Intel® Quartus® Prime Software
Intel® Quartus® Prime Design Software, Design Entry, Synthesis, Simulation, Verification, Timing Analysis, System Design (Platform Designer, formerly Qsys)
16611 Discussions

Questions Regarding Profiling of OpenCL Kernels

Altera_Forum
Honored Contributor II
‎11-15-2017 08:30 PM
1,035 Views

Hi, 

 

I have some basic questions regarding profiling metrics being reported after kernel execution. Let me mention, I have already took a look at Intel FPGA documentation, regarding profiling metrics, but some specific things are still vague for me. 

 

1) What exactly is the stall percentage, and which factors can cause stall in the pipeline? 

 

2) Which factors can play with occupancy factors? In cases with 0% stall and occupancy smaller than 100%, can we conclude there is an existence of overhead of scheduling threads on the compute units? 

 

3) The profiling shows execution time of the kernel, which is completely different from OpenCL event timing and the wall-clock execution time. What's the reason for that? 

 

Thanks, 

Saman
0 Kudos

Link Copied

1 Reply
Altera_Forum
Honored Contributor II
‎11-16-2017 01:24 AM
307 Views

Please check "Intel FPGA SDK for OpenCL Best Practices Guide, Section 4, Profiling Your Kernel to Identify Performance Bottlenecks". The answer to all of your questions and more is mentioned in that section.

0 Kudos
Copy link
Reply

Community support is provided during standard business hours (Monday to Friday 7AM - 5PM PST). Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.