API tools faq
paste
Advertisement
Guest User

CUTLASS tests: clang (32 shm, inline) vs. nvcc on sm_60

a guest
Dec 12th, 2017
71
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 10.64 KB | None | 0 0
raw download clone embed print report
  1.  
  2. ==========================================
  3. ## bin/dgemm_nn_sm35-60_clang_cuda_9.0
  4. ==========================================
  5.  
  6. ------------------------------------------------------------
  7. 7168x4096x4096, GEMM_nn, 29360128 C elements, 10 timing iterations
  8.  
  9. Avg runtime: 53.224 ms, total flops: 240576888832, GFLOP/s: 4520.08
  10.  
  11. Final wave_efficiency 1.0000, tiling_efficiency 8.0000
  12. Invoking kernel<<<(448, 256, 1), (1.y,64.x), 0, 0>>>(), 13 SM occupancy, 4096 split_k
  13. Avg runtime: 163.195 ms, total flops: 240576888832, GFLOP/s: 1474.17
  14.  
  15. Final wave_efficiency 1.0000, tiling_efficiency 16.0000
  16. Invoking kernel<<<(224, 128, 1), (1.y,64.x), 0, 0>>>(), 7 SM occupancy, 4096 split_k
  17. Avg runtime: 101.542 ms, total flops: 240576888832, GFLOP/s: 2369.24
  18.  
  19. Final wave_efficiency 1.0000, tiling_efficiency 32.0000
  20. Invoking kernel<<<(112, 64, 1), (1.y,256.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  21. Avg runtime: 67.952 ms, total flops: 240576888832, GFLOP/s: 3540.41
  22.  
  23. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  24. Invoking kernel<<<(56, 128, 1), (1.y,128.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  25. Avg runtime: 81.973 ms, total flops: 240576888832, GFLOP/s: 2934.82
  26.  
  27. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  28. Invoking kernel<<<(224, 32, 1), (1.y,128.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  29. Avg runtime: 58.602 ms, total flops: 240576888832, GFLOP/s: 4105.26
  30.  
  31. Final wave_efficiency 1.0000, tiling_efficiency 42.6667
  32. Invoking kernel<<<(112, 32, 1), (1.y,128.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  33. Avg runtime: 55.507 ms, total flops: 240576888832, GFLOP/s: 4334.18
  34.  
  35.  
  36.  
  37.  
  38. ==========================================
  39. ## bin/dgemm_nn_sm35-60_nvcc_9.0
  40. ==========================================
  41.  
  42. ------------------------------------------------------------
  43. 7168x4096x4096, GEMM_nn, 29360128 C elements, 10 timing iterations
  44.  
  45. Avg runtime: 53.219 ms, total flops: 240576888832, GFLOP/s: 4520.53
  46.  
  47. Final wave_efficiency 1.0000, tiling_efficiency 8.0000
  48. Invoking kernel<<<(448, 256, 1), (1.y,64.x), 0, 0>>>(), 13 SM occupancy, 4096 split_k
  49. Avg runtime: 106.596 ms, total flops: 240576888832, GFLOP/s: 2256.91
  50.  
  51. Final wave_efficiency 1.0000, tiling_efficiency 16.0000
  52. Invoking kernel<<<(224, 128, 1), (1.y,64.x), 0, 0>>>(), 7 SM occupancy, 4096 split_k
  53. Avg runtime: 61.048 ms, total flops: 240576888832, GFLOP/s: 3940.79
  54.  
  55. Final wave_efficiency 1.0000, tiling_efficiency 32.0000
  56. Invoking kernel<<<(112, 64, 1), (1.y,256.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  57. Avg runtime: 59.593 ms, total flops: 240576888832, GFLOP/s: 4036.99
  58.  
  59. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  60. Invoking kernel<<<(56, 128, 1), (1.y,128.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  61. Avg runtime: 57.847 ms, total flops: 240576888832, GFLOP/s: 4158.82
  62.  
  63. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  64. Invoking kernel<<<(224, 32, 1), (1.y,128.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  65. Avg runtime: 57.236 ms, total flops: 240576888832, GFLOP/s: 4203.25
  66.  
  67. Final wave_efficiency 1.0000, tiling_efficiency 42.6667
  68. Invoking kernel<<<(112, 32, 1), (1.y,128.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  69. Avg runtime: 54.754 ms, total flops: 240576888832, GFLOP/s: 4393.77
  70.  
  71.  
  72.  
  73.  
  74. ==========================================
  75. ## bin/igemm_nn_sm35-60_clang_cuda_9.0
  76. ==========================================
  77.  
  78. ------------------------------------------------------------
  79. 7168x4096x4096, GEMM_nn, 29360128 C elements, 10 timing iterations
  80.  
  81. CUDA error 30 [gemm.cu, 199]: unknown error
  82. CUDA error 30 [gemm.cu, 253]: unknown error
  83. CUDA error 30 [gemm.cu, 253]: unknown error
  84. CUDA error 30 [gemm.cu, 275]: unknown error
  85. CUDA error 30 [gemm.cu, 275]: unknown error
  86. CUDA error 30 [gemm.cu, 275]: unknown error
  87. CUDA error 30 [gemm.cu, 275]: unknown error
  88. CUDA error 30 [gemm.cu, 275]: unknown error
  89. CUDA error 30 [gemm.cu, 275]: unknown error
  90. CUDA error 30 [gemm.cu, 275]: unknown error
  91. CUDA error 30 [gemm.cu, 275]: unknown error
  92. CUDA error 30 [gemm.cu, 275]: unknown error
  93. CUDA error 30 [gemm.cu, 275]: unknown error
  94. Avg runtime: 0.000 ms, total flops: 240576888832, GFLOP/s: 1474123120.06
  95.  
  96. Final wave_efficiency 1.0000, tiling_efficiency 10.6667
  97. Invoking kernel<<<(448, 128, 1), (1.y,32.x), 0, 0>>>(), 28 SM occupancy, 4096 split_k
  98. Avg runtime: 54.396 ms, total flops: 240576888832, GFLOP/s: 4422.68
  99.  
  100. Final wave_efficiency 1.0000, tiling_efficiency 16.0000
  101. Invoking kernel<<<(224, 128, 1), (1.y,64.x), 0, 0>>>(), 20 SM occupancy, 4096 split_k
  102. Avg runtime: 48.562 ms, total flops: 240576888832, GFLOP/s: 4954.06
  103.  
  104. Final wave_efficiency 1.0000, tiling_efficiency 32.0000
  105. Invoking kernel<<<(112, 64, 1), (1.y,128.x), 0, 0>>>(), 7 SM occupancy, 4096 split_k
  106. Avg runtime: 41.002 ms, total flops: 240576888832, GFLOP/s: 5867.48
  107.  
  108. Final wave_efficiency 1.0000, tiling_efficiency 42.6667
  109. Invoking kernel<<<(56, 64, 1), (1.y,256.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  110. Avg runtime: 40.773 ms, total flops: 240576888832, GFLOP/s: 5900.44
  111.  
  112. Final wave_efficiency 1.0000, tiling_efficiency 42.6667
  113. Invoking kernel<<<(112, 32, 1), (1.y,256.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  114. Avg runtime: 40.614 ms, total flops: 240576888832, GFLOP/s: 5923.45
  115.  
  116. Final wave_efficiency 1.0000, tiling_efficiency 64.0000
  117. Invoking kernel<<<(56, 32, 1), (1.y,256.x), 0, 0>>>(), 1 SM occupancy, 4096 split_k
  118. Avg runtime: 42.676 ms, total flops: 240576888832, GFLOP/s: 5637.26
  119.  
  120.  
  121.  
  122.  
  123. ==========================================
  124. ## bin/igemm_nn_sm35-60_nvcc_9.0
  125. ==========================================
  126.  
  127. ------------------------------------------------------------
  128. 7168x4096x4096, GEMM_nn, 29360128 C elements, 10 timing iterations
  129.  
  130. CUDA error 30 [gemm.cu, 199]: unknown error
  131. CUDA error 30 [gemm.cu, 253]: unknown error
  132. CUDA error 30 [gemm.cu, 253]: unknown error
  133. CUDA error 30 [gemm.cu, 275]: unknown error
  134. CUDA error 30 [gemm.cu, 275]: unknown error
  135. CUDA error 30 [gemm.cu, 275]: unknown error
  136. CUDA error 30 [gemm.cu, 275]: unknown error
  137. CUDA error 30 [gemm.cu, 275]: unknown error
  138. CUDA error 30 [gemm.cu, 275]: unknown error
  139. CUDA error 30 [gemm.cu, 275]: unknown error
  140. CUDA error 30 [gemm.cu, 275]: unknown error
  141. CUDA error 30 [gemm.cu, 275]: unknown error
  142. CUDA error 30 [gemm.cu, 275]: unknown error
  143. Avg runtime: 0.001 ms, total flops: 240576888832, GFLOP/s: 170476816.05
  144.  
  145. Final wave_efficiency 1.0000, tiling_efficiency 10.6667
  146. Invoking kernel<<<(448, 128, 1), (1.y,32.x), 0, 0>>>(), 28 SM occupancy, 4096 split_k
  147. Avg runtime: 50.941 ms, total flops: 240576888832, GFLOP/s: 4722.66
  148.  
  149. Final wave_efficiency 1.0000, tiling_efficiency 16.0000
  150. Invoking kernel<<<(224, 128, 1), (1.y,64.x), 0, 0>>>(), 20 SM occupancy, 4096 split_k
  151. Avg runtime: 49.127 ms, total flops: 240576888832, GFLOP/s: 4897.09
  152.  
  153. Final wave_efficiency 1.0000, tiling_efficiency 32.0000
  154. Invoking kernel<<<(112, 64, 1), (1.y,128.x), 0, 0>>>(), 8 SM occupancy, 4096 split_k
  155. Avg runtime: 45.015 ms, total flops: 240576888832, GFLOP/s: 5344.39
  156.  
  157. Final wave_efficiency 1.0000, tiling_efficiency 42.6667
  158. Invoking kernel<<<(56, 64, 1), (1.y,256.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  159. Avg runtime: 42.090 ms, total flops: 240576888832, GFLOP/s: 5715.73
  160.  
  161. Final wave_efficiency 1.0000, tiling_efficiency 42.6667
  162. Invoking kernel<<<(112, 32, 1), (1.y,256.x), 0, 0>>>(), 3 SM occupancy, 4096 split_k
  163. Avg runtime: 41.746 ms, total flops: 240576888832, GFLOP/s: 5762.85
  164.  
  165. Final wave_efficiency 1.0000, tiling_efficiency 64.0000
  166. Invoking kernel<<<(56, 32, 1), (1.y,256.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  167. Avg runtime: 36.177 ms, total flops: 240576888832, GFLOP/s: 6650.04
  168.  
  169.  
  170.  
  171.  
  172. ==========================================
  173. ## bin/sgemm_nn_sm35-60_clang_cuda_9.0
  174. ==========================================
  175.  
  176. ------------------------------------------------------------
  177. 7168x4096x4096, GEMM_nn, 29360128 C elements, 10 timing iterations
  178.  
  179. Avg runtime: 26.732 ms, total flops: 240576888832, GFLOP/s: 8999.53
  180.  
  181. Final wave_efficiency 1.0000, tiling_efficiency 8.0000
  182. Invoking kernel<<<(448, 256, 1), (1.y,64.x), 0, 0>>>(), 14 SM occupancy, 4096 split_k
  183. Avg runtime: 72.281 ms, total flops: 240576888832, GFLOP/s: 3328.37
  184.  
  185. Final wave_efficiency 1.0000, tiling_efficiency 16.0000
  186. Invoking kernel<<<(224, 128, 1), (1.y,64.x), 0, 0>>>(), 16 SM occupancy, 4096 split_k
  187. Avg runtime: 42.854 ms, total flops: 240576888832, GFLOP/s: 5613.85
  188.  
  189. Final wave_efficiency 1.0000, tiling_efficiency 32.0000
  190. Invoking kernel<<<(112, 64, 1), (1.y,64.x), 0, 0>>>(), 8 SM occupancy, 4096 split_k
  191. Avg runtime: 30.098 ms, total flops: 240576888832, GFLOP/s: 7993.04
  192.  
  193. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  194. Invoking kernel<<<(56, 128, 1), (1.y,128.x), 0, 0>>>(), 5 SM occupancy, 4096 split_k
  195. Avg runtime: 34.436 ms, total flops: 240576888832, GFLOP/s: 6986.27
  196.  
  197. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  198. Invoking kernel<<<(224, 32, 1), (1.y,128.x), 0, 0>>>(), 6 SM occupancy, 4096 split_k
  199. Avg runtime: 31.245 ms, total flops: 240576888832, GFLOP/s: 7699.77
  200.  
  201. Final wave_efficiency 1.0000, tiling_efficiency 64.0000
  202. Invoking kernel<<<(56, 32, 1), (1.y,256.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  203. Avg runtime: 27.501 ms, total flops: 240576888832, GFLOP/s: 8748.00
  204.  
  205.  
  206.  
  207.  
  208. ==========================================
  209. ## bin/sgemm_nn_sm35-60_nvcc_9.0
  210. ==========================================
  211.  
  212. ------------------------------------------------------------
  213. 7168x4096x4096, GEMM_nn, 29360128 C elements, 10 timing iterations
  214.  
  215. Avg runtime: 26.723 ms, total flops: 240576888832, GFLOP/s: 9002.61
  216.  
  217. Final wave_efficiency 1.0000, tiling_efficiency 8.0000
  218. Invoking kernel<<<(448, 256, 1), (1.y,64.x), 0, 0>>>(), 24 SM occupancy, 4096 split_k
  219. Avg runtime: 98.982 ms, total flops: 240576888832, GFLOP/s: 2430.50
  220.  
  221. Final wave_efficiency 1.0000, tiling_efficiency 16.0000
  222. Invoking kernel<<<(224, 128, 1), (1.y,64.x), 0, 0>>>(), 16 SM occupancy, 4096 split_k
  223. Avg runtime: 35.309 ms, total flops: 240576888832, GFLOP/s: 6813.55
  224.  
  225. Final wave_efficiency 1.0000, tiling_efficiency 32.0000
  226. Invoking kernel<<<(112, 64, 1), (1.y,64.x), 0, 0>>>(), 8 SM occupancy, 4096 split_k
  227. Avg runtime: 27.971 ms, total flops: 240576888832, GFLOP/s: 8600.93
  228.  
  229. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  230. Invoking kernel<<<(56, 128, 1), (1.y,128.x), 0, 0>>>(), 6 SM occupancy, 4096 split_k
  231. Avg runtime: 29.545 ms, total flops: 240576888832, GFLOP/s: 8142.77
  232.  
  233. Final wave_efficiency 1.0000, tiling_efficiency 25.6000
  234. Invoking kernel<<<(224, 32, 1), (1.y,128.x), 0, 0>>>(), 6 SM occupancy, 4096 split_k
  235. Avg runtime: 30.165 ms, total flops: 240576888832, GFLOP/s: 7975.34
  236.  
  237. Final wave_efficiency 1.0000, tiling_efficiency 64.0000
  238. Invoking kernel<<<(56, 32, 1), (1.y,256.x), 0, 0>>>(), 2 SM occupancy, 4096 split_k
  239. Avg runtime: 27.154 ms, total flops: 240576888832, GFLOP/s: 8859.79
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!