cumo 0.7.0 → 0.9.0

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 (135) hide show
  1. checksums.yaml +4 -4
  2. data/.rubocop_todo.yml +590 -152
  3. data/CHANGELOG.md +63 -0
  4. data/README.md +245 -0
  5. data/bench/cumo_probe.rb +1 -1
  6. data/cumo.gemspec +7 -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/memory_pool_impl.cpp +32 -22
  11. data/ext/cumo/cuda/memory_pool_impl.hpp +3 -1
  12. data/ext/cumo/cuda/memory_pool_impl_test.cpp +19 -25
  13. data/ext/cumo/cuda/runtime.c +56 -0
  14. data/ext/cumo/cumo.c +4 -0
  15. data/ext/cumo/extconf.rb +2 -0
  16. data/ext/cumo/include/cumo/bit_reduce_kernel.h +6 -6
  17. data/ext/cumo/include/cumo/check.h +11 -3
  18. data/ext/cumo/include/cumo/cuda/cudnn.h +43 -6
  19. data/ext/cumo/include/cumo/cuda/runtime.h +2 -0
  20. data/ext/cumo/include/cumo/indexer.h +16 -0
  21. data/ext/cumo/include/cumo/intern.h +11 -0
  22. data/ext/cumo/include/cumo/narray.h +1 -0
  23. data/ext/cumo/include/cumo/reduce_kernel.h +283 -54
  24. data/ext/cumo/include/cumo/row_kernel.h +202 -0
  25. data/ext/cumo/include/cumo/row_method.h +149 -0
  26. data/ext/cumo/include/cumo/template_kernel.h +26 -0
  27. data/ext/cumo/include/cumo/types/bf16_def.h +82 -0
  28. data/ext/cumo/include/cumo/types/bf16_def_kernel.h +28 -0
  29. data/ext/cumo/include/cumo/types/bf16_macro.h +14 -0
  30. data/ext/cumo/include/cumo/types/bf16_macro_kernel.h +19 -0
  31. data/ext/cumo/include/cumo/types/bfloat.h +22 -0
  32. data/ext/cumo/include/cumo/types/bfloat_kernel.h +16 -0
  33. data/ext/cumo/include/cumo/types/bit.h +2 -0
  34. data/ext/cumo/include/cumo/types/bit_kernel.h +2 -0
  35. data/ext/cumo/include/cumo/types/complex_macro.h +2 -0
  36. data/ext/cumo/include/cumo/types/complex_macro_kernel.h +2 -0
  37. data/ext/cumo/include/cumo/types/f16_macro.h +191 -0
  38. data/ext/cumo/include/cumo/types/f16_macro_kernel.h +214 -0
  39. data/ext/cumo/include/cumo/types/float_def.h +8 -0
  40. data/ext/cumo/include/cumo/types/float_def_kernel.h +8 -0
  41. data/ext/cumo/include/cumo/types/float_macro.h +26 -17
  42. data/ext/cumo/include/cumo/types/float_macro_kernel.h +26 -19
  43. data/ext/cumo/include/cumo/types/half_macro.h +6 -180
  44. data/ext/cumo/include/cumo/types/half_macro_kernel.h +6 -185
  45. data/ext/cumo/include/cumo/types/robj_macro.h +2 -0
  46. data/ext/cumo/include/cumo/types/xint_macro.h +2 -6
  47. data/ext/cumo/include/cumo/types/xint_macro_kernel.h +2 -0
  48. data/ext/cumo/include/cumo.h +2 -2
  49. data/ext/cumo/narray/data.c +15 -6
  50. data/ext/cumo/narray/data_kernel.cu +110 -0
  51. data/ext/cumo/narray/gen/def/bfloat.rb +57 -0
  52. data/ext/cumo/narray/gen/def/bit.rb +12 -1
  53. data/ext/cumo/narray/gen/def/dcomplex.rb +12 -1
  54. data/ext/cumo/narray/gen/def/dfloat.rb +11 -1
  55. data/ext/cumo/narray/gen/def/hfloat.rb +11 -1
  56. data/ext/cumo/narray/gen/def/int16.rb +12 -1
  57. data/ext/cumo/narray/gen/def/int32.rb +12 -1
  58. data/ext/cumo/narray/gen/def/int64.rb +12 -1
  59. data/ext/cumo/narray/gen/def/int8.rb +12 -1
  60. data/ext/cumo/narray/gen/def/robject.rb +12 -1
  61. data/ext/cumo/narray/gen/def/scomplex.rb +12 -1
  62. data/ext/cumo/narray/gen/def/sfloat.rb +11 -1
  63. data/ext/cumo/narray/gen/def/uint16.rb +12 -1
  64. data/ext/cumo/narray/gen/def/uint32.rb +12 -1
  65. data/ext/cumo/narray/gen/def/uint64.rb +12 -1
  66. data/ext/cumo/narray/gen/def/uint8.rb +12 -1
  67. data/ext/cumo/narray/gen/narray_def.rb +35 -6
  68. data/ext/cumo/narray/gen/spec.rb +14 -14
  69. data/ext/cumo/narray/gen/tmpl/accum_arg.c +1 -1
  70. data/ext/cumo/narray/gen/tmpl/accum_binary.c +63 -5
  71. data/ext/cumo/narray/gen/tmpl/accum_binary_kernel.cu +29 -13
  72. data/ext/cumo/narray/gen/tmpl/accum_index.c +1 -1
  73. data/ext/cumo/narray/gen/tmpl/batch_norm.c +4 -4
  74. data/ext/cumo/narray/gen/tmpl/batch_norm_backward.c +5 -5
  75. data/ext/cumo/narray/gen/tmpl/binary.c +3 -9
  76. data/ext/cumo/narray/gen/tmpl/complex_accum_kernel.cu +15 -5
  77. data/ext/cumo/narray/gen/tmpl/conv.c +90 -123
  78. data/ext/cumo/narray/gen/tmpl/conv_grad_w.c +91 -83
  79. data/ext/cumo/narray/gen/tmpl/conv_transpose.c +90 -123
  80. data/ext/cumo/narray/gen/tmpl/cum.c +8 -7
  81. data/ext/cumo/narray/gen/tmpl/cum_kernel.cu +7 -6
  82. data/ext/cumo/narray/gen/tmpl/fixed_batch_norm.c +4 -4
  83. data/ext/cumo/narray/gen/tmpl/float_accum_kernel.cu +22 -10
  84. data/ext/cumo/narray/gen/tmpl/gemm.c +11 -41
  85. data/ext/cumo/narray/gen/tmpl/layer_norm.c +38 -0
  86. data/ext/cumo/narray/gen/tmpl/layer_norm_kernel.cu +54 -0
  87. data/ext/cumo/narray/gen/tmpl/lib.c +1 -0
  88. data/ext/cumo/narray/gen/tmpl/lib_kernel.cu +1 -0
  89. data/ext/cumo/narray/gen/tmpl/median.c +1 -1
  90. data/ext/cumo/narray/gen/tmpl/poly.c +0 -3
  91. data/ext/cumo/narray/gen/tmpl/poly_kernel.cu +1 -1
  92. data/ext/cumo/narray/gen/tmpl/pooling_backward.c +3 -2
  93. data/ext/cumo/narray/gen/tmpl/pooling_forward.c +3 -2
  94. data/ext/cumo/narray/gen/tmpl/quantize_symmetric.c +74 -0
  95. data/ext/cumo/narray/gen/tmpl/quantize_symmetric_kernel.cu +67 -0
  96. data/ext/cumo/narray/gen/tmpl/rand.c +9 -9
  97. data/ext/cumo/narray/gen/tmpl/rand_kernel.cu +9 -9
  98. data/ext/cumo/narray/gen/tmpl/rand_norm.c +4 -4
  99. data/ext/cumo/narray/gen/tmpl/rand_norm_kernel.cu +7 -7
  100. data/ext/cumo/narray/gen/tmpl/real_accum_kernel.cu +6 -6
  101. data/ext/cumo/narray/gen/tmpl/rms_norm.c +38 -0
  102. data/ext/cumo/narray/gen/tmpl/rms_norm_kernel.cu +57 -0
  103. data/ext/cumo/narray/gen/tmpl/softmax.c +21 -0
  104. data/ext/cumo/narray/gen/tmpl/softmax_kernel.cu +157 -0
  105. data/ext/cumo/narray/gen/tmpl/sort.c +9 -32
  106. data/ext/cumo/narray/gen/tmpl/sort_index.c +16 -95
  107. data/ext/cumo/narray/gen/tmpl/store_array.c +12 -9
  108. data/ext/cumo/narray/gen/tmpl/store_from.c +1 -16
  109. data/ext/cumo/narray/gen/tmpl_bit/mask.c +11 -11
  110. data/ext/cumo/narray/gen/tmpl_bit/store_array.c +5 -9
  111. data/ext/cumo/narray/gen/tmpl_bit/where.c +50 -11
  112. data/ext/cumo/narray/gen/tmpl_bit/where2.c +11 -12
  113. data/ext/cumo/narray/gen/tmpl_bit/where_kernel.cu +1 -2
  114. data/ext/cumo/narray/index.c +35 -25
  115. data/ext/cumo/narray/index_kernel.cu +27 -0
  116. data/ext/cumo/narray/math.c +40 -8
  117. data/ext/cumo/narray/narray.c +85 -15
  118. data/ext/cumo/narray/ndloop.c +210 -5
  119. data/ext/cumo/narray/sort_kernel.cu +61 -41
  120. data/ext/cumo/narray/struct.c +3 -0
  121. data/lib/cumo/cuda/cudnn.rb +1 -1
  122. data/lib/cumo/narray/extra.rb +2 -2
  123. data/test/bfloat_test.rb +238 -0
  124. data/test/bit_test.rb +52 -14
  125. data/test/cuda/memory_pool_test.rb +8 -7
  126. data/test/cuda/runtime_test.rb +25 -0
  127. data/test/cudnn_test.rb +108 -24
  128. data/test/extra_test.rb +1 -1
  129. data/test/fused_test.rb +587 -0
  130. data/test/math_test.rb +182 -2
  131. data/test/narray_alt_coverage_test.rb +1 -0
  132. data/test/narray_test.rb +344 -22
  133. data/test/test_helper.rb +1 -0
  134. metadata +32 -5
  135. 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: 0f27ebdaa195fa967c144d6178ad3f9fbeca551162f8edacac205d5edd5d0fb6
4
+ data.tar.gz: 3d87faa55ab7c4c0e52bee3afdb0116d9d463fa778c408a46e200bff0cabdf81
5
5
  SHA512:
6
- metadata.gz: 314d2b224cb753761618b6bc833af4715d8d1cb2e1fd8de05b0ad65483ec940c31167997abe7cc76e2f056b12b442885731d8e6b4d4b47ec174c19c70c2326e9
7
- data.tar.gz: 5a4909d952dbe926b52ccaf04010b3ffcf08617a2af710ff49a70764ff7e626ffddcef04ebcb4ca6b5aab8c7b14170b9b3328768a203dd7961d2fbab5c973002
6
+ metadata.gz: 4b498a760e59725c15aa3e651ecbada01e52517833fd3132a860dbfe459e354f4afba47207cc33f2c94963f58f19a8a1d7e52ce722e51065876d637c8ddfc94b
7
+ data.tar.gz: bbda42b2ef7ce7ea882982a015bec06ff058ad42a2b72e4c8953e75facfcb64657c8da4b69d63666cf4c06d3c265dc7e0acf54c431450f181b39cbeebed9560e