ARM Cortex M7 64 bits floating point

Charles G. wrote:
> Does the GCC compiler support the ARM Cortex M7 64 bits floating point?
> If so what are the switches an where can I find the documentation.


There is no "ARM Cortex M7 64 bits floating point".
There are only ARM Cortex M7 with ARM FPv4 (for example some STM32F7 
Devices) and ARM Cortex M7 with ARM FPv5 (for example some STM32F7 or 
STM32H7 devices) extension.

For Options of gcc look at gcc docs, of course:
https://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html

Look for armv7e-m, fpv5.

Charles G. wrote:
> I was under the impression that the ARM Cortex M7 support 32 and 64 bits
> hardware floating point

This may be the only M7 with 64 bits:

https://www.microchip.com/design-centers/32-bit/sam-32-bit-mcus/sam-s-mcus

But in any case Cortex A or R may be better suitable because of NEON

>>>>>>>
 We are having a product that does 32 bit
floating point using a different processor and ran into an accuracy
issue when trying to compute a 15 order polynomial.

Check this:

https://en.wikipedia.org/wiki/Horner%27s_method

This may help with accuracy issues in polynomials. Cranking up from 32 
to 64 bit is no cure in most of the cases. Been there, done that.

Cheers
Detlef

Lothar wrote:
> Charles G. wrote:
>> I was under the impression that the ARM Cortex M7 support 32 and 64 bits
>> hardware floating point
>
> This may be the only M7 with 64 bits:
>
> https://www.microchip.com/design-centers/32-bit/sam-32-bit-mcus/sam-s-mcus


No the STM32H7 Series also have the FPv5 extension for double precision 
and is faster than the Atmel.
http://www.st.com/en/microcontrollers/stm32h7-series.html?querycriteria=productId=SS1951

In the slower STM32F7 Series - which is slower - there are also some 
double precision devices.
http://www.st.com/en/microcontrollers/stm32f7-series.html?querycriteria=productId=SS1858