Hello,
While playing with the NXP RDB for the T1040, some colleagues and I performed some (not so) quick experiments to have an idea of its processing power: we ran some sort of benchmark code directly over U-Boot, in a single core, and compared it to other single-core processors we have.
Our code took some 212 microseconds to run in core 0 with all other cores disabled (226 if they were kept in spin-lock), which matched our expectations, but when we ran it in core 1 or 2 (disabling the other secondary cores), it took no less than 657 us. Even more surprisingly, core 3 took 701 us (i.e. it was always looking a bit slower than other secondary cores).
We ran the same code in a "not too different board" (an EPT1042xS, which had less peripherals but used a T1042 that looks very similar to the T1040) and the results looked less odd: running our code in core 0 took 273 us, and cores 1/2/3 took around 376 us. Since the NXP board has the T1040 running at 1.4GHz (vs. the T1042 @1.125GHz in the EP board) and its DDR and buses are also faster, we did expect the NXP board to perform faster and we also supposed the secondary cores would run our code slower than core 0 due to it being active, running U-Boot code, but we did not expect the secondary cores in the T1040D4-RDB to be so slow when compared to the primary one.
What are the causes for such timing discrepancies? Anything particular in the T1040D4-RDB U-Boot implementation? While my title states "unexpected", I admit it is unexpected only for beginners...
Thanks!
Best regards,
Ricardo
