Intel® Integrated Performance Primitives
Community support and discussions relating to developing high-performance vision, signal, security, and storage applications.
Announcements
This community is designed for sharing of public information. Please do not share Intel or third-party confidential information here.

Cannot reduce OpenMP threads

daven-hughes
Beginner
152 Views

I can't get good performance with ippsFir_32f() past the point where it starts using FFTs internally (CORRECTION: that is, past the point where FFTs start getting processed in parallel at order 13). I get about 80% wait time and it's all caused by _kmp_launch_worker threads. 

I've tried

- ippsSetNumThreads(1)
- kmp_set_blocksize(200) via dll import 

Yet I still see multiple kmp threads in Vtune and overall cpu usage is about 75% between 4 cores. What could I be doing wrong here?

0 Kudos
1 Reply
Chao_Y_Intel
Employee
152 Views
Hello, Could you attached a simple file that can show how the function is usded and some profiling result? For high CPU usage, can you add the following APIs, to reduced OpenMP BLOCK time? kmp_set_blocktime(0). Thanks, Chao
Reply