cumo 0.7.0 → 0.8.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (117) hide show
  1. checksums.yaml +4 -4
  2. data/.rubocop_todo.yml +590 -152
  3. data/CHANGELOG.md +29 -0
  4. data/README.md +140 -0
  5. data/bench/cumo_probe.rb +1 -1
  6. data/cumo.gemspec +1 -1
  7. data/ext/cumo/cuda/cudnn.c +18 -0
  8. data/ext/cumo/cuda/cudnn_impl.cpp +60 -14
  9. data/ext/cumo/cuda/driver.c +11 -9
  10. data/ext/cumo/cuda/runtime.c +56 -0
  11. data/ext/cumo/cumo.c +4 -0
  12. data/ext/cumo/extconf.rb +2 -0
  13. data/ext/cumo/include/cumo/bit_reduce_kernel.h +6 -6
  14. data/ext/cumo/include/cumo/cuda/cudnn.h +40 -0
  15. data/ext/cumo/include/cumo/cuda/runtime.h +2 -0
  16. data/ext/cumo/include/cumo/intern.h +9 -0
  17. data/ext/cumo/include/cumo/narray.h +1 -0
  18. data/ext/cumo/include/cumo/reduce_kernel.h +32 -28
  19. data/ext/cumo/include/cumo/row_kernel.h +170 -0
  20. data/ext/cumo/include/cumo/row_method.h +48 -0
  21. data/ext/cumo/include/cumo/template_kernel.h +26 -0
  22. data/ext/cumo/include/cumo/types/bf16_def.h +82 -0
  23. data/ext/cumo/include/cumo/types/bf16_def_kernel.h +28 -0
  24. data/ext/cumo/include/cumo/types/bf16_macro.h +186 -0
  25. data/ext/cumo/include/cumo/types/bf16_macro_kernel.h +213 -0
  26. data/ext/cumo/include/cumo/types/bfloat.h +22 -0
  27. data/ext/cumo/include/cumo/types/bfloat_kernel.h +16 -0
  28. data/ext/cumo/include/cumo/types/bit.h +2 -0
  29. data/ext/cumo/include/cumo/types/bit_kernel.h +2 -0
  30. data/ext/cumo/include/cumo/types/complex_macro.h +2 -0
  31. data/ext/cumo/include/cumo/types/complex_macro_kernel.h +2 -0
  32. data/ext/cumo/include/cumo/types/float_def.h +8 -0
  33. data/ext/cumo/include/cumo/types/float_def_kernel.h +8 -0
  34. data/ext/cumo/include/cumo/types/float_macro.h +19 -17
  35. data/ext/cumo/include/cumo/types/float_macro_kernel.h +19 -19
  36. data/ext/cumo/include/cumo/types/half_macro.h +15 -17
  37. data/ext/cumo/include/cumo/types/half_macro_kernel.h +15 -0
  38. data/ext/cumo/include/cumo/types/robj_macro.h +2 -0
  39. data/ext/cumo/include/cumo/types/xint_macro.h +2 -6
  40. data/ext/cumo/include/cumo/types/xint_macro_kernel.h +2 -0
  41. data/ext/cumo/include/cumo.h +2 -2
  42. data/ext/cumo/narray/gen/def/bfloat.rb +56 -0
  43. data/ext/cumo/narray/gen/def/bit.rb +11 -1
  44. data/ext/cumo/narray/gen/def/dcomplex.rb +11 -1
  45. data/ext/cumo/narray/gen/def/dfloat.rb +10 -1
  46. data/ext/cumo/narray/gen/def/hfloat.rb +10 -1
  47. data/ext/cumo/narray/gen/def/int16.rb +11 -1
  48. data/ext/cumo/narray/gen/def/int32.rb +11 -1
  49. data/ext/cumo/narray/gen/def/int64.rb +11 -1
  50. data/ext/cumo/narray/gen/def/int8.rb +11 -1
  51. data/ext/cumo/narray/gen/def/robject.rb +11 -1
  52. data/ext/cumo/narray/gen/def/scomplex.rb +11 -1
  53. data/ext/cumo/narray/gen/def/sfloat.rb +10 -1
  54. data/ext/cumo/narray/gen/def/uint16.rb +11 -1
  55. data/ext/cumo/narray/gen/def/uint32.rb +11 -1
  56. data/ext/cumo/narray/gen/def/uint64.rb +11 -1
  57. data/ext/cumo/narray/gen/def/uint8.rb +11 -1
  58. data/ext/cumo/narray/gen/narray_def.rb +0 -5
  59. data/ext/cumo/narray/gen/spec.rb +11 -14
  60. data/ext/cumo/narray/gen/tmpl/accum_arg.c +1 -1
  61. data/ext/cumo/narray/gen/tmpl/accum_binary_kernel.cu +9 -9
  62. data/ext/cumo/narray/gen/tmpl/accum_index.c +1 -1
  63. data/ext/cumo/narray/gen/tmpl/batch_norm.c +3 -3
  64. data/ext/cumo/narray/gen/tmpl/batch_norm_backward.c +3 -3
  65. data/ext/cumo/narray/gen/tmpl/complex_accum_kernel.cu +15 -5
  66. data/ext/cumo/narray/gen/tmpl/conv.c +88 -121
  67. data/ext/cumo/narray/gen/tmpl/conv_grad_w.c +89 -81
  68. data/ext/cumo/narray/gen/tmpl/conv_transpose.c +88 -121
  69. data/ext/cumo/narray/gen/tmpl/cum.c +8 -7
  70. data/ext/cumo/narray/gen/tmpl/cum_kernel.cu +7 -6
  71. data/ext/cumo/narray/gen/tmpl/fixed_batch_norm.c +3 -3
  72. data/ext/cumo/narray/gen/tmpl/float_accum_kernel.cu +22 -10
  73. data/ext/cumo/narray/gen/tmpl/gemm.c +11 -35
  74. data/ext/cumo/narray/gen/tmpl/layer_norm.c +91 -0
  75. data/ext/cumo/narray/gen/tmpl/layer_norm_kernel.cu +54 -0
  76. data/ext/cumo/narray/gen/tmpl/lib.c +1 -0
  77. data/ext/cumo/narray/gen/tmpl/lib_kernel.cu +1 -0
  78. data/ext/cumo/narray/gen/tmpl/median.c +1 -1
  79. data/ext/cumo/narray/gen/tmpl/poly.c +0 -3
  80. data/ext/cumo/narray/gen/tmpl/poly_kernel.cu +1 -1
  81. data/ext/cumo/narray/gen/tmpl/pooling_backward.c +2 -1
  82. data/ext/cumo/narray/gen/tmpl/pooling_forward.c +2 -1
  83. data/ext/cumo/narray/gen/tmpl/rand.c +9 -9
  84. data/ext/cumo/narray/gen/tmpl/rand_kernel.cu +9 -9
  85. data/ext/cumo/narray/gen/tmpl/rand_norm.c +4 -4
  86. data/ext/cumo/narray/gen/tmpl/rand_norm_kernel.cu +7 -7
  87. data/ext/cumo/narray/gen/tmpl/real_accum_kernel.cu +6 -6
  88. data/ext/cumo/narray/gen/tmpl/softmax.c +56 -0
  89. data/ext/cumo/narray/gen/tmpl/softmax_kernel.cu +157 -0
  90. data/ext/cumo/narray/gen/tmpl/sort.c +9 -32
  91. data/ext/cumo/narray/gen/tmpl/sort_index.c +16 -95
  92. data/ext/cumo/narray/gen/tmpl/store_array.c +12 -9
  93. data/ext/cumo/narray/gen/tmpl_bit/mask.c +11 -11
  94. data/ext/cumo/narray/gen/tmpl_bit/store_array.c +5 -9
  95. data/ext/cumo/narray/gen/tmpl_bit/where.c +50 -11
  96. data/ext/cumo/narray/gen/tmpl_bit/where2.c +11 -12
  97. data/ext/cumo/narray/gen/tmpl_bit/where_kernel.cu +1 -2
  98. data/ext/cumo/narray/index.c +2 -1
  99. data/ext/cumo/narray/math.c +2 -0
  100. data/ext/cumo/narray/narray.c +27 -1
  101. data/ext/cumo/narray/ndloop.c +73 -4
  102. data/ext/cumo/narray/sort_kernel.cu +61 -41
  103. data/ext/cumo/narray/struct.c +3 -0
  104. data/lib/cumo/cuda/cudnn.rb +1 -1
  105. data/lib/cumo/narray/extra.rb +2 -2
  106. data/test/bfloat_test.rb +238 -0
  107. data/test/cuda/memory_pool_test.rb +8 -7
  108. data/test/cuda/runtime_test.rb +25 -0
  109. data/test/cudnn_test.rb +108 -24
  110. data/test/extra_test.rb +1 -1
  111. data/test/fused_test.rb +311 -0
  112. data/test/math_test.rb +77 -2
  113. data/test/narray_alt_coverage_test.rb +1 -0
  114. data/test/narray_test.rb +70 -10
  115. data/test/test_helper.rb +1 -0
  116. metadata +20 -4
  117. data/ext/cumo/narray/gen/tmpl/qsort.c +0 -205
checksums.yaml CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: 97c7f575209616cb9fa484e5dd4096bac11170aef18e99f1f6f733e599daa9af
4
- data.tar.gz: 5efabd7aface13fac85dcef81fef54dd9bb22d3748666177ff221741a8419486
3
+ metadata.gz: 695bd66c93bd2511f905d74cf242a7d2d7d0ac4aea8f5915fb722e013e658e9a
4
+ data.tar.gz: fa51db7ef5f01745ded6d71834bde62fa238fb92cba9bb358374293a18b7577b
5
5
  SHA512:
6
- metadata.gz: 314d2b224cb753761618b6bc833af4715d8d1cb2e1fd8de05b0ad65483ec940c31167997abe7cc76e2f056b12b442885731d8e6b4d4b47ec174c19c70c2326e9
7
- data.tar.gz: 5a4909d952dbe926b52ccaf04010b3ffcf08617a2af710ff49a70764ff7e626ffddcef04ebcb4ca6b5aab8c7b14170b9b3328768a203dd7961d2fbab5c973002
6
+ metadata.gz: fbad56229183188dc36b2341105a68a5ca670de3a86ac9d50887544bebf4200e80c0f571783aefe24657c609aa86bc3f6af0da8d5943b56c0c91031ea7c22244
7
+ data.tar.gz: a5a3e9cfa2e99873b3e7f555fec3e397ed8c9814fbd0255ccc2e0cd52ac9d337fd654741a00412f1b0eb46a777ed346fa46ae98c7e4bf45ce0d7eb8de964fa75