[Embench] Notes from Embench call 20 Apr 2020

Jeremy Bennett jeremy.bennett at embecosm.com
Fri May 8 21:20:21 CEST 2020

Hash: SHA1

Hi all,

These are the notes from last month's meeting, summarizing actions.

2. GitHub pull requests
- -----------------------

The following two pull requests have been merged:

- - Update ARC GCC target
- - Thanks to Claudiu Zissulescu for this patch

- - Remove redundant code from STM32F4 Discovery board support
- - Thanks to Paolo Savini for this patch

3. New GitHub issues
- --------------------

The following new issues have been raised.

- - Problems with spaces when setting tool flag options
- - This appears to be a bug - assigned to Paolo Savini to fix
- - Thanks to Mark Hill for raising the issue.

- - Incorrect reporting of code size.
- - This is a more comprehensive version of issue 40, with suggested fix.
- - Assigned to Paolo Savini to fix.
- - Thanks to Mark Hill for raising the issue.

- - Inconsistent speed results on VexRiscv core
- - This appears to be an issue with a non-bare-metal multi-core system
- - Some suggestions provided - an issue for future Embench
- - Thanks to genesys2 for raising the issue.

5. Update on benchmarks
- -----------------------

Work is in progress on the embench-iot-results repository. Goal is to
have this clean and ready to accept new submissions by the time of the
next meeting.

ACTION: Jeremy Bennett, Paolo Savini. Complete work in results repository.

Ofer Shinaar advised that Western Digital would start working on
adopting their enterprise build system for Embench later this year.

Ofer Shinnar advised that Western Digital hoped to take forward the
work on benchmarking interrupt latency and context switch performance
later this year. This is building on the initial work by Cesare Garlati.

A question was raised about where the reference platform (Cortex M4,
ST Discovery Board) could be obtained.

ACTION: Jeremy Bennett to publicise current vendors on the mailing
list, and update the documentation with more details to make the board
easier to find in future.

8. Publicity
- ------------

David Patterson has submitted an abstract to Hot Chips. This will look
at the latest results for a wider range of processors and also the
work by AdaCore on an Ada version of Embench.

There was a request for more coverage of Embench at RISC-V meetups.
This has already been done at the London meetup by Jeremy Bennett.

ACTION: Jeremy Bennett to update his slides and make them generally
available to the community for others to use.

10. AOB: Embench 0.6
- --------------------

There was a wide-ranging discussion of the requirements for Embench
0.6. Ideas suggested were:
- - a floating point suite
  - BEEBS has some that were rejected from current Embench
  - increasing the size of the benchmarks for FP platforms to say 1MB
  - can we add some machine learning benchmarks
  - use M4/M7 as baseline
- - easier configuration
- - extending the interrupt latency, context-switching benchmarks
- - getting rid of dummy libraries

ACTION: All to suggest benchmarks that could be used for a FP
benchmark suite.

9. Next meeting
- ---------------

We currently meet regularly on the third Monday of each month 8-10am
California time. Our future meetings will be
- - Mon 11 May 2020 (changed)
- - Mon 15 Jun 2020
- - Mon 20 Jul 2020
- - Mon 10 Aug 2020 (changed)
- - Mon 21 Sep 2020
- - Mon 19 Oct 2020
- - Mon 16 Nov 2020
- - Mon 14 Dec 2020 (changed)

Note that some dates have been changed to accommodate conferences and

I have a calendar entry for our regular attendees, to provide
automated reminders. Please ask if you wish to be added to this.

Our May videoconference is next Monday, 11 May at 08:00 California
time. I shall send out an agenda over the weekend.

Best wishes,




More information about the Embench mailing list