Commits
Wes McKinney authored d54ab9a23aa
ARROW-737: [C++] Enable mutable buffer slices, SliceMutableBuffer function This patch also results in better microperformance on my laptop. ``` Run on (4 X 1200 MHz CPU s) 2017-04-01 22:35:30 Benchmark Time CPU Iterations ---------------------------------------------------------------------------------------- BM_WriteRecordBatch/1/min_time:1.000/real_time 53617 ns 53423 ns 26903 18.2136GB/s BM_WriteRecordBatch/4/min_time:1.000/real_time 48118 ns 47999 ns 26823 20.295GB/s BM_WriteRecordBatch/16/min_time:1.000/real_time 46679 ns 46562 ns 30409 20.9209GB/s BM_WriteRecordBatch/64/min_time:1.000/real_time 50583 ns 50344 ns 27308 19.3061GB/s BM_WriteRecordBatch/256/min_time:1.000/real_time 93474 ns 93259 ns 11720 10.4474GB/s BM_WriteRecordBatch/1024/min_time:1.000/real_time 254056 ns 253461 ns 5434 3.84389GB/s BM_WriteRecordBatch/4k/min_time:1.000/real_time 892292 ns 888924 ns 1210 1120.71MB/s BM_WriteRecordBatch/8k/min_time:1.000/real_time 1799323 ns 1795023 ns 754 555.764MB/s BM_ReadRecordBatch/1/min_time:1.000/real_time 2501 ns 2452 ns 586633 390.521GB/s BM_ReadRecordBatch/4/min_time:1.000/real_time 6921 ns 5577 ns 227670 141.095GB/s BM_ReadRecordBatch/16/min_time:1.000/real_time 15561 ns 14505 ns 67493 62.758GB/s BM_ReadRecordBatch/64/min_time:1.000/real_time 48583 ns 48242 ns 30070 20.1008GB/s BM_ReadRecordBatch/256/min_time:1.000/real_time 184637 ns 183306 ns 6660 5.2891GB/s BM_ReadRecordBatch/1024/min_time:1.000/real_time 734128 ns 729692 ns 1905 1.33024GB/s BM_ReadRecordBatch/4k/min_time:1.000/real_time 3027028 ns 3008020 ns 445 330.357MB/s BM_ReadRecordBatch/8k/min_time:1.000/real_time 6472022 ns 6426801 ns 211 154.511MB/s ``` before ``` Run on (4 X 1200 MHz CPU s) 2017-04-01 22:37:36 Benchmark Time CPU Iterations ---------------------------------------------------------------------------------------- BM_WriteRecordBatch/1/min_time:1.000/real_time 59317 ns 59202 ns 22727 16.4633GB/s BM_WriteRecordBatch/4/min_time:1.000/real_time 56322 ns 56191 ns 24944 17.3389GB/s BM_WriteRecordBatch/16/min_time:1.000/real_time 52027 ns 51880 ns 26682 18.7703GB/s BM_WriteRecordBatch/64/min_time:1.000/real_time 56324 ns 56202 ns 24724 17.3384GB/s BM_WriteRecordBatch/256/min_time:1.000/real_time 108096 ns 107868 ns 11284 9.03423GB/s BM_WriteRecordBatch/1024/min_time:1.000/real_time 279508 ns 278730 ns 4957 3.49386GB/s BM_WriteRecordBatch/4k/min_time:1.000/real_time 1013229 ns 1009772 ns 1191 986.944MB/s BM_WriteRecordBatch/8k/min_time:1.000/real_time 2011309 ns 2005377 ns 661 497.189MB/s BM_ReadRecordBatch/1/min_time:1.000/real_time 2507 ns 2501 ns 558949 389.563GB/s BM_ReadRecordBatch/4/min_time:1.000/real_time 5120 ns 5110 ns 275798 190.735GB/s BM_ReadRecordBatch/16/min_time:1.000/real_time 15800 ns 15706 ns 85481 61.8072GB/s BM_ReadRecordBatch/64/min_time:1.000/real_time 55678 ns 55476 ns 25022 17.5393GB/s BM_ReadRecordBatch/256/min_time:1.000/real_time 218083 ns 217596 ns 6163 4.47794GB/s BM_ReadRecordBatch/1024/min_time:1.000/real_time 875861 ns 873419 ns 1591 1.11497GB/s BM_ReadRecordBatch/4k/min_time:1.000/real_time 3545586 ns 3538141 ns 383 282.041MB/s BM_ReadRecordBatch/8k/min_time:1.000/real_time 7118830 ns 7104433 ns 194 140.473MB/s ``` Author: Wes McKinney <wes.mckinney@twosigma.com> Closes #479 from wesm/ARROW-737 and squashes the following commits: b663ca4 [Wes McKinney] Enable mutable buffer slices, SliceMutableBuffer function