roywei commented on issue #15429: Operator Performance Regression on CPU URL: https://github.com/apache/incubator-mxnet/issues/15429#issuecomment-508667518 @sandeep-krishnamurthy @Zha0q1 https://github.com/apache/incubator-mxnet/pull/15240 This PR introduced regression from 0619-0621. broadcast_add speed before this PR: 0.004, after this PR: 0.0047 Note speed is different than pip installed version due to build flags but the regression is there. build: `sudo make -j$(nproc) USE_MKLDNN=1` script: ``` import mxnet as mx from mxnet import nd mx.random.seed(10) from benchmark.opperf.utils.benchmark_utils import run_performance_test add_res = run_performance_test(nd.broadcast_add, run_backward=False, dtype='float32', ctx=mx.cpu(), inputs=[{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]} ], warmup=10, runs=500) print("speed:", add_res[0]['broadcast_add'][0]['avg_time_forward_broadcast_add']) ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
