并行示例

现象

多物理内核的情况下（128个），单并行线程只有1个，程序运行时长较久。

调优思路

通过最基本的omp编程方式实现并行计算；因为在多核环境场景下，提升并行度是最直接的优化手段。

操作步骤

1. 运行矩阵行列大小为2048的parallel_matmult示例。

   ./matmul 2048 1

   返回信息如下：

   Size is 2048, Matrix multiplication method is: 1, Check correctness is: 0
   Initialization time = 0.175117s
   Matrix multiplication time = 2.971563s

   矩阵行列大小为2048情况下，并行计算耗时3秒左右。

2. 创建矩阵行列大小为2048的parallel_matmult示例的roofline任务。

   devkit tuner roofline -o parallel_matmult_2048 -m region ./matmul 2048 1

   返回信息如下：

   Note:
     1. Roofline task is currently only supported on the 920 platform.
     2. The application must be a binary file in ELF format.
     3. Roofline task collection needs to ensure the application has finished running.
     4. The estimated time of roofline collection is about 3 * application estimated time.
     5. You can learn about the roofline profiling method by looking at document /usr/local/devkit/tuner/docs/ROOFLINE_KNOW_HOW.MD
   RFCOLLECT: Start collection for ./matmul
   RFCOLLECT: Launch application to collect performance metrics of ./matmul
   Size is 2048, Matrix multiplication method is: 1, Check correctness is: 0
   Initialization time = 0.174164s
   ROOFLINE_EVENTS are initialized.
   Matrix multiplication time = 2.996051s
   RFCOLLECT: Launch application to do binary instrumentation of ./matmul
   Size is 2048, Matrix multiplication method is: 1, Check correctness is: 0
   Initialization time = 0.523171s
   Matrix multiplication time = 3.427321s
   RFCOLLECT: Launch benchmarks for measuring roofs
   RFCOLLECT: Processing all collected data
   RFCOLLECT: Result is captured at /matrix_multiplication/rfcollect-20240506-154009.json
   RFCOLLECT: Run "rfreport /matrix_multiplication/rfcollect-20240506-154009.json" to get report.
   
   Get roofline report ...
   The roofline json report: /matrix_multiplication/parallel_matmult_2048.json
   The roofline html report: /matrix_multiplication/parallel_matmult_2048.html

3. 查看parallel_matmult_2048_html报告。
   图2 parallel_matmult_2048_html报告
   

   此时获取的roofs的并行度为128，获取到Elapsed Time 2.953s, GFLOP Count 17.18，Performance 5.818 GFLOPS。

优化效果

在多核环境场景下，提升并行度是最直接的优化手段。