
  backward
  6.234 ms (1623 allocations: 23.92 MiB)
  forw and back
  9.189 ms (2293 allocations: 38.51 MiB)
RNN Vec CUDA n=200, ts=16
  forward
  563.140 μs (1489 allocations: 102.67 KiB)
  backward
  1.662 ms (5960 allocations: 316.31 KiB)
  forw and back
  2.424 ms (8674 allocations: 552.09 KiB)
RNN Vec CPU n=200, ts=64
  forward
  15.238 ms (384 allocations: 29.15 MiB)
  backward
  25.381 ms (6423 allocations: 97.52 MiB)
  forw and back
  39.505 ms (8917 allocations: 156.73 MiB)
RNN Vec CUDA n=200, ts=64
  forward
  1.961 ms (5953 allocations: 410.56 KiB)
  backward
  6.072 ms (23817 allocations: 1.24 MiB)
  forw and back
  9.022 ms (34308 allocations: 2.13 MiB)
RNN Vec CPU n=1000, ts=1
  forward
  7.714 ms (6 allocations: 7.63 MiB)
  backward
  13.385 ms (122 allocations: 22.91 MiB)
  forw and back
  19.695 ms (221 allocations: 38.20 MiB)
RNN Vec CUDA n=1000, ts=1
  forward
  527.965 μs (168 allocations: 9.81 KiB)
  backward
  1.195 ms (476 allocations: 22.62 KiB)
  forw and back
  1.896 ms (834 allocations: 48.88 KiB)
RNN Vec CPU n=1000, ts=4
  forward
  62.978 ms (24 allocations: 41.97 MiB)
  backward
  97.428 ms (422 allocations: 137.40 MiB)
  forw and back
  145.270 ms (635 allocations: 221.40 MiB)
RNN Vec CUDA n=1000, ts=4
  forward
  2.298 ms (537 allocations: 30.45 KiB)
  backward
  4.818 ms (1745 allocations: 84.58 KiB)
  forw and back
  7.162 ms (2679 allocations: 154.89 KiB)
RNN Vec CPU n=1000, ts=16
  forward
  266.436 ms (96 allocations: 179.30 MiB)
  backward
  424.925 ms (1623 allocations: 595.36 MiB)
  forw and back
  621.772 ms (2293 allocations: 954.19 MiB)
RNN Vec CUDA n=1000, ts=16
  forward
  7.745 ms (2013 allocations: 113.05 KiB)
  backward
  19.628 ms (6821 allocations: 331.95 KiB)
  forw and back
  28.773 ms (10059 allocations: 578.11 KiB)
RNN Vec CPU n=1000, ts=64
  forward
  1.092 s (384 allocations: 728.62 MiB)
  backward
  1.725 s (6423 allocations: 2.37 GiB)
  forw and back
  2.519 s (8917 allocations: 3.79 GiB)
RNN Vec CUDA n=1000, ts=64
  forward
  30.842 ms (7917 allocations: 443.44 KiB)
  backward
  79.725 ms (27126 allocations: 1.29 MiB)
  forw and back
  117.308 ms (39581 allocations: 2.22 MiB)
RNN Block CPU n=2, ts=1
  forward
  1.175 μs (14 allocations: 640 bytes)
  backward
  10.660 μs (132 allocations: 6.89 KiB)
  forw and back
  29.692 μs (261 allocations: 36.22 KiB)
RNN Block CUDA n=2, ts=1
  forward
  124.708 μs (137 allocations: 8.02 KiB)
  backward
  249.134 μs (439 allocations: 21.31 KiB)
  forw and back
  417.532 μs (747 allocations: 46.67 KiB)
RNN Block CPU n=2, ts=4
  forward
  1.690 μs (23 allocations: 1.39 KiB)
  backward
  28.258 μs (423 allocations: 24.09 KiB)
  forw and back
  58.840 μs (642 allocations: 98.61 KiB)
RNN Block CUDA n=2, ts=4
  forward
  214.211 μs (503 allocations: 30.58 KiB)
  backward
  547.575 μs (1678 allocations: 86.39 KiB)
  forw and back
  837.096 μs (2466 allocations: 150.77 KiB)
RNN Block CPU n=2, ts=16
  forward
  3.546 μs (59 allocations: 4.50 KiB)
  backward
  92.498 μs (1588 allocations: 92.56 KiB)
  forw and back
  155.328 μs (2168 allocations: 347.64 KiB)
RNN Block CUDA n=2, ts=16
  forward
  554.636 μs (1967 allocations: 120.86 KiB)
  backward
  1.607 ms (6634 allocations: 346.55 KiB)
  forw and back
  2.333 ms (9342 allocations: 566.83 KiB)
RNN Block CPU n=2, ts=64
  forward
  10.812 μs (203 allocations: 16.88 KiB)
  backward
  345.428 μs (6244 allocations: 366.75 KiB)
  forw and back
  534.571 μs (8264 allocations: 1.31 MiB)
RNN Block CUDA n=2, ts=64
  forward
  1.886 ms (7823 allocations: 482.00 KiB)
  backward
  5.731 ms (26459 allocations: 1.35 MiB)
  forw and back
  8.162 ms (36848 allocations: 2.18 MiB)
RNN Block CPU n=20, ts=1
  forward
  3.241 μs (14 allocations: 5.77 KiB)
  backward
  14.694 μs (132 allocations: 18.84 KiB)
  forw and back
  35.691 μs (261 allocations: 56.69 KiB)
RNN Block CUDA n=20, ts=1
  forward
  123.940 μs (137 allocations: 8.02 KiB)
  backward
  250.611 μs (439 allocations: 21.31 KiB)
  forw and back
  416.753 μs (747 allocations: 46.67 KiB)
RNN Block CPU n=20, ts=4
  forward
  11.016 μs (23 allocations: 26.28 KiB)
  backward
  48.984 μs (423 allocations: 90.91 KiB)
  forw and back
  88.382 μs (642 allocations: 203.81 KiB)
RNN Block CUDA n=20, ts=4
  forward
  222.157 μs (503 allocations: 30.58 KiB)
  backward
  550.462 μs (1681 allocations: 86.44 KiB)
  forw and back
  855.562 μs (2512 allocations: 153.28 KiB)
RNN Block CPU n=20, ts=16
  forward
  42.542 μs (60 allocations: 108.61 KiB)
  backward
  179.538 μs (1589 allocations: 379.05 KiB)
  forw and back
  280.389 μs (2170 allocations: 792.23 KiB)
RNN Block CUDA n=20, ts=16
  forward
  568.705 μs (1967 allocations: 120.86 KiB)
  backward
  1.617 ms (6649 allocations: 346.78 KiB)
  forw and back
  2.371 ms (9400 allocations: 569.53 KiB)
RNN Block CPU n=20, ts=64
  forward
  163.198 μs (204 allocations: 438.25 KiB)
  backward
  704.353 μs (6245 allocations: 1.50 MiB)
  forw and back
  1.056 ms (8266 allocations: 3.07 MiB)
RNN Block CUDA n=20, ts=64
  forward
  1.954 ms (7823 allocations: 482.00 KiB)
  backward
  5.747 ms (26522 allocations: 1.36 MiB)
  forw and back
  8.229 ms (36954 allocations: 2.18 MiB)
RNN Block CPU n=200, ts=1
  forward
  206.012 μs (17 allocations: 470.09 KiB)
  backward
  289.603 μs (138 allocations: 1.08 MiB)
  forw and back
  455.803 μs (272 allocations: 1.87 MiB)
RNN Block CUDA n=200, ts=1
  forward
  131.274 μs (140 allocations: 8.06 KiB)
  backward
  262.678 μs (442 allocations: 21.36 KiB)
  forw and back
  450.127 μs (797 allocations: 49.25 KiB)
RNN Block CPU n=200, ts=4
  forward
  926.796 μs (35 allocations: 2.29 MiB)
  backward
  1.545 ms (453 allocations: 6.14 MiB)
  forw and back
  2.288 ms (692 allocations: 9.72 MiB)
RNN Block CUDA n=200, ts=4
  forward
  245.877 μs (515 allocations: 30.77 KiB)
  backward
  601.164 μs (1696 allocations: 86.67 KiB)
  forw and back
  924.747 μs (2543 allocations: 153.77 KiB)
RNN Block CPU n=200, ts=16
  forward
  3.883 ms (107 allocations: 9.62 MiB)
  backward
  6.601 ms (1714 allocations: 26.37 MiB)
  forw and back
  9.613 ms (2374 allocations: 41.11 MiB)
RNN Block CUDA n=200, ts=16
  forward
  655.443 μs (2015 allocations: 121.61 KiB)
  backward
  1.815 ms (6712 allocations: 347.77 KiB)
  forw and back
  2.756 ms (9527 allocations: 571.52 KiB)
RNN Block CPU n=200, ts=64
  forward
  15.904 ms (395 allocations: 38.92 MiB)
  backward
  26.704 ms (6754 allocations: 107.30 MiB)
  forw and back
  41.180 ms (9094 allocations: 166.69 MiB)
RNN Block CUDA n=200, ts=64
  forward
  2.262 ms (8015 allocations: 485.00 KiB)
  backward
  6.885 ms (26777 allocations: 1.36 MiB)
  forw and back
  9.792 ms (37465 allocations: 2.19 MiB)
RNN Block CPU n=1000, ts=1
  forward
  8.044 ms (19 allocations: 11.45 MiB)
  backward
  13.822 ms (138 allocations: 26.72 MiB)
  forw and back
  20.431 ms (274 allocations: 45.83 MiB)
RNN Block CUDA n=1000, ts=1
  forward
  611.569 μs (216 allocations: 11.44 KiB)
  backward
  1.185 ms (538 allocations: 25.05 KiB)
  forw and back
  1.819 ms (969 allocations: 56.31 KiB)
RNN Block CPU n=1000, ts=4
  forward
  66.125 ms (37 allocations: 57.23 MiB)
  backward
  101.235 ms (453 allocations: 152.66 MiB)
  forw and back
  150.217 ms (694 allocations: 240.47 MiB)
RNN Block CUDA n=1000, ts=4
  forward
  2.263 ms (681 allocations: 35.55 KiB)
  backward
  4.997 ms (1945 allocations: 92.75 KiB)
  forw and back
  7.396 ms (2958 allocations: 164.62 KiB)
RNN Block CPU n=1000, ts=16
  forward
  279.050 ms (109 allocations: 240.33 MiB)
  backward
  441.440 ms (1714 allocations: 656.40 MiB)
  forw and back
  642.376 ms (2376 allocations: 1019.05 MiB)
RNN Block CUDA n=1000, ts=16
  forward
  8.139 ms (2541 allocations: 132.02 KiB)
  backward
  20.460 ms (7573 allocations: 363.41 KiB)
  forw and back
  29.609 ms (10914 allocations: 597.56 KiB)
RNN Block CPU n=1000, ts=64
  forward
  1.131 s (397 allocations: 972.76 MiB)
  backward
  1.792 s (6754 allocations: 2.61 GiB)
  forw and back
  3.213 s (9096 allocations: 4.04 GiB)
RNN Block CUDA n=1000, ts=64
  forward
  32.230 ms (9981 allocations: 517.91 KiB)
  backward
  83.352 ms (30086 allocations: 1.41 MiB)
  forw and back
  121.319 ms (42740 allocations: 2.27 MiB)