This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
Hello,
in the document SPRAD45, https://www.ti.com/lit/an/sprad45/sprad45.pdf?ts=1672766415727&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FAM623we
The benchmark data was carried out with code optimization (Compiler and flags are set to -O3), this is shown in table 2-1.
Since this optimization is extreme, is there a possibility to get the measurements in table 2-1 Dhrystone Benchmarks with -O2 optimization?
Thanks,
Marwan
As a general note Dhrystone is a very old benchmark (1980's) and there is no certification body or organization that defines it. The nightly tests we run and benchmark are for performance accross the portfolio and result in ~3 DMIPS per MHz on any A53 core. With a modern compiler like GCC since about 2017 on any ARMv8-A you will get the best score with just flags "-march=ARMv8 -O3 ".
You can download (from for example https://github.com/Keith-S-Thompson/dhrystone) and build Dhrystone with other flags or disabling modern compiler features. Or maybe https://github.com/ARM-software/workload-automation/tree/master/wa/workloads/dhrystone/src can be an even simpler path to just modify. For this tiny program it is likely simplest to build and run on target. Just download the Makefile and dhrystone.c from the second link and type make -f makefile . Modify the flags from the Makefile to match what you are trying to measure.
In case it is not obvious all A53's score the same DMIPS/MHz using the same binary (or compiler). With Linux you can also just copy the binary and run.
Pekka