On 04/14/2016 04:23 PM, Piotr Masoń wrote:
> 
> Hi, during debugging of code compiled by means of
> OSELAS.Toolchain-2013.12.1/arm-cortexa8-linux-gnueabihf/gcc-4.8.3-glibc-2.18-binutils-2.24-kernel-3.12-sanitized
> for procesoor TI AM3558, I get a strange behavoiur for double to float
> conversion.
> 
> //cValue.getValue() returns double
> float floatValue=static_cast<float>(cValue.getValue());
> 
> and when cValue.getValue() = 5.0 then after conversion floatValue =
> 8.40779079e-45 and it seams, that the conversion doesn't work.
> 
> 
> when I only add a line with simple assigning to double variable at the
> same branch after the conversion line, after compilation, float values
> work fine.
> 
> 
> float floatValue=static_cast<float>(cValue.getValue());
> double doubleValue=cValue.getValue();
> 
> I get doubleValue = 5.0 and floatValue = 5.0
> 
> How compiler decides, which instructions of NEON or VFP choose, or does
> it use a "hard-float" at all in this case. Why it does work with an
> extra line of  code with assignig a float value to double value?

Can you try a newer compiler?
Can you give us a simple compiling test case that illustrates the problem?

Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |