Floating Point MIPS

Technical Support
, , ,
1

Hi,
I have an issue to implement a software core which needs to do 16 floating-point MIPS and I don’t know which product supports floating-point instructions in this scale.

Cloud anyone please help me if there any chance to use colibri module for this purpose?

2

Dear @shaw.mehdi

Can you explain in more detail what you mean by 16 floating-point MIPS? All our Products have a floating point unit available most of them also support ARM NEON which is a DSP extension for ARM processors. E.g. have a look at the iMX6 processor which has floating point support as well as ARM NEON:
https://www.toradex.com/computer-on-modules/apalis-arm-family/nxp-freescale-imx-6

Best regards,
Stefan

3

Hi @stefan_e.tx, Thanks for your quick answer,

Let me explain the problem in detail; I want to deploy a core software which needs to execute 16000000 floating-point instructions and I don’t know if there is any module which supports this; and if so how much time the module needs to process these 16m instructions. actually the processing time for these instructions is very important for me.

Thanks in advance.

4

Dear @shaw.mehdi,

This shouldn’t be a problem normally. For example the Apalis iMX6 runs at 800-1000 MHz. Most operations only require one clock and even if they would require multiple clocks you are still way faster than 16e6 FLOPS.

But please keep in mind that all of our modules run on an Operating system. This means it may happen that a process gets interrupted by another process. So keep that in mind when designing your software.

Regards,
Stefan

5

Thank you @stefan_e.tx. as a result I can not be sure about the os and it’s better to develop this process on a hardware like fpga or dsp.