floating point benchmarks

Tobyn VanVeghten

Hey,

I'm looking for some performance benchmarks for the Tiva with the optional floating point hardware unit. Ideally, I am looking for something like this link for the TI C6000 series:

http://www.ti.com/lsds/ti/dsp/c6000_dsp/c674x/benchmarks.page

Short of that, what is the range of clock cycles I can typically expect for calculating a single precision floating point multiply and accumulate (MAC? From this, I can extrapolate out to my required filters. At this point, I am looking at using the TI supplied compiler and RTOS (if I use an RTOS).

Thanks,

Tobyn

over 11 years ago

0 Petrei over 11 years ago

Guru 26105 points

Hi,

You can use the procedures shown in this thread.

Petrei

0 Tobyn VanVeghten over 11 years ago in reply to Petrei

Intellectual 700 points

I am not looking to make benchmarks of my code. I am trying to determine if the Tiva will work for my application and am looking for typical performance metrics for single precision floating point FIR filters.

Thanks,

Tobyn

0 Bobby Bradford over 11 years ago in reply to Tobyn VanVeghten

TI__Genius 10040 points

I am trying to determine if this information is available for the Cortex M4F (in Tiva) in combination with the compilers used in CCS. I will post a response back here as soon as I get more information.

--Bobby

0 f. m. over 11 years ago in reply to Bobby Bradford

Guru 11940 points

After my experience, the performance results are the same for all Cortex M4F, given identical compilers, settings and ABI - at least up to the point when Flash waitstates are kicking in. So general benchmarks like coremark might give you an idea. I tried a 2048-point FFT example from ARM, with less than 10 percent difference for the same core frequency (STM32F4, LM4F120).

Arm-based microcontrollers

Arm-based microcontrollers forum

floating point benchmarks